EPA 600/R-12/531 | May 2012 | www.epa.

gov/ord

EPA Traceability Protocol

for Assay and
Certification of Gaseous
Calibration Standards

Office of Research and Development

National Risk Management Research Laboratory
 EPA/600/R-12/531
May 2012

EPA Traceability Protocol for Assay

and Certification of Gaseous
Calibration Standards
by

Robert S. Wright
Air Pollution Prevention and Control Division
National Risk Management Research Laboratory
Research Triangle Park, NC 27711

National Risk Management Research Laboratory

Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
 EPA Traceability Protocol for Gaseous Calibration Standards

Notice

This document has been reviewed in accordance with U.S.

Environmental Protection Agency policy and approved for publication.
Mention of trade names or commercial products does not constitute
endorsement or recommendation for use.

Cover photo: Gas cylinders stored at the National Institute of Standards and
Technology laboratory in Gaithersburg, Maryland (photo courtesy of
National Institute of Standards and Technology)

Industrial facility at sunset (photo courtesy of Mikael Miettinen,

Göteborg, Sweden, http://www.flickr.com/photos/mikaelmiettinen/)

ii
 EPA Traceability Protocol for Gaseous Calibration Standards

Contents

Section 1 - Summary of Changes ......................................................................... 1

Section 2 - EPA Traceability Protocol for Assay and Certification of Compressed

Gas Calibration Standards .................................................................................... 5
2.1 General Information ................................................................................. 5
2.2 Procedure G1: Assay and Certification of a Compressed Gas
Calibration Standard without Dilution .............................................................. 37
2.3 Procedure G2: Assay and Certification of a Compressed Gas
Calibration Standard using Dilution ................................................................. 47
2.4 Procedure G3: Assay and Certification of a Zero Air Material as
Conforming to 40 CFR Part 72.2 ..................................................................... 63

Section 3 - EPA Traceability Protocol for Assay and Certification of Permeation

Device Calibration Standards ............................................................................. 71
3.1 General Information ............................................................................... 71
3.2 Procedure P2: Assay and Certification of Permeation Device Calibration
Standards based on a Compressed Gas Reference Standard ....................... 79
3.3 Procedure P3: Assay and Certification of Permeation Device Calibration
Standards based on a Mass Reference Standard ........................................... 89

Section 4 - EPA Traceability Protocol for Assay and Certification of Dynamic Gas
Dilution Systems ................................................................................................. 97
4.1 Procedure D1: Assay and Certification of a Dynamic Gas Dilution
System ............................................................................................................ 97

Section 5 - References ..................................................................................... 115

Appendix A - Statistical Spreadsheet for Procedures G1 and G2 ..................... 121

Appendix B - Statistical Spreadsheet for Procedure P3 .................................... 133

Appendix C - Statistical Spreadsheet for Stability Determination ...................... 139

Appendix D - Matrix Notation ............................................................................ 142

Appendix E - Statistical Spreadsheet for Procedure G3 ................................... 151

Appendix F - Statistical Spreadsheet for Procedure D1 .................................... 156

iii
 EPA Traceability Protocol for Gaseous Calibration Standards

Figures
Number Page

2-1 Example regression curve and confidence bands for multipoint 15

calibration

2-2 Apparatus for the assay of calibration standards without dilution 38

2-3 Two designs for assay apparatus with dilution (Procedure G2) 49

3-1 One possible design of the apparatus for the assay of 81

permeation device calibration standards referenced to a
compressed gas reference standard

3-2 Chamber for storing permeation tubes 91

3-3 Example of Appendix B spreadsheet graphic output 95

4-1 One possible design of a gas dilution system for the generation 99
of pollutant gas mixtures using dynamic dilution methods

Tables

Number Page

2-1 Summary of Compressed Gas SRMs, NTRMs, and RGMs that 8

are available from NIST and their Concentration Ranges

2-2 NIST and VSL Primary Gas Mixture Suites that are declared to 9
be Equivalent

2-3 Maximum Certification Periods for Calibration Standards in 31

Passivated Aluminum Cylinders

iv
 EPA Traceability Protocol for Gaseous Calibration Standards

Acknowledgments
This traceability protocol is a revision of an earlier protocol1 that was itself
revised in 1997. The original protocol was prepared in 1978. This revision was
reviewed by personnel of U.S. EPA, the National Institute of Standards and
Technology (NIST), the South Coast Air Quality Management District (AQMD),
and the Van Swinden Laboratorium (VSL, the national metrology institute of the
Netherlands) as it was being prepared. The author thanks the following
individuals for their assistance in providing technical information, reviewing this
revision of the protocol, and providing comments:

Lara Autry (EPA)

Thien Bui (EPA)
Raul Dominguez (AQMD)
Dave Duewer (NIST)
William Dorko (NIST)
Lyn Gameson (NIST)
Frank Guenther (NIST)
Michael Papp (EPA)
John Schakenbach (EPA)
Avraham Teitz (EPA)
Blaza Toman (NIST)
Rob Wessel (VSL)

Several specialty gas producers and other organizations provided useful

comments and suggestions for improving this revision of the protocol. The
author thanks the following organizations for their assistance in providing
technical information, reviewing this revision, and providing comments:

Airgas
Air Liquide
Applied Gas
Bios International
Environics
Linde
Matheson
NorLab
Praxair
Scott-Marrin
Takachiho

The spreadsheet modifications for this revision were made by Dr. Brian
Lopes and Dr. Kevin Ross, who were then working in the Statistics Department
of the University of North Carolina at Chapel Hill, and by Avraham Teitz of EPA
Region 2. The spreadsheets were first constructed by Dr. Michael Messner, who
was then working at RTI International.

v
EPA Traceability Protocol for Gaseous Calibration Standards

vi
 EPA Traceability Protocol for Gaseous Calibration Standards

Section 1 - Summary of Changes

The U.S. Environmental Protection Agency (EPA) in Research Triangle
Park, North Carolina has revised the 1997 version of its traceability protocol for
the assay and certification of compressed gas and permeation-device calibration
standards.1 The protocol allows producers of these standards, users of gaseous
standards, and other analytical laboratories to establish traceability of EPA
Protocol Gases to gaseous reference standards produced by the National
Institute of Standards and Technology (NIST). Parts 50, 58, 60, 72, and 75 of
Title 40 of the Code of Federal Regulations (CFR) require using Standard
Reference Materials (SRMs) or gaseous standards traceable to SRMs for
calibrating and auditing ambient air and stationary source pollutant monitoring
systems2-7.

This revision of the protocol has several major and minor changes from
the 1997 revision, including those listed below:

1. NIST-Traceable Reference Material Primes (NTRM*s) and Research

Gas Materials (RGMs) are acceptable analytical reference standards
for the assay of candidate standards (see Subsection 2.1.3);

2. Candidate standards can be made traceable to an expanded list of

NIST and VSL reference standards and the concentration ranges of
these reference standards have been expanded. This list includes
zero air materials (see Subsection 2.1.3);

3. The statistical test that is used to determine the stability of reactive

gas mixtures has been changed from Student's t-test to Schuirmann's
two one-sided tests (TOST), which will be used to determine if the
mean concentrations from the two assays differ by 1.0 percent or less.
TOST is used in the pharmaceutical industry to show analytical
method equivalency. It is superior to Student's t-test because it does
not allow substantial concentration differences to pass the statistical
significance test if measurement precision is poor. That is, being able
to find that a candidate standard's concentration is stable (i.e., within
the TOST acceptance criterion) is more important than not being able
to find that the concentration is unstable (see Subsection 2.1.5.2);

4. The uncertainty of the concentration of a candidate standard will now

be expressed as the expanded uncertainty (U) as described in the
Guide to the Expression of Uncertainty in Measurements (GUM),
published by International Organization for Standardization (ISO).
The calculated value of U must be given in the certification
documentation. The use of an uncalculated or blanket estimate of U
(e.g. +/- 1 percent) is not acceptable. (see Subsection 2.1.6);

1
 EPA Traceability Protocol for Gaseous Calibration Standards

5. The Excel spreadsheets associated with the protocol have been

updated to reflect the revisions to the protocol. If a specialty gas
producer wishes to use an equivalent statistical technique to calculate
the expanded uncertainty, the producer must first submit a detailed
description of the technique and any supporting software to EPA for
statistical evaluation and approval (see Subsection 2.1.6);

6. The specialty gas producer's Protocol Gas Verification Program

(PGVP) vendor identification number is required in the certification
documentation (see Subsection 2.1.7);

7. The certification expiration date is defined as the certification date plus

the certification period plus one day and is required in the certification
documentation (see Subsection 2.1.7);

8. Maximum certification periods for certified standards have been

extended in Table 2-3. Specialty gas producers may elect to certify
candidate standards for less than these periods if they believe that
they cannot prepare standards whose stability attains the maximum
certification period. The default certification period is that given in
Table 2-3. Each producer has discretion in this matter. (see
Subsection 2.1.9);

9. If the time period between the initial and second assays is at least 6
months and if the mean concentrations from the two assays are
demonstrated to be within 1.0 percent of each other using TOST,
candidate standards whose concentrations are lower than the
concentration range given in Table 2-3 may be certified for the full
certification period given in Table 2-3 rather than only six months as
was specified in earlier versions of the protocol (see Subsection
2.1.9);

10. Standards may be recertified if the cylinder pressure after the assays
have been completed is greater than 100 psig (see Subsection
2.1.11);

11. The minimum cylinder pressure has been changed to 100 psig (see
Subsection 2.1.12);

12. The PGVP is the primary accuracy assessment for EPA Protocol
Gases. EPA air pollution monitoring regulations specify that the
regulated community must use EPA Protocol Gases from PGVP
participants (see Subsection 2.1.13);

13. A new procedure and an associated spreadsheet have been prepared

for the assay and certification of a compressed gas zero air material

2
 EPA Traceability Protocol for Gaseous Calibration Standards

as conforming to 40 CFR Part 72.2 (see Subsection 2.4). At this time,

EPA does not require the regulated community to use NIST-traceable
zero air materials for the calibration of ambient air or continuous
emission monitors that are required by 40 CFR Parts 50, 58, 60, and
75. However, end users may elect to purchase these standards from
specialty gas producers who elect to assay and certify them;

14. Section 3 concerns the assay and certification of permeation device

reference standards. It is largely unchanged from the 1997 version
although some wording and reference citations have changed.
However, Procedure P1 for comparison of a candidate standard to a
permeation device reference standard has been omitted because
such reference standards are no longer sold by NIST.

15. A new procedure has been written and a new spreadsheet has been
prepared for the assay and certification of dynamic gas dilution
systems (see Section 4). At this time, EPA does not require the
regulated community to use NIST-traceable dynamic gas dilution
systems for the calibration of ambient air or continuous emission
monitors that are required by 40 CFR Parts 50, 58, 60, and 75.
However, end users may elect to use these systems for calibrations.

The producers of reference standards that are assayed and certified

under this protocol are allowed one year after the publication of this revision of
the protocol to implement the changes that have been made to this protocol.
After the year has elapsed, producers must follow the procedures given in this
revision of the protocol if they wish to sell reference standards as "EPA Protocol
Gases". They must also participate in the PGVP (see Subsection 2.1.13).

3
 EPA Traceability Protocol for Gaseous Calibration Standards

Section 2 - EPA Traceability Protocol for Assay and

Certification of Compressed Gas Calibration Standards
2.1 General Information
2.1.1 Purpose and Scope of the Protocol

Use of this protocol is mandatory for certifying the calibration gases being
used for the calibration and audit of ambient air quality analyzers and continuous
emission monitors that are required by 40 CFR Parts 50, 58, 60, and 75 2-5. This
protocol describes three procedures for assaying candidate standards (i.e., not-
yet-certified standards) and for certifying that their concentrations are traceable
to a National Institute of Standards and Technology (NIST) compressed gas
reference standards [i.e., Standard Reference Materials (SRMs), NIST-Traceable
Reference Materials3 [NTRMs, including NTRM-Primes (NTRM*s) for which NIST
analyzes every sample in the production lot], and NIST Research Gas Materials
(RGMs)]. It may be used to assay and certify only candidate standards that have
the same components and concentration ranges as NIST-traceable reference
standards. A multiple-component candidate standard may be assayed and
certified under this protocol if NIST-traceable reference standards that contain
the same individual components as are in the candidate standard exist.

This protocol may be used by specialty gas producers, end users, or other
laboratories to assay candidate standards. The assay involves the direct
comparison of the candidate standards to NIST-traceable reference standards
without dilution (i.e., Procedure G1) or the indirect comparison of the candidate
standards to reference standards with dilution (i.e., Procedure G2). A candidate
standard having a concentration that is lower or higher than that of the reference
standard may be certified under this protocol if both standards' concentrations (or
diluted concentrations) fall within the well-characterized region of the analyzer's
calibration curve. This protocol places no restrictions on cylinder sizes and the
same analytical procedures must be used in assays of all cylinder sizes.

Standards that are certified under this protocol must remain in the
cylinders in which they were originally assayed. Unassayed standards in
cylinders that have been transfilled from cylinders assayed under this protocol
cannot be certified as being EPA Protocol Gases

This protocol may be used to assay and certify zero air materials.
Certified impurity concentrations can be certified using Procedures G1 or G2.
Procedure G3 can be used to certify that zero air materials conform to the purity
specifications of 40 CFR Part 72.2 (i.e., concentrations of SO2, NOX, and THC
are not greater than 0.1 ppm; whose concentration of CO is not greater than1
ppm; and whose concentration of CO2 is not greater than 400 ppm)7.

5
 EPA Traceability Protocol for Gaseous Calibration Standards

2.1.2 Using the Protocol

The assay/certification protocol described here is designed to minimize

both systematic and random errors in the assay process. Therefore, the protocol
must be carried out exactly as it is described. The assay procedures in this
protocol include one or more possible designs for the assay apparatus. The
analyst is not required to use these designs and may use alternative components
and configurations that produce equivalent-quality measurements. The protocol
is like a basic recipe that cooks will follow while substituting different ingredients.

2.1.3 Reference Standards

The EPA monitoring regulations define a "traceable" standard as one that

has been compared and certified, either directly or via not more than one
intermediate standard, to a primary standard such as an SRM or a Certified
Reference Materials (CRM).3,4 The monitoring regulations require that calibration
gases used for calibration and audit of ambient air quality analyzers and
continuous emission monitors be traceable to NIST-traceable reference
standards, which are listed in Table 2-1.

NIST's Policy on Metrological Traceability (see www.nist.gov/traceability/)

has adopted the international definition of metrological traceability: "property of a
measurement result whereby the result can be related to a reference through a
documented unbroken chain of calibrations, each contributing to the
measurement uncertainty." This protocol extends metrological traceability to
standards that are assayed and certified by maintaining the unbroken chain of
calibrations to NIST reference standards, by documenting the assay procedure
and the assay results, and by estimating the uncertainty of certified standards.

In 2010, NIST and VSL (Van Swinden Laboratorium, the National

Metrology Institute of the Netherlands) issued a joint Declaration of Equivalence
(DOE, see http://www.vsl.nl/files/DOE_Scan.pdf) that specific VSL Primary
Reference Materials (PRMs) and VSL CRMs are equivalent to the corresponding
SRMs within stated uncertainties. The current SRM-equivalent PRMs and CRMs
that are available from VSL are listed in Table 2-2. Other gas mixtures may be
added to the DOE in the future. PRMs or CRMs from another national metrology
organization will be considered equivalent to SRMs when a DOE is issued jointly
by NIST and the organization. PRMs and CRMs that are not declared to be
equivalent to SRMs cannot be used as reference standards under this protocol.
The generic terms "PRM" and "CRM" are used to refer to any SRM-equivalent
standard that is listed in a DOE.

Comparison of a candidate standard directly to an SRM, an NTRM, an

RGM, a PRM or a CRM is preferred and recommended. However, the use of a
Gas Manufacturer's Intermediate Standard (GMIS) (see Subsection 2.1.3.1) in
the comparison is permitted. A GMIS is an intermediate reference standard that

6
 EPA Traceability Protocol for Gaseous Calibration Standards

has been compared directly to an SRM, an NTRM, an RGM, a PRM or a CRM

according to Procedure G1. It is an acceptable reference standard for the assay
of candidate standards. However, purchasers of standards that have been
compared to a GMIS should be aware that, in conformity with the above
definition, such a standard could only be used directly for calibration or audit.
Such a standard could not be used as a second-generation intermediate
reference standard to assay other candidate standards.

Accordingly, the reference standard used for assaying and certifying a

candidate standard under this protocol must be an SRM, an NTRM, an RGM, a
PRM, a CRM or a GMIS. A summary of the compressed gas SRMs, NTRMs,
and RGMs that are certified by NIST is given in Table 2-1. A summary of the
PRMs and CRMs that are equivalent to NIST reference standards is given in
Table 2-2. The reference standard must be within its certification period.

Flow rate reference standards or gas dilution systems must be traceable

to NIST flow rate, volume, mass or time reference standards by original
calibration at a NIST-accredited state weights and measures laboratory or at a
testing and calibration laboratory that is accredited by the National Voluntary
Laboratory Accreditation Program (NVLAP), the American Association for
Laboratory accreditation (A2LA) or by the International Laboratory Accreditation
Conference (ILAC) under ISO/IEC 17025 (General Requirements for the
Competence of Testing and Calibration Laboratories).10,11, 12 They must be
recertified on an annual basis using NIST-traceable reference standards
although this recertification may be performed in the analytical laboratory in
which the candidate standards are assayed. These volume reference standards
are required for assays using Procedure G2 (see Subsection 2.3.4).

Alternatively, flow rate reference standards or gas dilution systems may

be traceable to the flow rate, volume, mass or time reference standards of other
national metrology institutes (e.g., KRISS in South Korea, NEL in the United
Kingdom, PTB in Germany) provided that these institutes are participants in the
Working Group for Fluid Flow (WGFF) of the International Committee for Weights
and Measures (CIPM).

7
 EPA Traceability Protocol for Gaseous Calibration Standards

TABLE 2-1. Summary of Compressed Gas SRMs, NTRMs, and RGMs that are
available from NIST and their Concentration Rangesa
Certified Balance SRM b range NTRMb range RGMb range
Certification
component gas period
(years)
Ammonia Nitrogen Not available Not available 5 to 50 ppm 1
Carbon dioxide Air 390 to 400 ppm 360 to 420 ppm 360 to 420 ppm 8
Carbon dioxide Nitrogen 500 ppm to 16% 5 ppm to 20% Contact NIST 8
Carbon
Air 60 to 160 ppb 50 to 500 ppb 40 to 500 ppb TBDc
monoxide
Carbon
Air 10 to 45 ppm 10 to 45 ppm Contact NIST 8
monoxide
Carbon
Nitrogen 10 ppm to 13% 101 ppm to 15% Contact NIST 8
monoxide
Formaldehyde Nitrogen Not available Not available 0.5 to 10 ppm 1
Hydrogen
Nitrogen Not available Not available Contact NIST TBD
chloride
Hydrogen
Nitrogen 5 to 20 ppm 1 to 400 ppm Contact NIST 3
sulfide
Methane Air 1 to 100 ppm 1 to 1000 ppm 1.7 to 2.2 ppm 8
Methane Nitrogen Not available 0.5 ppm to 10% Contact NIST 8
Methanol or Nitrogen
Not available Not available 75 to 500 ppm 4
ethanol or Air
Natural gas Natural
Not Available Contact NIST Contact NIST 4
componentsd gas
Nitric oxide Nitrogen 0.5 to 50 ppm 0.5 to 50 ppm Contact NIST 3
Nitric oxide Nitrogen 50 to 3000 ppm 50 to 3000 ppm Contact NIST 8
Nitrous oxide Air TBD 300 ppb to 5% 300 to 350 ppb 8
Total oxides of 3
Air 100 ppm 10 to 100 ppm 3 to 100 ppm
nitrogen (NOX)e (SRM 6)
Oxygen Nitrogen 2 to 21 % 0.4 to 25 % Contact NIST 8
Propane Air 0.1 to 500 ppm 0.1 to 500 ppm Contact NIST 8
Propane Nitrogen 100 to 2000 ppm 5 ppb to 2% Contact NIST 8
Sulfur dioxide Nitrogen 5 ppm 5 to 50 ppm Contact NIST 4
Sulfur dioxide Nitrogen 50 to 3500 ppm 50 to 5000 ppm Contact NIST 8
Volatile organics Nitrogen 5 ppb Contact NIST Contact NIST 4
Zero air
Air Not available Not available Contact NIST TBD
materialf
a
All SRMs may not be available at all times. Other SRMs may be developed in the future and
could be used as reference standards. Contact NIST for information about SRM availability.
b
Concentrations are by mole; ppb = parts per billion; ppm = parts per million.
c
To be determined.
d Natural gas components are methane, ethane, propane, n-butane, iso-butane, n-pentane, iso-
pentane, helium, nitrogen, and carbon dioxide.
e
NIST defines its total NOX standards as containing nitrogen dioxide plus contaminant nitric acid.
f
Concentrations of SO2, NOX, and total hydrocarbons (THC) are not >0.1 ppm; concentration of
CO is not >1 ppm; and concentration of CO2 is not >400 ppm as per 40 CFR Part 72.2.

8
 EPA Traceability Protocol for Gaseous Calibration Standards

TABLE 2-2. NIST and VSLa Primary Gas Mixture Suites that are declared to be
Equivalent
Certified Balance gas Concentrationb Maximum Stability period
component range for gas allowable (years)
mixture suite differencec
Carbon
Nitrogen 10 ppm to 20 % 0.3 % relative 3
dioxide
Carbon
Air 100 to 500 ppm 0.5 % relative 3
dioxide
Carbon Nitrogen or
1 ppm to 10 % 0.3 % relative 3
monoxide Air
Nitrogen or
Ethanol 75 to 500 ppm 0.5 % relative 3
Air
Hydrogen
Nitrogen 10 to 1000 ppm 1.0 % relative 2 or 3f
sulfide
0.5 % relative
d
Natural gas Nitrogen Typical (0.3 % for 3
CH4)
Nitric oxide Nitrogen 0.5 ppm to 1 % 0.5 % relative 2 or 3f
Nitrogen Nitrogen or
10 ppm to 1 % 0.5 % relative 2
dioxide Air
Oxygen Nitrogen 10 ppm to 25 % 0.2 % relative 3
Nitrogen or
Propane 1 ppm to 1 % 0.3 % relative 1, 2 or 3f
Air
Sulfur dioxide Nitrogen 1 ppm to 1 % 0.5 % relative 2 or 3f
Volatile
organic Nitrogen 1 ppb to 1 ppm 2 % relative 2
compoundse
a
Information about Reference Gas Mixtures can be obtained from:
VSL [i.e., the Van Swinden Laboratorium, the National Metrology Institute of the Netherlands]
Thijsseweg 11, 2629 JA Delft, NL P.O. Box 654, 2600 AR Delft NL
Telephone: 31 (0) 15 269 1550 FAX: 31 (0) 15 261 2971
E-mail: [email protected] Website: www.vsl.nl
b
Within the listed ranges, any concentration is available. PRMs are prepared individually in 5-L
cylinders according to ISO Standard 6142 (Gas Analysis–Preparation of calibration gas mixtures-
weighing methods). After preparation, the composition is verified against VSL Primary Standard
Gas Mixtures. CRMs are available in larger size cylinders and are gravimetrically prepared by an
accredited supplier. They are certified by VSL against VSL Primary Standard Gas Mixtures.
c
Maximum allowable difference between NIST and VSL primary standard gas mixture suites.
d
Methane, ethane, propane, n-butane, iso-butane, n-pentane, iso-pentane, 2,2-dimethylpropane
(neopentane), n-hexane, carbon dioxide, and helium.
e
Ethane, ethene, propane, propene, iso-butane, iso-butene, 1-butene, n-butane, 2-methyl
butane, iso-pentane, n-pentane, 1-pentene, 1,3-butadiene, trans-2-pentene, 2-methyl pentane,
2,2,4-trimethyl pentane, n-hexane, n-heptane, benzene, toluene, n-octane, and o-xylene.
f
Stability period is dependent on the concentration of the PRM/CRM.

9
 EPA Traceability Protocol for Gaseous Calibration Standards

2.1.3.1 Gas Manufacturer's Intermediate Standard—

A GMIS is a compressed gas calibration standard that has been assayed
by direct comparison to an SRM, an NTRM, an RGM, a PRM or a CRM, that has
been assayed and certified according to Procedure G1, and that also meets the
following requirements:

1. A candidate GMIS must be assayed on at least three separate dates that are
uniformly spaced over at least a 3-month period. During each of these
assays, the candidate GMIS must be measured at least three times. All these
assays must use the same SRM, NTRM, RGM, PRM or CRM as the
reference standard to avoid errors associated with the use of different
reference standards for different assays.

2. For each assay, the analyst must calculate the mean concentration and U for
the three or more measurements of the candidate GMIS according to the
statistical procedures described in Appendix A or an EPA-approved
equivalent statistical technique (see Subsection 2.1.6). The value of U must
be less than or equal to 1.0 percent of the mean concentration.

3. After the three or more assays have been completed, the analyst must
calculate the overall mean concentration and U for the candidate GMIS using
the spreadsheet described in Appendix C or an EPA-approved equivalent
statistical technique (see Subsection 2.1.6). If an EPA-approved equivalent
statistical technique is used, it must be identified as such in the certificate
documentation and it must be described in media readily accessible to end
users.

4. If the mean estimated concentrations from individual assays differ by less

than 1.0 percent using Schuirmann's two one-sided test (TOST)13-18, the
candidate GMIS can be considered to be stable and can be used as a
reference standard for assays of candidate standards. If the TOST
acceptance criterion is not attained, the candidate GMIS may be unstable or
there may be analytical problems associated with the assays or the reference
standards. The analyst must either disqualify the candidate GMIS or
investigate why the TOST acceptance criterion is not attained. The analyst
may discard the data from a questionable assay and then conduct another
assay. The candidate GMIS can be used as a reference standard if it is
stable, but it cannot be used if it appears to be unstable.

5. The certification period for a GMIS is the same as for an EPA Protocol Gas of
the same composition and concentration.

6. A GMIS must be recertified after its original certification period ends. Three
or more discrete measurements of the candidate standard must be made
during a single recertification assay. Use the spreadsheet described in
Appendix C or an EPA-approved equivalent statistical technique (see

10
 EPA Traceability Protocol for Gaseous Calibration Standards

Subsection 2.1.6) to compare the mean concentration from the recertification

assay with the mean concentration from the previous assays. If the mean
concentrations differ by less than 1.0 percent using TOST, the GMIS can be
recertified. If the reassayed GMIS fails to meet the TOST acceptance
criterion, it must undergo a full certification as described in Step 1 above
before it can be used again. There is no requirement that the same reference
standard must be used in the original assays and the recertification assay, but
this practice is desirable if possible.

2.1.3.2 Reference Standards for Multipoint Calibrations—

The reference standards for the multipoint calibration must be diluted or
undiluted SRMs, RGMs, PRMs, CRMs, NTRMs, or GMISs (see Subsection
2.1.3) or dynamically diluted pure gases. Pure gases may be dynamically diluted
to prepare gas mixtures for use in multipoint calibrations, but such mixtures may
not be used as the reference standards for the span gas check or for the assay
of the candidate standard. Pure gases may not be diluted by more than a factor
of 100. Information concerning this standard (e.g., cylinder identification number,
certified concentration, expanded uncertainty, certification expiration date,
cylinder pressure, etc.) must be recorded in the laboratory's records.

2.1.3.3 Reference Standards for Span Gas Checks—

The reference standard for the span gas checks must be a diluted or
undiluted SRM, RGM, PRM, CRM, NTRM, or GMIS as specified in Subsection
2.1.3. The reference standard for the span gas check need not be the same as
one of those used for the multipoint calibration or for the assay of the candidate
standard. Information concerning this standard (e.g., cylinder identification
number, certified concentration, expanded uncertainty, certification expiration
date, cylinder pressure, etc.) must be recorded in the laboratory's records.

2.1.3.4 Reference Standards for Assay of Candidate Standards—

The reference standard used for the assay of the candidate standard must
be an SRM, an RGM, a PRM, a CRM, an NTRM or a GMIS. This standard need
not be the same as any of the reference standards used for the span gas check
or for the multipoint calibration. Information concerning the reference standard
(e.g., cylinder identification number, certified concentration, expanded
uncertainty, certification expiration date, cylinder pressure, etc.) must be
recorded in the laboratory's records.

2.1.3.5 Zero Gas for Multipoint Calibrations and Zero Gas Checks—
Zero gas used for multipoint calibrations, zero gas checks or for dilution of any
candidate or reference standard must be clean, dry, zero-grade air or nitrogen
containing no detectable concentration of the pollutant of interest. It may come
from compressed gas cylinders or from zero gas generators. The use of NIST-
traceable zero air material is recommended, but not required. The zero gas must
match the balance gas in the candidate standard and the reference standard,
unless it has been demonstrated that the pollutant gas analyzer is insensitive to

11
 EPA Traceability Protocol for Gaseous Calibration Standards

differences in the balance gas composition. It also must contain no contaminant

that causes a detectable response in the analyzer or that suppresses or
enhances the analyzer's response to the pollutant. The oxygen content of zero
air must be approximately that of ambient air, unless it has been demonstrated
that varying the oxygen content does not suppress or enhance the analyzer's
response. The water vapor concentration in the zero gas must be less than 5
ppm. Information concerning the zero gas (e.g., cylinder identification number,
cylinder pressure, etc.) must be recorded in the laboratory's records.

For analyzers such as gas chromatographs, the analyst may suspect that
the zero gas reading may not accurately represent the zero-intercept of the
calibration equation. The analyst may substitute a low-concentration, NIST-
traceable reference standard for the zero gas, providing that the concentration of
this standard is less than the concentration of the candidate standard.

2.1.3.6 Zero Gas for Assay of Candidate Standards and Zero Air Material—
The zero gas used for the assay of candidate standards need not be the
same zero gas as used for the multipoint calibrations and zero gas checks, but it
must conform to the specifications in Subsection 2.1.3.5. The zero gas used for
the assay of zero air material must be an NIST-traceable zero air material.
Information concerning the zero gas (e.g., cylinder identification number, cylinder
pressure, etc.) must be recorded in the laboratory's records.

2.1.3.7 Minimum Cylinder Pressure for Reference Standards

A reference standard must not be used when its gas pressure is below 0.7
megapascals (i.e., 100 psig). NIST has found that some gas mixtures (e.g., nitric
oxide in nitrogen) have exhibited a concentration change when the cylinder
pressure fell below this value. There is no minimum cylinder pressure
specification for zero gas.

2.1.3.8 Recertification of Reference Standards—

Recertification requirements for SRMs, NTRMs, and RGMs are specified
by NIST. Recertification requirements for PRMs and CRMs are specified by
VSL. See Subsection 2.1.3.1 for GMIS recertification requirements.

2.1.4 Analyzer Calibration

2.1.4.1 General Analyzer Calibration Requirements—

The assay procedures described in this protocol employ a data reduction
technique to calculate the concentration of a candidate standard that corrects for
minor analyzer calibration variations (i.e., drift). This technique does not require
the absolute accuracy of the analyzer's calibration curve at the time of the assay.
The analyzer must: (1) have a calibration curve that is well-characterized for the
pollutant of interest (see Subsection 2.1.4.2); (2) have good resolution and low
noise; and (3) have a calibration that is known and that is reasonably stable or
recoverable during the assay session.

12
 EPA Traceability Protocol for Gaseous Calibration Standards

2.1.4.2 Analyzer Multipoint Calibration—

The analyzer used for the assay must have had a multipoint calibration
within 1 month prior to the assay date. When the assay is on the same day as
the calibration, the calibration data are directly used to calculate the certified
concentration and expanded uncertainty of the candidate standard using the
spreadsheets described in Appendix A or using an EPA-approved equivalent
statistical technique (see Subsection 2.1.6). When the assay date is different
from the calibration date, these data are not used to directly calculate the
certified concentration although they are used to estimate the expanded
uncertainty. Zero gas and reference standard measurements on a different
assay date are compared to the multipoint calibration data (see Subsections
2.2.6.4 and 2.3.8.4). If only minor calibration drift has occurred since the
multipoint calibration, these measurements are used to calculate the certified
concentration of the candidate standard. If excessive calibration drift has
occurred, the multipoint calibration must be repeated.

The analyzer's zero and span controls may be adjusted before the start of
the multipoint calibration. If a zero or span adjustment is made, allow the
analyzer to stabilize for at least one hour before beginning the multipoint
calibration. The waiting period is necessary because some analyzers'
calibrations drift for a period of time following a zero or span control adjustment.

The multipoint calibration must consist of one or more measurements of

the analyzer responses to at least five different concentrations. The use of an
NIST-traceable zero air material in the calibration is recommended, but is not
required (see Section 2.1.3.3). Record these measurements and the analyzer's
zero and span control settings in the laboratory's records. These calibration
concentrations should be approximately evenly spaced over the concentration
range. The concentrations may be produced by undiluted reference standards or
by dilution of reference standards using a gas dilution system. See Subsection
2.1.3.2 for reference standard requirements.

If a gas dilution system is used in the assay apparatus, it must have a

specified accuracy of no worse than 1.0 percent of the undiluted reference
standard concentration. Additionally, the gas dilution system must be checked
by the analyst at monthly intervals to verify that its calibration has not drifted
significantly since its last calibration or recertification. Use an NIST-traceable
flow rate reference standard to check at least one flow rate setting for each
pollutant and dilution gas stream in the assay apparatus. Record the indicated
and actual flow rates in the laboratory's records. Calculate the relative difference
(in percent) between the indicated and actual volumes or flow rates. That is,

Relative Difference = 100 (Indicated Flow Rate - Actual Flow Rate)

(Actual Flow Rate)

13
 EPA Traceability Protocol for Gaseous Calibration Standards

If the relative differences for the pollutant and diluent flow rates are less than 1.0
percent, the calibrations of the gas dilution system has not drifted significantly
since its last calibration or recertification and the pollutant gas analyzer
calibration can proceed. If the relative differences for either the pollutant flow
rate or the diluent flow rate exceed 1.0 percent, significant drift has occurred and
the device must be recertified before the assay is conducted.

If the analyzer has multiple concentration ranges, a multipoint calibration

must be done for all ranges that will be used later for the assay of candidate
standards. A multipoint calibration that is conducted on one range is not valid for
an assay that is conducted on another range.

Data from the multipoint calibration must be evaluated using least-

squares regression analysis.19 This analysis technique will be used to determine
the analyzer's calibration curve and to characterize the uncertainty associated
with the calibration. The concentration values are the independent (i.e., X)
values in the analysis and their units may be parts per million, mole percent, or
any other appropriate units. The analyzer response values are the dependent
(i.e., Y) values in the analysis and their units may be volts, millivolts, percent of
scale or any other measurable analyzer response units. The analyzer response
values must have a resolution that is less than or equal to 1 percent of the
maximum measured analyzer response.

Because an analyzer’s response has a random error component,

repeated measurements of the same reference standard will not produce
identical analyzer responses. The analyst may investigate the analyzer’s
precision by making replicate measurements of reference standards at different
concentrations. Least-squares regression analysis is normally conducted under
the assumption that the precision is the same at all concentrations. However,
this statistical assumption may not be true for some real-world analyzers and the
analyst may need to use alternate statistical procedures to analyze the multipoint
calibration data. Performing replicate measurements at different concentrations
and calculating the standard deviation of the analyzer responses at each
concentration may allow the analyst to assess whether the precision is the same
at all concentrations.

Calculate the least-squares regression coefficients of the calibration

equation using the spreadsheets described in Appendix A or using an EPA-
approved equivalent statistical technique (see Subsection 2.1.6). If an EPA-
approved equivalent statistical technique is used, it must be identified as such in
the certificate documentation and it must be described in media readily
accessible to end users.

The spreadsheets allow the multipoint calibration data to be fitted to

straight-line, quadratic, cubic, or quartic linear regression models. EPA
discourages the use of the cubic and quartic models and believes that better fits

14
 EPA Traceability Protocol for Gaseous Calibration Standards

of the data can be obtained by performing multipoint calibrations over more

limited concentration ranges and by using straight-line or quadratic models.
Inclusion of cubic and quartic models in the spreadsheets is for experimental use
or for situations in which there is a theoretical basis for the use of such higher-
order models. Analysts should be aware that apparent higher-order calibration
curves may be caused by artifacts such as inaccurate reference standards or
leaks in a gas dilution system. They should not use higher-order regression
models to fit multipoint calibration data that have inadequate precision and that
should be fitted to lower-order regression models. Additionally, a multipoint
calibration should not change orders from one month to the next.

The spreadsheet described in Appendix A will suggest the best regression

model for the multipoint calibration data, but the analyst is expected to choose
the model that best fits the measurement process on theoretical grounds.

Plot the values from the multipoint calibration and the regression curve
with confidence bands as shown in Figure 2-1. These plots will provide a
graphical representation of the calibration and will permit a qualitative
assessment of the uncertainty associated with the calibration. Record the
regression coefficients, their expanded uncertainties, and other statistical results
from the spreadsheet in the laboratory's records. Enough information should be
recorded so that the analyst can calculate the expanded uncertainty of the
certified concentration of a candidate standard that is assayed at some later date
and to reconstruct these calculations at some later date should questions arise
about the calibration data or the spreadsheet calculations.

Figure 2-1. Example regression curve and confidence bands for multipoint calibration

However, a quantitative assessment of the calibration's uncertainty is

needed to allow the analyst to determine whether the multipoint calibration data
adequately characterizes the "true" calibration curve for the analyzer. The

15
 EPA Traceability Protocol for Gaseous Calibration Standards

criterion to be used to evaluate the uncertainty of the multipoint calibration is U

for a concentration predicted from the regression line using measured values of
the analyzer response. This value can be calculated using the spreadsheets
described in Appendix A or using an EPA-approved equivalent statistical
technique. Record the uncertainty calculations in the laboratory's records.

All measurements of candidate standards must fall within the well-

characterized region of the analyzer's calibration curve, which lies between the
largest and smallest measured concentrations of the multipoint calibration and for
which U for the regression-predicted analyzer response is ≤±1 percent of the
measured response for the largest concentration in the calibration. For example,
assume that a calibration was conducted between 0 and 100 ppm and that the
measured responses ranged between 0 and 10 volts. The example calibration is
well-characterized for all concentrations between 0 and 100 ppm for which U is
≤±0.1 volt (i.e., ±1 percent of 10 volts). Step 4 of the spreadsheet described in
Appendix A allows the analyst to enter various concentrations and obtain the
corresponding regression-predicted analyzer response and confidence limits.

In effect, U is a measure of how well the multipoint calibration data fit an

equation which the analyst assumes is the "true" calibration equation for the
analyzer. Comparison of values of U from straight-line and quadratic equations
permits the analyst to select the equation that best represents the data.

A multipoint calibration may fail to meet this uncertainty criterion for

several possible reasons:

• inadequate analytical precision;

• inaccuracy of the reference standards or the gas dilution system; or
• excessive uncertainty in the analyzer's calibration equation due to too
few measurements (either too few replicate measurements at the
same concentration or too few different concentrations in the
calibration) or to incorrect assumptions about the form of the equation.

The effect of inadequate analytical precision can be reduced by

increasing the number of replicate measurements at each calibration
concentration or by increasing the number of different concentrations used in the
multipoint calibration. Additionally, precision can be improved by using an
averaged analyzer response, rather than an instantaneous analyzer response,
for each measurement. The most accurate reference standards that are
available should be used. An inaccurate gas dilution system can be detected by
comparing measurements of the concentration of a diluted reference standard to
the theoretically equal concentration of another, undiluted reference standard. It
can also be detected by comparing measurements of two theoretically equal
concentrations obtained by dilution of two reference standards having
significantly different concentrations. An inaccurate gas dilution system must not
be used for the multipoint calibration. The effect of excessive uncertainty in a

16
 EPA Traceability Protocol for Gaseous Calibration Standards

straight-line calibration equation can be eliminated by using a quadratic

calibration equation or by transforming the calibration data mathematically so that
they may be fitted to a straight line regression equation. See Reference 16 for a
discussion of such linearizing transformations of data.

Note that possibly a more restrictive uncertainty criterion applies for the
assay of the candidate standard. The value of U for the estimated concentration
of the candidate standard must be ≤±1 percent of the concentration of the
reference standard (see Subsections 2.2.2 and 2.3.2). For example, assume
that (1) an analyzer's calibration equation is well characterized from 0 to 100 ppm
and (2) a 70-ppm candidate standard is being assayed at a later date using a 50-
ppm reference standard. Then, U must be ≤±0.5 ppm, rather than ≤±1.0 ppm
which is derived from the multipoint calibration.

2.1.4.3 Uncertainty of the Calibration Curve—

The data reduction technique used in this protocol is based on the
assumption that the analyzer has a well-characterized calibration curve. The
accuracy of the certified concentration of a candidate standard is dependent
upon this assumption. The analyst cannot assume that the analyzer's calibration
curve is a straight line between the measured values for the zero gas and the
reference standard. The analyst must calculate the calibration equation and the
uncertainty for its predicted concentrations by statistical analysis of the
measurements obtained during the multipoint calibration.

The total uncertainty of the certified concentration for a candidate standard

is composed of several components. The first component is the uncertainty
associated with the certified concentration of the reference standard. This
uncertainty is minimized by using an SRM, an RGM, a PRM, a CRM, an NTRM,
or a GMIS as the reference standard. The second component is the precision of
the measurements of the reference and candidate standards. This uncertainty is
minimized by making replicate measurements of these standards. The third
component is the uncertainty associated with the concentrations that are
predicted from the analyzer's calibration curve. This uncertainty concerns
whether an assumed calibration equation accurately represents the "true"
calibration curve.

This third component of uncertainty does not exist if the concentrations of

the reference and candidate standards are equal. The assumed calibration
equation and the true calibration curve will pass through the data for the
reference standard regardless of whether they diverge elsewhere and the
equation will be accurate for that single concentration. However, the uncertainty
does exist if the concentrations of the reference and candidate standards differ.
The assumed and true calibration curves may pass through different points for
concentrations not equal to that of the reference standard. Analytical errors will
develop because of this difference.

17
 EPA Traceability Protocol for Gaseous Calibration Standards

The measure of this uncertainty that is most directly useful to the analyst
is the value of U for a regression-predicted concentration given one or more
measurements of the candidate standard. This value may be calculated using
the spreadsheet described in Appendix A or using an EPA-approved equivalent
statistical technique (see Subsection 2.1.6). Several points should be noted
about this value. First, its magnitude decreases as n increases where n is the
number of measurements in the multipoint calibration. Second, its magnitude
decreases as n increases, where n is the number of measurements of the
candidate standard. Third, its magnitude increases as the mean measured
analyzer response ( Y ' ) for the candidate standard diverges from the overall
mean measured analyzer response ( Y ) for the multipoint calibration. These
points mean that it becomes easier to satisfy the uncertainty criterion as one
increases the number of measurements in the multipoint calibration and in the
assay of the candidate standard. Additionally, the absolute uncertainty of the
regression predicted concentration is larger at the extremes of the calibrated
concentration range than at the middle of the range.

For analyzers having an inherently nonlinear, but precise response, the

calibration equation can be calculated using quadratic or higher-order polynomial
regression analysis. Alternatively, a nonlinear equation may be linearized with a
simple mathematical transformation of the multipoint calibration data. Examples
of some linearizing transformations are given in Reference 19. The multipoint
calibration data may need to undergo several different transformations before the
optimum transformation is determined. Using appropriately transformed
calibration data, a calibration equation can be calculated with an acceptable
value of U for the regression-predicted concentration. Subsequently, data
obtained from the assay of the candidate standard must be similarly transformed
to calculate a concentration for the candidate standard.

2.1.4.4 Zero and Span Gas Checks—

On any day after the multipoint calibration that the analyzer will be used
for the assay of a candidate standard, its calibration drift must be measured.
This drift is calculated relative to the analyzer response during the multipoint
calibration. The purpose of the zero and span gas checks is to verify that the
calibration drift has remained within acceptable limits since the multipoint
calibration. The criterion that is used to assess the drift is the relative difference
between the analyzer's current response and the corresponding value from the
multipoint calibration. The following equation is used for this calculation:

Relative Difference = 100 [ Current Response - Calibration Response ]

[Calibration Response for Reference Standard]

This calculation is performed in Step 6 of the Appendix A spreadsheet.

Note that the relative difference is always calculated relative to the

calibration response for the reference standard, even when the zero gas is being

18
 EPA Traceability Protocol for Gaseous Calibration Standards

measured. This calculation is based on the statistical assumption that the

precision is the same at all concentrations. Long-term calibration drift cannot be
determined to be excessive if the zero and span gas measurements do not
exceed what can be explained by the short-term precision of the measurements.
This calculation is performed for the zero gas measurements and for the
reference standard measurements. If the reference standard was not measured
during the multipoint calibration, use the regression-predicted response for a
concentration equal to that of the reference standard.

If the relative differences for the zero and span gas checks are each less
than or equal to 5.0 percent, the analyzer's current calibration is considered to be
approximately the same as during the multipoint calibration and the assay may
be conducted. The zero and span controls do not have to be adjusted following
the zero and span checks because the data reduction technique used in this
protocol does not depend on the absolute accuracy of the analyzer calibration
equation at the time of the assay.

If the relative differences for the zero or span gas checks are greater than
5.0 percent, the analyzer is considered to be out of calibration. A new multipoint
calibration may be conducted before the candidate standard is assayed or the
analyzer's zero and span controls may be adjusted to return the analyzer's
response to the original calibration levels. For some analyzers such as
nondispersive infrared instruments, daily changes in environmental variables
such as barometric pressure may shift the calibration. After any adjustment of
controls, the analyst should repeat the zero and span gas checks and recalculate
the relative differences to verify that the analyzer is in calibration.

The zero gas and reference standard measurements that are performed
for the assay of the candidate standard may also be used for the zero and span
gas checks.

Between the time of the multipoint calibration and the time of the zero and
span gas checks, the analyst may adjust the analyzer’s zero and span controls
for assays that will not be certified according to this protocol. However, these
controls must be returned to their settings at the multipoint calibration before the
zero and span gas checks or assays under this protocol.

2.1.5 Assay/Certification of Candidate Standards

See Subsections 2.2, 2.3, and 2.4 for Procedures G1, G2, and G3,
respectively

2.1.5.1 Incubation of Newly Prepared Candidate Standards—

Newly prepared candidate standards must be incubated at least 4 days
before being assayed and certified.

19
 EPA Traceability Protocol for Gaseous Calibration Standards

2.1.5.2 Stability Test for Reactive Gas Mixtures—

Newly prepared candidate standards that contain reactive gas mixtures,
including ammonia (NH3), hydrogen chloride (HCl), hydrogen sulfide (H2S), nitric
oxide (NO), oxides of nitrogen (NOx), and sulfur dioxide (SO2), and that have not
been previously certified, must be assayed on at least two dates that are
separated by at least 7 days. Candidate standards that contain nonreactive gas
mixtures, including carbon dioxide (CO2), carbon monoxide (CO), methane
(CH4), nitrous oxide (N2O), oxygen (O2), and propane (C3H8), do not require a
second assay. Zero air materials do not require a second assay.

Conduct an initial assay of the candidate standard and determine a

concentration for the standard. Make three or more discrete measurements of
the candidate standard during the initial assay. "Discrete" means that the analyst
must change the gas mixture being sampled by the analyzer between
measurements by turning gas flow control valves or by other means. Record
these measurements in the laboratory's records.

Reassay the candidate standard at least 7 days after the first assay.
Make three or more discrete measurements of the candidate standard during the
second assay. Record these measurements in the laboratory's records.

Use the spreadsheet described in Appendix C or an EPA-approved

equivalent statistical technique (see Subsection 2.1.6) to determine if the mean
estimated concentrations from first and second assays differ by less than 1.0
percent or less. If this acceptance criterion is attained, the candidate standard is
considered to be stable and can be certified. If this acceptance criterion is not
attained, the candidate standard may be unstable or there may be analytical
problems associated with the assays or the reference standards. The analyst
must either disqualify the candidate standard or investigate why the acceptance
criterion was not attained.

In earlier versions of this protocol, Student's t-test was used to determine

whether there was a statistically significant difference between the mean
concentrations from the two assays. It penalized high-precision measurements
because a slight, analytically inconsequential concentration difference could be
found to be statistically significant if the uncertainties of the two assays are
sufficiently small. Conversely, a substantial concentration difference could pass
Student's t-test, if measurement precision was large. In this version of the
protocol, Schuirmann's two one-sided test (TOST)13-18 is used in Appendix C to
determine if the two mean estimated concentrations differ by 1.0 percent or less.
TOST is used in the pharmaceutical industry to show analytical method
equivalency. It is superior to Student's t-test because it does not allow
substantial concentration differences to pass the statistical significance test if
measurement precision is poor. That is, being able to find that a candidate
standard's concentration is stable (i.e., within the TOST acceptance criterion) is

20
 EPA Traceability Protocol for Gaseous Calibration Standards

more important than not being able to find that the concentration is unstable.
See Appendix D for a further discussion of TOST calculations with examples.

If a candidate standard's concentration is not found to be stable (i.e., not

within the TOST acceptance criterion), the analyst may elect to discard the
candidate standard or may elect to conduct a third assay of the candidate
standard to assess whether stability has been achieved. The analyst must add
the additional data to the Appendix C spreadsheet. The analyst must wait an
additional 7 days or more and conduct the third assay. If the data for the third
assay is found to be equivalent to the data for either of the two previous assays,
the candidate standard can be certified using the data from the two equivalent
assays, which will be used to calculate the overall estimated concentration and
the total uncertainty. Data from a nonequivalent assay should be discarded. The
analyst must disqualify the candidate standard if none of the three sets of data
are found to be equivalent. The analyst is expected to investigate and document
the cause of the lack of agreement among the three assays and is expected to
correct any problems that are discovered. Record the equivalent data and any
discarded data in the laboratory’s records.

If the data from the two assays are found to be statistically equivalent and
yet the analyst is concerned that these data suggest potential instability, the
analyst may elect to conduct a third assay to confirm that stability has been
achieved. The analyst must wait an additional 7 days or more and conduct the
third assay and must add the additional data to the Appendix C spreadsheet.
Record the additional data in the laboratory’s records.

2.1.5.3 Assay/Certification of Multicomponent Candidate Standards—

This protocol may be used to assay and certify a multiple-component
standard if compressed gas SRMs, NTRMs, RGMs, PRMs or CRMs exist that
contain the individual components of the multiple-component standard. Some or
all of the components may be assayed and certified according to this protocol
and the remaining components may be assayed and certified by other methods.
The certification documentation and certification label (see Subsections 2.1.7 and
2.1.8) must clearly state which components of the standard are certified
according to this protocol and which are not.

If any component in the multiple-component standard interferes with the

assay of any other component, the analyst must conduct an interference study to
determine an interference correction equation. For example, NIST found multiple
interferences (i.e., NDIR analysis of NO or SO2 in the presence of CO2; NDUV
analysis of NO in the presence of SO2; and chemiluminescence analysis of NO in
presence of CO2) during its 2010 audit of multiple-component EPA Protocol
Gases for EPA37.

This interference study must be conducted using the same analyzer or

analyzers as will be used to assay the standard. The study must use single-

21
 EPA Traceability Protocol for Gaseous Calibration Standards

component and multiple-component reference standards that have been assayed

using interference-free analyzers. The study must cover the same range of
concentrations for all components as will exist for the standards being assayed
and certified according to this protocol.

Data from the interference study must be evaluated using multiple-

variable least-squares regression analysis. The analyst should consult with a
statistician before beginning the study or evaluating its data. The regression
analysis must produce an interference correction equation and an estimate of the
standard uncertainty (uCORRECTED) associated with the corrected concentrations
for the assayed components. The interference correction equation will be valid
for the range of concentrations covered in the study for which the uncertainty of
the corrected concentration is ≤1 percent of the corrected concentration. The
analyst must add the interference correction uncertainty to the total uncertainty of
the standard. The certification documentation must include a statement that the
certified concentration of a specified component has been corrected for
interferences from other specified components. An interference study is not
needed if the assay analyzer is interference free.

2.1.5.4 Assay/Certification of Trace Contaminants in Candidate Standards—

This protocol may be used to assay and certify the concentration of a
trace contaminant in a candidate standard if compressed gas SRMs, NTRMs,
RGMs, PRMs or CRMs exist for the trace contaminant. Such an assay and
certification may be needed if the main component and the trace contaminant
both produce an analyzer response in an end user's pollutant gas analyzer. For
example, both NO and NO2 produce a response in the NOX channel of a
chemiluminescent NO/NOX analyzer. A specialty gas producer may wish to
certify the NOX concentration of a standard by separately assaying the standard's
NO and NO2 concentrations using a Fourier transform infrared (FTIR) analyzer,
rather than a chemiluminescent NO/NOX analyzer. If the producer uses a
chemiluminescent NO/NOX analyzer to certify the NO and NOX concentrations of
a standard, then no special procedure is required because this analyzer has
separate NO and NOX channels and can measure NOX directly.

The following steps give a simplified procedure for determining the trace
contaminant concentration:

1. Perform monthly multipoint calibrations of the pollutant gas analyzer for the
main component and the trace contaminant using reference standards for
both components and using either Procedure G1 or Procedure G2.

2. On the day of the assay of the candidate standard, perform zero and span
gas checks for the main component as described in Subsection 2.2.6.4 or in
Subsection 2.3.8.4. If the calibration drift for the main component is within the
acceptance criterion (i.e., relative differences for the zero and span gas
checks less than or equal to 5.0 percent), then the analyst may assume that

22
 EPA Traceability Protocol for Gaseous Calibration Standards

excessive calibration drift has not occurred for the trace contaminant.
However, if the calibration drift is excessive, then the analyzer is considered
to be out of calibration. Follow the procedures for out-of-calibration analyzers
that are described in Subsection 2.2.6.4 or in Subsection 2.3.8.4.

3. Perform the assays for the trace contaminant during the same assay session
as the assays of the main component of the candidate standard.
Measurement of the trace contaminant reference standard is not necessary.

4. Calculate the concentration of the trace contaminant and its uncertainty using
the spreadsheet described in Appendix A (or an EPA-approved equivalent
statistical technique, see Subsection 2.1.6). Enter data from the trace
contaminant measurements in the spreadsheet cells in Step 5 where
candidate standard data for the same day as the calibration normally would
be entered. Record these values in the laboratory's records. The trace
contaminant concentration cannot be greater than 2.0 percent of the main
component concentration.

The certification documentation and certification label (see Subsections

2.1.7 and 2.1.8) must clearly state the trace contaminant concentration and its
uncertainty. A producer can elect to also report the sum of the main component
concentration and the trace contaminant concentration and the uncertainty of that
sum (see Subsection 2.1.6).

2.1.6 Expanded Uncertainty of the Concentration of the Candidate Standard

The expanded uncertainty (U) is the combined standard uncertainty (uc)

multiplied by a coverage factor (k), which is equal to 2.8,9 The value of U is
equivalent to a 95- percent confidence interval in statistics. It is due to many
different error sources, including the uncertainty in the reference standards,
uncertainty in the analyzer multipoint calibration, uncertainty in the zero/span
correction factors, random measurement error, uncertainty in interference
corrections, and uncertainty in gas dilution. This protocol provides statistical
spreadsheets to calculate and combine these uncertainty components. There
may be additional uncertainty sources that cannot be assessed with the limited
data that are produced when implementing this protocol. The calculated value of
U must be given in the certification documentation. The use of an uncalculated
or blanket estimate of U (e.g. +/- 1 percent) is not acceptable.

The analyst must use the Appendix A spreadsheet (or an EPA-approved

equivalent statistical technique) to calculate an estimated concentration and an
estimated U for each assay of the candidate standard. The estimate of U
includes only the uncertainty component associated with the assay. It does not
include the uncertainty components associated with the reference standard,
interference correction or dilution. These uncertainty components are added in

23
 EPA Traceability Protocol for Gaseous Calibration Standards

the Appendix C spreadsheet after the overall estimated concentration is

determined from all assays of the candidate standard.

The analyst must use the Appendix C spreadsheet (or an EPA-approved

equivalent statistical technique) to calculate the overall estimated concentration
and U for all assays of the candidate standard. This spreadsheet uses the output
from the Appendix A spreadsheet for one, two or three assays. Only one assay
is needed for candidate standards that contain nonreactive gas mixtures. Two or
three assays are needed for candidate standards that contain reactive gas
mixtures. The overall estimated concentration and U for the candidate standard
are to be reported in the certification documentation.

If a specialty gas producer or other laboratory wishes to use an equivalent

statistical technique to calculate the overall estimated concentration and U for the
candidate standard, then a detailed description of the technique and any
supporting software must be submitted to EPA Traceability Protocol Project,
Technical Services Branch, U.S. EPA, Mail Code E343-03, Research Triangle
Park, NC 27711 for statistical evaluation and approval. The description must be
in sufficient detail to demonstrate equivalence to the spreadsheets described in
Appendices A and C. Example calculations demonstrating equivalence must be
included. Any information about the technique that is submitted to EPA will be
treated as confidential business information. Upon approval by EPA, the use of
this equivalent statistical technique must be noted in the certificate
documentation and a description of the technique must be presented in media
readily accessible to end users.

Although this protocol does not specify an acceptance criterion for U for
candidate standard concentrations, EPA's Acid Rain Program has specified in 40
CFR Part 75 that an EPA Protocol Gas must have a specialty gas producer-
certified uncertainty (95- percent confidence interval) that must not be greater
than 2.0 percent of the certified concentration (tag value) of the gas mixture5. In
general, an end user's purchase specifications for an EPA Protocol Gas should
include a specification for U that meets its needs and any regulatory
requirements concerning the uncertainty of the certified concentration.

The certified value of U for SRMs, NTRMs, RGMs, PRMs, and CRMs is
expressed on their certificates of analysis as the combined standard uncertainty
(uc) multiplied by a coverage factor (k), which is equal to 2.8,9 The combined
standard uncertainty includes the uncertainties of known sources of systematic
error as well as the random error of measurement. A value of one-half of U for
the reference standard should be used in calculating U of candidate standards
that are certified under this protocol (see Appendix C).

For those cases when the candidate standard is assayed on the same day
as the multipoint calibration, the candidate standard's concentration is
determined directly from the calibration curve. The combined standard

24
 EPA Traceability Protocol for Gaseous Calibration Standards

uncertainty of the concentration is calculated by using the spreadsheets

described in Appendices A and C or an EPA-approved equivalent statistical
technique (see below). It combines the uncertainty of the assay (uASSAY) with the
uncertainty of the reference standard (uSTANDARD) using the following equation:

=uc (uSTANDARD )2 + (uASSAY )2

For those cases in which the candidate standard is assayed on a date

following the multipoint calibration, uASSAY includes the uncertainty associated
with the drift in the zero and span gas measurements.

If a trace contaminant in the candidate standard has been measured in

addition to the measurement of the main component, calculate the standard
uncertainty of the sum of the main component and the trace component from the
uncertainty of the main component (uMAIN) and the uncertainty of the trace
component (uTRACE) using the following equation:

= uc (uMAIN )2 + (uTRACE )2
If an interference-correction equation has been used to obtain a corrected
concentration for the candidate standard, the standard uncertainty for the
corrected concentration (uCORRECTION) must be included in the assessment of the
total analytical uncertainty of the candidate standard's concentration using the
following equation:

=
UC (USTANDARD )2 + (UASSAY )2 + (UCORRECTION )2

If dilution has been used in the assay of the candidate standard, the
standard uncertainty for the dilution (uDILUTION) must be included in the
assessment of the total analytical uncertainty of the candidate standard's
concentration using the following equation:

=
UC (USTANDARD )2 + (UASSAY )2 + (UCORRECTION )2 + (UDILUTION )2

The UDILUTION does not have to be included in the calculation of UC if

exactly the same settings for the Procedure G2 gas dilution apparatus are used
for the analysis of the reference standard and the candidate standard. In this
case, no variability in the results is introduced by using a constant dilution setting.

Generally, U should be rounded to one significant figure unless the

leading figure is a 1 in which case two significant figures should be reported17.

The estimate of U (e.g., +/- 20 ppm) may be supplemented, but not

replaced) by a statement of the equivalent fractional uncertainty (e.g., +/- 1
percent), if desired, using the following equation:

25
 EPA Traceability Protocol for Gaseous Calibration Standards

UFRACTIONAL = 100(U/ Certified Concentration)

The same rule-of-thumb regarding significant figures applies to the

fractional uncertainty.

2.1.7 Certification Documentation for Certified Standards and Zero Air Materials

2.1.7.1 Certification Documentation for Assays using Procedure G1 or G2—

For each standard (i.e., an EPA Protocol Gas, a GMIS or a zero air
material) that is assayed using Procedure G1 or G2, the assay results must be
documented in a written report, which contains at least the following information:

1. Cylinder identification number (e.g., stamped cylinder number).

2. The certified concentrations for the assayed components of the standard, in

ppb by mole, ppm by mole or in mole percent. Use the spreadsheet
described in Appendices A and C (or an EPA-approved equivalent statistical
technique) to calculate the certified concentrations, which are the weighted
means of all assayed concentrations for which the standard is considered to
be stable. They generally should be reported to at least 3 significant figures.
The last significant figure that is reported must be of the same order of
magnitude as U (e.g., 2530 ppm +/- 20 ppm, not 2533 ppm +/- 20 ppm) as is
discussed below. If a candidate standard contains components that interfere
with it measurement of components being certified by more than 1.0 percent
of the estimated concentration of the certified component, then the magnitude
of the interference must be included in the calculation of the certified
concentration. For example, consider a candidate standard containing NO,
SO2, and CO2 in nitrogen whose NO concentration is being measured using a
chemiluminescent analyzer. The CO2 in the candidate standard will interfere
with the NO measurement and the CO2 interference must be included in the
calculation of the NO concentration if the magnitude of interference exceeds
1.0 percent of the estimated NO concentration.

If a zero air material was assayed using Procedure G1 or G2, the certification
documentation must include the impurity concentrations of the assayed
components and a statement to the effect that it conforms to the purity
specifications of 40 CFR Part 72.2. Use the spreadsheet described in
Appendix A or an EPA-approved equivalent statistical technique to calculate
the impurity concentrations, which are the means of all measurements of the
zero air material. The significant figures in the certified impurity
concentrations generally should be the same as U (e.g., 0.01 ppm +/- 0.03
ppm, not 0.012 ppm +/- 0.30 ppm) as is discussed below.

3. A calculated estimate of U for the standard (see Section 2.1.6). The use of
an uncalculated, blanket value for U (e.g., +/- 1 percent) is not acceptable.
The estimate is expressed as a 95-percent confidence interval, which is the

26
 EPA Traceability Protocol for Gaseous Calibration Standards

combined standard uncertainty (uc) multiplied by a coverage factor (k), which

is equal to 2.8,9 It must include the uncertainties associated with the reference
standards, the analyzer multipoint calibration, the analyses of the standard,
any gas dilutions, and any interference correction. Use the spreadsheet
described in Appendix C or an EPA-approved equivalent statistical technique
(see Section 2.1.6) to calculate U. If an EPA-approved equivalent statistical
technique is used rather that the Appendix C spreadsheet, it must be
identified as such in the certificate documentation and must be described in
media readily accessible to end users. As a rule-of-thumb, U should be
rounded to one significant figure unless the leading figure is a 1 in which case
two significant figures should be reported20. The estimate of U (e.g., +/- 20
ppm) may be supplemented by a statement of the equivalent fractional
uncertainty (e.g., +/- 1.0 percent) if desired

If a zero air material was assayed using Procedure G1 or G2, the certification
documentation must include a calculated estimate of U for the impurity
concentrations in the zero air material using the statistical procedures
described above.

4. The assayed component(s) and balance gas in the gas mixture. If the
composition or origin of the balance gas (e.g, oxygen percentage or synthetic
vs. scrubbed ambient air) has a measurable effect on the end user's pollutant
gas analyzers, it must described in the certification documentation.

5. Cylinder pressure at certification and the statement that the standard should
not be used when the pressure is below 0.7 megapascals (i.e., 100 psig).

6. Dates of the assays and certification. The certification date is the date of the
last assay.

7. Certification expiration date (i.e., the certification date plus the certification
period plus one day) (see Subsection 2.1.9).

8. Information about the reference standard used in the assay: NIST SRM
number, NIST sample number, cylinder identification number, certified
concentration, expanded uncertainty, and certification expiration date for an
SRM; cylinder identification number, certified concentration, expanded
uncertainty, and certification expiration date for an NTRM, an RGM, a PRM, a
CRM or a GMIS. The certification documentation must identify the reference
standard as being an SRM, an NTRM, an RGM, a PRM, a CRM or a GMIS.
For a candidate standard that is assayed using a GMIS, the certification
documentation must include information about the reference standard that
was used for the GMIS' assay (see above).

9. Statement that the assay/certification was performed according to this

protocol and that lists the assay procedure (i.e., Procedure G1 or G2) used.

27
 EPA Traceability Protocol for Gaseous Calibration Standards

10. The analytical method that was used in the assay and the date of the most
recent multipoint calibration of the instrument.

11. Identification of the specialty gas producer or other laboratory (i.e., producer's
or laboratory's name, city, and state, and PGVP vendor ID issued by EPA for
the production location) where the candidate standard was assayed, even if
another organization prepared the candidate standard or will sell it to an end
user. This identification must be given in the same or larger font as the other
required information in the report.

The IDs are listed at www.epa.gov/airmarkets/emissions/pgvp-vendorID.html.

12. Chronological record of all certifications for the standard.

13. If applicable, a statement that the certified concentration for one component
has been corrected for interferences from other specified components.

This certification documentation must be given to the end user of the

standard. The specialty gas producer or other laboratory that assayed the
standard must maintain laboratory records and certification documentation for at
least the standard's certification period or until such time as the standard is blown
down after being returned by the end user to the producer. A specialty gas
producer or other vendor may redocument an assayed and certified standard that
it has purchased from another specialty gas producer and that it wishes to sell to
a third party. However, the new certification documentation must clearly identify
the specialty gas producer or other laboratory (i.e., name, city, and state) where
the standard was assayed and present all the information that is contained in the
original report. The documentation may, at the producer's discretion, include
identification of the producer or laboratory that prepared the candidate standard if
different from the producer or laboratory that assayed and certified the standard.

2.1.7.2 Certification Documentation for Assays using Procedure G3—

For a zero air material that is assayed using Procedure G3, the assay
results must be documented in a written report, which contains at least the
following information:

1. Cylinder identification number (e.g., stamped cylinder number).

2. If a zero air material was assayed using Procedure G3, use the spreadsheet
described in Appendix E or an EPA-approved equivalent statistical technique
(see Subsection 2.1.6) to calculate whether the purity specifications of 40
CFR Part 72.2 have been attained. If the specifications have been attained,
the certification documentation must include a statement to this effect.

28
 EPA Traceability Protocol for Gaseous Calibration Standards

3. If the zero air material was assayed using Procedure G3, no uncertainty
estimate need be calculated or documented.

4. A statement that there is no expiration date for the certification

5. Information about the NIST-traceable zero air material used in the assay:
cylinder identification number, certified concentration, expanded uncertainty,
and certification expiration date. The certification documentation must identify
the NIST-traceable zero air material as being an SRM, an NTRM, an RGM, a
PRM, a CRM or a GMIS. For a GMIS, the NIST-traceable zero air material
that was used for its assay must be identified in the documentation for the
zero air material being certified.

6. Information about the reference standard used in the assay: cylinder

identification number, certified concentration, expanded uncertainty, and
certification expiration date for the reference standard. The certification
documentation must identify the reference standard as being an SRM, an
NTRM, an RGM, a PRM, a CRM or a GMIS. For a GMIS, the reference
standard that was used for its assay must be identified in the documentation
for the zero air material being certified.

7. Statement that the assay/certification was performed according to this

protocol and that lists the assay procedure (i.e., Procedure G3) used.

8. The analytical method that was used in the assay and the date of the most
recent multipoint calibration of the instrument.

9. Identification of the specialty gas producer or other laboratory (i.e., producer's

or laboratory's name, city, and state, and PGVP vendor ID issued by EPA for
the production location) where the zero air material was assayed, even if
another organization prepared the zero air material or will sell it to an end
user. This identification must be given in the same or larger font as the other
required information in the report.

10. If the composition or origin of the balance gas (e.g, oxygen percentage or
synthetic vs. scrubbed ambient air) has a measurable effect on the end user's
pollutant gas analyzers, it must described in the certification documentation.

This certification documentation must be given to the end user of the zero
air material. The specialty gas producer or other laboratory that assayed the
standard must maintain laboratory records and certification documentation for at
least the standard's certification period or until such time as the standard is blown
down after being returned by the end user to the producer. A specialty gas
producer or other vendor may redocument a zero air material that it has
purchased from another specialty gas producer and that it wishes to sell to a third
party. However, the new certification documentation must clearly identify the

29
 EPA Traceability Protocol for Gaseous Calibration Standards

specialty gas producer or other laboratory (i.e., name and location) where the
zero air material was assayed and present all the information that is contained in
the original report.

2.1.8 Certification Label

A label or tag must be attached to the standard bearing the information

described in Items 1-5, 7, 9, and 11 of Subsection 2.1.7.1 for each standard
assayed using Procedure G1 or G2 and in Items 1, 2, 4, 7, and 9 of Subsection
2.1.7.2 for each zero air material assayed using Procedure G3.

2.1.9 Certification Periods for Standards

The certification of a standard is valid for only a specified period following

its certification date, which is the date of its last assay. In general, the
certification period should be no longer than the period for which similar gas
mixtures (e.g., SRMs or similar standards) over specific concentration ranges
have been shown to be stable as documented in the peer review literature or in
concentration stability data submitted by NIST and specialty gas producers for
review by EPA.21-29 Maximum certification periods for various standards that are
certified or recertified under this protocol are specified in Table 2-3. The
certification period for a GMIS is the same as for an EPA Protocol Gas. A
multiple-component standard can be certified for a maximum period equal to that
of its most briefly certifiable component. For example, a standard containing
sulfur dioxide, carbon monoxide, and propane in nitrogen can be certified for up
to 4 years because the shortest certification period is 4 years. These certification
periods are for standards that are contained in passivated aluminum cylinders,
except for hydrogen chloride mixtures where nickel-coated steel cylinders may
also be used and for zero air material where steel cylinders may be used. In
general, the certification period for standards that are contained in nonaluminum
cylinders is 6 months. However, an exception is made for the following three
nonreactive gas mixtures: CO2 with a concentration >0.5 percent; O2 with a
concentration >0.5 percent; and C3H8 with a concentration >0.1 percent. The
certification period for standards containing these three nonreactive gas mixtures
in nonaluminum cylinders is the same as is given in Table 2-3 for passivated
aluminum cylinders.

Specialty gas producers may elect to certify candidate standards for less
than the maximum certification period in Table 2-3 if they believe that they cannot
prepare standards whose stability attains the maximum certification period. The
default certification period is that given in Table 2-3. Each producer has
discretion in this matter.

30
 EPA Traceability Protocol for Gaseous Calibration Standards

TABLE 2-3. Maximum Certification Periodsa for Calibration Standards in

Passivated Aluminum Cylinders
Components Balance gas Concentration range Period (years)
Ammonia Nitrogen 5 to 50 ppm 1
b
Carbon dioxide Air 360 to 420 ppm 8
Carbon dioxide Nitrogen 5 ppm to 20% 8
Carbon monoxide Air 40 to 500 ppb TBD
Carbon monoxide Air 500 ppb to 10% 8
Carbon monoxide Nitrogen 1 ppm to 15% 8
Formaldehyde Nitrogen 0.5 to 10 ppm 1
c
Hydrogen chloride Nitrogen 10 to 5000 ppm 2
Hydrogen sulfide Nitrogen 1 to 1000 ppm 3
Methane Air 1 to 1000 ppm 8
Methane Nitrogen 500 ppb to 10% 8
Methanol or
Nitrogen or Air 75 to 500 ppm 4
ethanol
Natural gas 4
Natural gas Contact NIST
componentsd
Nitric oxide O2-free nitrogene 0.5 to 50 ppm 3
e
Nitric oxide O2-free nitrogen 50 ppm to 1% 8
Nitrous oxide Air 300 ppb to 5% 8
f
Oxides of nitrogen Air 3 ppm to 1% 3
Oxygen Nitrogen 10 ppm to 25% 8
Propane Air 0.1 to 500 ppm 8
Propane Nitrogen 5 ppb to 2% 8
Sulfur dioxide Nitrogen 1 to 50 ppm 4
Sulfur dioxide Nitrogen 50 ppm to 1%- 8
Volatile organics Nitrogen 1 ppb to 1 ppm 4
Zero air materialg Air Not applicable Unlimited
Multicomponent — — See text
mixtures
Mixtures with lower — — See text
concentrations
a
Specialty gas producers may elect to certify candidate standards for less than the maximum
certification period. Each producer has discretion in this matter. See text.
b
"Air" is defined as a mixture of oxygen and nitrogen where the minimum concentration of
oxygen is 10 percent and the concentration of nitrogen is greater than 60 percent.
c
Hydrogen chloride may be contained in passivated aluminum or nickel-coated steel cylinders.
d
Natural gas components are methane, ethane, propane, n-butane, iso-butane, n-pentane, iso-
pentane, helium, nitrogen, and carbon dioxide.
e
O2-free nitrogen contains ≤ 100 ppb of oxygen.
f
NIST defines its total NOX standards as containing nitrogen dioxide plus contaminant nitric acid.
g
Concentrations of SO2, NOX, and THC are not >0.1 ppm; concentration of CO is not >1 ppm;
and concentration of CO2 is not >400 ppm as per 40 CFR Part 72.2. Zero air material may be
contained in steel cylinders.

31
 EPA Traceability Protocol for Gaseous Calibration Standards

There is a risk that regulator-related contamination or improper storage by

the end user may alter the stability of the certified concentrations in an EPA
Protocol Gas. This protocol cannot account for the misuse of EPA Protocol
Gases. The solution to this potential problem is to educate end users in the
proper storage and use of EPA Protocol Gases to avoid stability problems. It is
also possible that instability may be associated with the cylinder passivation
technique, which varies among producers and which is generally proprietary
information. Each producer should investigate its own passivation techniques
and select certification periods that are appropriate for those techniques.

This protocol allows the certification of standards with concentrations

that are lower than those given in Table 2-3. If the concentration of the candidate
standard is less than the applicable concentration range given in Table 2-3 and if
the time period between the initial and second assays is less than 6 months, then
the initial certification period for this standard is 6 months. After this period, the
standard must be recertified before further use. If the time period between the
initial and second assays is at least 6 months and if the mean concentrations
from the two assays are demonstrated to be within 1.0 percent of each other
using Schuirmann's TOST (see Subsection 2.1.5.2), then the standard can be
certified for the period shown in Table 2-3.

2.1.10 Certification Periods for Zero Air Materials

The certification of a zero air material is unlimited and there is no

expiration date for the certification.

2.1.11 Recertification of Standards

If a standard is to be used after its certification period has ended, it must

be recertified in accordance with this protocol. It may be recertified only if the
gas pressure remaining in the cylinder after the assays have been completed is
greater than 0.7 megapascals (i.e., 100 psig). Zero air materials do not need to
be recertified.

The recertification assay must be performed using the same analytical

procedure (e.g., Procedure G1) as was used for the original assay of the
standard. The purpose of this assay is to determine whether the standard has
remained stable since its original certification. To recertify a standard, three or
more discrete measurements of the candidate standard must be made during a
single recertification assay. There is no requirement that the same reference
standard must be used in the original and recertification assays, although this
practice is desirable if possible. Record the results of the recertification assay in
the laboratory's records.

Use the spreadsheet described in Appendix C (or an EPA-approved

equivalent statistical technique) to compare the measured concentrations from
the recertification assay with the measured concentrations from the previous

32
 EPA Traceability Protocol for Gaseous Calibration Standards

assays. If the TOST acceptance criterion is attained, the standard can be

recertified. The recertification period is the same as that given in Table 2-3. If
the measured concentrations are shown to be not equivalent, the standard must
undergo a full certification (e.g., Procedure G1) before it can be used again.

The spreadsheet described in Appendix C can include data from only

three assays. If more than three assays are conducted, only the data from the
three most recent assays should be used in the spreadsheet.

A standard that was certified under this protocol may be recertified by a

laboratory other than the one that performed the original certification. In such a
case, the recertification documentation must list the information from the original
certification documentation plus the corresponding information from the
recertification assays. Both the original and the recertification laboratories must
be identified in the recertification documentation.

If the TOST acceptance criterion is not attained, a second recertification

assay may be conducted. If the two recertification assays attain the TOST
acceptance criterion, the standard can be recertified.

If the TOST acceptance criterion is still not attained after the second
recertification assay, the analyst must either disqualify the standard for further
use under this protocol or investigate why there is an apparent difference
between the original assays and the recertification assay. This difference may
be due to an actual instability of the gas mixture, to a reference standard
problem, to an analytical instrumentation problem, or to some other problem. If
the analyst can find a reasonable explanation for the difference and if this reason
is not instability, then the standard can be recertified. The analyst must append a
brief report on the results of the investigation to the recertification documentation
and to the laboratory's records.

Standards having certified concentrations that are lower than those given
in Table 2-3 may be recertified for the period given in Table 2-3 provided at least
6 months have elapsed between the initial certification and the recertification.
For example, a 0.5-ppm sulfur dioxide in nitrogen standard will have an initial
certification period of 6 months. After a successful recertification, this standard
will have a recertification period of 4 years. The certification date is the date of
the last assay.

A multiple-component standard can be recertified for a period equal to

that of its most briefly certifiable component. For example, a standard containing
ammonia, carbon monoxide, and propane in nitrogen can be recertified for 2
years because the shortest certification period is 2 years for ammonia. In this
case, the certification date is the date of the last ammonia assay.

33
 EPA Traceability Protocol for Gaseous Calibration Standards

2.1.12 Minimum Cylinder Pressure

A certified standard should not be used when its gas pressure is below
0.7 megapascals (i.e., 100 psig). NIST has found that some gas mixtures (e.g.,
nitric oxide in nitrogen) have exhibited a concentration change when the cylinder
pressure fell below this value. There is no minimum cylinder pressure
specification for zero air materials.

2.1.13 Protocol Gas Verification Program (PGVP)

Periodically, the U.S. EPA will assess the accuracy of a blind sample of
compressed gas calibration standards that have been assayed and certified
according to this protocol. The analytical results, identifying the specialty gas
producers or other analytical laboratories that assayed and certified the
standards, will be published as public information.

In the 1980s and 1990s, EPA conducted a series of EPA-funded accuracy

assessments of EPA Protocol Gases sold by producers30-34. The intent of these
audits was to:

• increase the acceptance and use of EPA Protocol Gases as calibration

gases;
• provide a quality assurance (QA) check for the producers of these gases; and
• help end users identify producers who can consistently provide accurately
certified gases.

Either directly or through third parties, EPA procured EPA Protocol Gases
from the producers, assessed the accuracy of the gases' certified concentrations
through independent analyses, and inspected the accompanying certificates of
analysis for completeness and accuracy. The producers were not aware that
EPA had procured the gases for these audits.

The accuracy of the EPA Protocol Gases' certified concentrations was

assessed using SRMs as the reference standards. If the difference between the
audit's measured concentration and the producer's certified concentration was
more than +/- 2.0 percent or if the documentation was incomplete or inaccurate,
EPA notified the producer to resolve and correct the problem. The results of the
accuracy assessments were published in peer-reviewed journals and were
posted on EPA's Technology Transfer Network website.

The accuracy assessments were discontinued in 1998. In 2002, there

was interest by the specialty gas producers and EPA to reestablish this program.

In 2009, the Office of the Inspector General (OIG) published the report
EPA Needs an Oversight Program for Protocol Gases35. OIG recommended
that the Office of Air and Radiation (OAR) implement oversight programs to

34
 EPA Traceability Protocol for Gaseous Calibration Standards

assure the quality of EPA Protocol Gases used to calibrate continuous emissions
monitoring systems and ambient air monitors. It also recommended that EPA's
Office of Research and Development (ORD) update and maintain this traceability
protocol to ensure that the monitoring programs' objectives are met.

To address the OIG findings for ambient air monitoring, EPA's Office of
Air Quality Planning and Standards (OAQPS), in cooperation with EPA Regions
2 and 7 developed an Ambient Air Protocol Gas Verification Program (AA-
PGVP). This program establishes gas metrology laboratories in Regions 2 and 7
to verify the certified concentrations of EPA Protocol Gases used to calibrate
ambient air quality monitors. It is expected to:

• ensure that producers selling EPA Protocol Gases participate in the AA-
PGVP, and
• provide end users with information about participating producers and
verification results.

The AA-PGVP QA requirements in 40 CFR Part 58, Appendix A require:

"2.6 Gaseous and Flow Rate Audit Standards. Gaseous pollutant concentration standards
(permeation devices or cylinders of compressed gas) used to obtain test concentrations for
CO, SO2, NO, and NO2 must be traceable to either a National Institute of Standards and
Technology (NIST) Traceable Reference Material (NTRM), NIST Standard Reference
Materials (SRM) and Netherlands Measurement Institute (NMi) Primary Reference Materials
(valid as covered by Joint Declaration of Equivalence) or a NIST-certified Gas Manufacturer's
Internal Standard (GMIS), certified in accordance with one of the procedures given in
reference 4 of this appendix. Vendors advertising certification with the procedures provided in
reference 4 of this appendix and distributing gases as 'EPA Protocol Gas' must participate in
the EPA Protocol Gas Verification Program or not use 'EPA' in any form of advertising."

The AA-PGVP is considered a verification program because its current

level of evaluation does not allow for a large enough number of samples of EPA
Protocol Gases from any one specialty gas producer to yield a statistically
rigorous assessment of the accuracy of any specific producer's standards. It will
not provide end users with a scientifically defensible estimate of whether
standards of acceptable quality can be purchased from any specific producer.
Rather, the results provide information to end users that the specialty gas
producer is participating in the program and the information in the verification
report may be helpful when selecting a producer.

The results from the first two years of the AA-PGVP have been
published36. More information about the AA-PGVP can be found at OAQPS' AA-
PGVP website, www.epa.gov/ttnamti1/aapgvp.html.

In a parallel effort, EPA's Office of Air and Radiation (OAR) Clean Air
Markets Division (CAMD) conducted assessments in 2003 and 2010 of a blind
sample of EPA Protocol Gases containing CO2, NO, and SO2 that are used to
calibrate continuous emission monitors under the Acid Rain Program37. EPA has

35
 EPA Traceability Protocol for Gaseous Calibration Standards

promulgated amendments to 40 CFR Part 75 that require that EPA Protocol

Gases used for Part 75 purposes be obtained from specialty gas producers that
participate in a source-level PGVP38. These amendments state:

"On and after May 27, 2011 for each unit subject to this part that uses EPA Protocol gases,
the owner or operator must obtain such gases from either an EPA Protocol gas production
site that is on the EPA list of sites participating in the PGVP on the date the owner or operator
procures such gases or from a reseller that sells to the owner or operator unaltered EPA
Protocol gases produced by an EPA Protocol gas production site that was on the EPA list of
participating sites on the date the reseller procured such gases."

Any specialty gas producer choosing to participate in the source-level

PGVP must notify EPA each year. Each year, EPA will hire a third party to
blindly procure up to 4 cylinders from each participating producer. After being
notified that its cylinders are being audited by EPA (after cylinders have been
received), a producer would cancel its invoice or credit the third party’s account,
and pay NIST to analyze its cylinders and provide a report to EPA. NIST expects
to analyze all cylinders and provide a report to EPA within 6 months of receipt of
the cylinders and has established cost containment/efficiency measures to help it
do so. EPA will post audit results on-line. The results will indicate whether
cylinder meets the Part 75 performance specification of + 2.0% of cylinder tag
value. End users can decide from whom they wish to purchase gases.

Recordkeeping/reporting will be required to help ensure that EPA Protocol

Gases being used by Part 75 sources are from Acid Rain PGVP participants, and
to inform cylinder selection for future audits.

More information about the Emission PGVP can be found at CAMD's

website, www.epa.gov/airmarkets/emissions/rules.html.

36
 EPA Traceability Protocol for Gaseous Calibration Standards

2.2 Procedure G1: Assay and Certification of a

Compressed Gas Calibration Standard without Dilution
2.2.1 Applicability

This procedure may be used to assay the concentration of a candidate

compressed gas calibration standard, based on the concentration of a
compressed gas reference standard of the same gas mixture. This procedure
allows a specialty gas producer, a standard user, or other analytical laboratory to
certify that the assayed concentration for the candidate standard is traceable to
the reference standard. The procedure employs a pollutant gas analyzer to
compare the candidate and reference standards' concentrations by direct
measurement without dilution of either gas.

This procedure may be used for the assay of more than one candidate
standard during the same assay session. Criteria that apply to the assay of one
candidate standard apply to the assay of multiple candidate standards.

2.2.2 Limitations

The concentration of the candidate standard may be greater than or lesser

than the concentration of the reference standard. However, both concentrations
must lie within the well-characterized region of the multipoint calibration (see
Subsection 2.1.4.2). Additionally, the expanded uncertainty associated with each
assay of the candidate standard (i.e. Cell B248 times Cell B251 or Cell B333
times Cell B336 in the Appendix A spreadsheet) must be ≤±1 percent of the
reference standard concentration. This criterion may be more restrictive than the
corresponding criterion for the multipoint calibration, but it allows the analyst
greater flexibility in the selection of a reference standard for the assay of a
particular candidate standard while still keeping control of the uncertainty due to
the assay of the candidate standard. For example, assume that a 70-ppm
candidate standard is being assayed using a 50-ppm reference standard and that
the analyzer’s calibration was found to be well-characterized between 20 and 80
ppm. The expanded uncertainty for the example candidate standard’s estimated
concentration must be less than or equal to ±0.5 ppm.

The balance gas must be the same in both the candidate standard and the
reference standard, unless it has been demonstrated that the analyzer's
response is insensitive to differences in the balance gas composition.

2.2.3 Assay Apparatus

Figure 2-2 illustrates one possible design of the apparatus for the assay of
compressed gas calibration standards without dilution. This apparatus is
designed to allow the convenient routing of the gas mixtures to the pollutant gas
analyzer. Inert materials (e.g., Teflon®, stainless steel, borosilicate glass or

37
 EPA Traceability Protocol for Gaseous Calibration Standards

silanized glass) and clean, noncontaminating components should be used in

those portions of the apparatus that are in contact with the gas mixtures being
assayed.

Figure 2-2. Apparatus for the assay of calibration standards without dilution

The gas mixture to be measured is selected by rotation of two three-way valves

(i.e., V1 and V2). Pressure regulators and gas flow controllers (i.e., C1 and C2)
control the flow rates from the individual cylinders.

The gas flow controllers may be needle values, capillary tubes, thermal mass
flow controllers, or other flow control devices. The gas mixtures are routed to the
analyzer through a union tee tube fitting. Gas in excess of the analyzer's
demand is vented, which helps to ensure that the gas entering the analyzer is at
near-ambient pressure. Normally, the excess gas is vented to the atmosphere
without any obstructions in the tubing. However, the excess gas can be routed
through an uncalibrated rotameter by rotation of a three-way valve (i.e., V3). The

38
 EPA Traceability Protocol for Gaseous Calibration Standards

rotameter is used to demonstrate that the total gas flow rate exceeds the sample
flow rate of the analyzer and that no room air is being drawn in through the vent
line.

The apparatus may be modified in several ways that will not diminish its
performance. For example, the two three-way valves could be replaced by
solenoid valves, by a single four-way valve with three input ports and one output
port or by a rotating multiposition step valve. See also Subsection 2.1.2
concerning the acceptability of alternative designs for the assay apparatus.

2.2.4 Pollutant Gas Analyzer

The pollutant gas analyzer must have a well-characterized calibration

curve and must be capable of measuring directly the concentration of both the
candidate and the reference standards without dilution. See Subsection 2.1.4.1
concerning general analyzer calibration requirements and Subsection 2.1.4.1
concerning the well-characterized calibration curve. It must have good
resolution, good precision, a stable response, and low output signal noise. In
addition, the analyzer should have good specificity for the pollutant of interest so
that it has no detectable response to any other component or contaminant that
may be contained in either the candidate or reference standards. When
selecting an analyzer for the assay of a specific gas mixture, the analyst should
investigate potential analytical inferences between gas mixture components. If
any component in a multiple-component standard interferes with the assay of any
other component, the analyst must conduct an interference study to determine an
interference correction equation. If the candidate and reference standards
contain dissimilar balance gases (e.g., air versus nitrogen or different pro-
portions of oxygen in the balance air), it must have been demonstrated that the
analyzer's response is not sensitive to differences in the balance gas
composition. This demonstration can be accomplished by showing that no
difference exists in the analyzer's response when measuring a compressed gas
calibration standard that has been diluted with identical flow rates of different
balance gases.

The analyzer should be connected to a high-precision data acquisition

system (DAS), which must produce a numeric and graphic record of the
analyzer's response during the assay. The DAS should have four-digit resolution
for the numerical values of the response. More precise values will be obtained if
this system has a data-averaging capability. The specialty gas producer or other
laboratory that assayed the candidate standard must maintain laboratory records
for at least the standard's certification period or until such time as the standard is
blown down after being returned by the end user to the producer.

If the analyzer has not been in continuous operation, turn it on and allow it
to stabilize (e.g., for at least 12 hours) before beginning the measurements.

39
 EPA Traceability Protocol for Gaseous Calibration Standards

2.2.5 Assay Gases

2.2.5.1 Candidate Standard—

See Subsections 2.1.5 and 2.2.2.

2.2.5.2 Reference Standard—

See Subsection 2.1.3.2. If the multipoint calibration data have been fitted
to a linear (i.e., straight-line) regression model, then only a single reference
standard need be measured during the assay of the candidate standard. If these
data have been fitted to a quadratic or higher-order regression model, then at
least two reference standards must be measured. One reference standard is
adequate to determine the slope of a linear equation, but additional reference
standards are needed to determine the curvature of quadratic and higher-order
polynomial equations. The concentrations of the additional reference standards
should be located at the maximum difference between the polynomial curve and
the corresponding straight line between the zero gas and the highest-
concentration reference standard.

2.2.5.3 Zero Gas—

See Subsection 2.1.3.4. The analyst may substitute a low-concentration,
NIST-traceable reference standard for the zero gas during zero gas checks and
assays if there is reason to believe that the zero gas reading may not accurately
represent the zero-intercept of the calibration equation. Information concerning
the zero gas should be recorded in the laboratory's records.

2.2.6 Analyzer Calibration

2.2.6.1 Multipoint Calibration—

See Subsections 2.1.4.2 and 2.1.4.3.

2.2.6.2 Analyzer Range—

The range of the analyzer must include the concentrations of the zero gas,
the candidate standard and the reference standard. The concentrations of the
candidate and reference standards must fall within the well-characterized region
of the analyzer's calibration curve. In general, the analyst should use a range
that will produce the largest on-scale analyzer response.

2.2.6.3 Linearity—
The data reduction technique used in this procedure requires that the
analyzer have a well-characterized, but not necessarily linear, calibration curve
(see Subsection 2.1.4.2 concerning the well-characterized calibration curve).
High-concentration-range analyzers of the type that are required for this
procedure may not be inherently linear, but in such cases they usually have a
predictable, non-linear calibration curve that can be described by a polynomial
equation or can be mathematically transformed to produce a straight-line
calibration curve that is suitable for use in this procedure (see Subsection 2.1.4.3

40
 EPA Traceability Protocol for Gaseous Calibration Standards

concerning the transformation). Any such polynomial equation or mathematical

transformation should be verified during the multipoint calibration. Caution
should be exercised in using a transformed calibration curve because zero or
span control adjustments to the analyzer may produce unexpected effects in the
transformed calibration curve.

2.2.6.4 Zero and Span Gas Checks—

See Subsections 2.1.7.3 and 2.1.7.4. Prior to carrying out the assay of
the candidate standard, use zero and span gases to check for calibration drift in
the analyzer since the multipoint calibration. Zero gas and span gas checks
must be performed on any day after the multipoint calibration that candidate
standards are assayed. If multiple assays are being performed on the same
analyzer range, the analyst needs to perform only a single set of zero gas and
span gas checks for this range. However, another set must be performed if the
range is changed.

The gas mixtures to be used during the zero and span gas checks need
not be, but can be, the reference standards used for the assay of the candidate
standard or for the multipoint calibration. The reference standard for the span
gas check must be traceable to an SRM, an RGM, a PRM, a CRM, or an NTRM.
Information concerning this standard (e.g., cylinder identification number,
certified concentration, expanded uncertainty, certification expiration date, etc.)
must be recorded in the laboratory's records. The zero gas must meet the
requirements in Subsection 2.1.3.4. The use of NIST-traceable zero air material
for the zero gas check is recommended, but not required. The analyst may
substitute a low-concentration, NIST-traceable reference standard for the zero
gas if there is reason to believe that the zero gas reading may not accurately
represent the zero-intercept of the calibration equation.

Make three or more discrete measurements of the zero gas and three or
more discrete measurements of the span gas reference standard. "Discrete"
means that the analyst must change the gas mixture being sampled by the
analyzer between measurements. For example, the analyst might alternate
between measurements of the reference standard and measurements of the zero
gas. Record these measurements in the laboratory's records.

Next, verify that the analyzer's precision is acceptable. Calculate the

mean and standard deviation of the analyzer's responses to the zero gas.
Repeat the calculations for the reference standard measurements. These
calculations are performed in Step 6 of the spreadsheet described in Appendix A.
Record these calculations in the laboratory's records. The standard error of the
mean for each set of measurements must be less than or equal to 1.0 percent of
the mean response to the reference standard. That is,

s R
≤ RS
n 100

41
 EPA Traceability Protocol for Gaseous Calibration Standards

where
s = standard deviation of the analyzer's response;
n = the number of measurements of the gas mixture; and
RRS = the mean analyzer response to the reference standard.

The value of the standard error of the mean can be made smaller by
increasing the number of measurements. This calculation will enable the analyst
to determine how many replicate measurements are needed during the assay of
the candidate standard to obtain acceptable precision. The analyst may wish to
use a data logger or data acquisition system with averaging capability to obtain
more precise measurements. If the value of the standard error is not acceptable,
then the analyzer must be repaired or another analyzer must be used.

Next, verify that excessive zero drift has not occurred since the multipoint
calibration. For the zero gas measurements, calculate the relative difference (in
percent) between the current mean analyzer response during the zero gas check
and the corresponding response that is predicted from the multipoint calibration
regression equation. That is,

Relative Difference = 100 (Current Zero Response - Calibration Zero Response)

(Zero-Corrected Response for Reference Standard)

Note that the relative difference is calculated relative to the zero-corrected

response for the reference standard from the multipoint calibration. If the
reference standard was not measured during the multipoint calibration, use the
regression-predicted response for a concentration equal to that of this standard.

Repeat this relative difference calculation for the span gas reference
standard measurements. That is,

Relative Difference = 100 (Current Span Response - Calibration Span Response)

(Zero-Corrected Response for Reference Standard)

These relative difference calculations are performed in Step 6 of the

Appendix A spreadsheet. Record these calculations in the laboratory's records.

Then, if the relative differences for the zero and span gas checks are less
than or equal to 5.0 percent, the analyzer's current calibration is considered to be
approximately the same as during the multipoint calibration and the assay may
be conducted. The zero and span controls need not be adjusted and the assay
may be conducted. The data reduction technique does not require that some
calibration drift has not occurred since the multipoint calibration. Some minor
calibration drift is acceptable because the effect of any drift will be corrected
during the reduction of the assay data.

42
 EPA Traceability Protocol for Gaseous Calibration Standards

However, if the relative difference for either the zero or the span gas
check is greater than 5.0 percent, then the analyzer is considered to be out of
calibration. A new multipoint calibration may be conducted before the candidate
standard is assayed or the analyzer's zero and span controls may be adjusted to
return the analyzer's response to the original calibration levels. For some
analyzers such as nondispersive infrared instruments, daily changes in
environmental variables such as barometric pressure may shift the calibration.
After any adjustment of controls, the analyst should repeat the zero and span gas
checks and recalculate the relative differences to verify that the analyzer is
sufficiently in calibration. The analyzer will be considered to be out of calibration
if the relative differences remain greater than 5.0 percent.

2.2.7 Assay Procedure

1. Verify that the assay apparatus is properly configured, as described in

Subsection 2.2.3 and shown in Figure 2-1. Inspect the analyzer to verify that
it appears to be operating normally and that all controls are set to their
expected values. Record these control values in the laboratory's records.

2. Verify that a multipoint calibration has been performed within 1 month prior to
the assay date and that it demonstrates acceptable instrument performance
(see Subsection 2.1.4.2). Additionally, verify that the zero and span gas
checks indicate that the analyzer is in calibration (see Subsection 2.2.6.4).
Finally, verify that the concentrations of the candidate and reference
standards fall within the well-characterized region of the analyzer's calibration
curve (see Subsection 2.1.4.2).

3. Measure and adjust the flow rates of the gas mixtures (i.e., reference
standard(s), candidate standard, and zero gas) to approximately the same
value that will provide enough flow for the analyzer and sufficient excess to
assure that ambient air will not be drawn into the vent line.

4. In succession, measure the zero gas, the reference standard(s), and the
candidate standard(s) using the analyzer. Use valves V1 and V2 to select
each of the gas mixtures for measurement. Allow sufficient time between
measurements to completely purge the analyzer of the previous gas and for
the analyzer to achieve a stable reading. If the response for each
measurement is not stable, the precision of the measurements will decline
and the candidate standard may not be certifiable under this protocol. Record
the analyzer response for each measurement in the laboratory's records,
using the same response units (e.g., volts, area counts, etc.) as was used for
the multipoint calibration. At this point, do not convert these data into
concentration values using the calibration equation or perform any
mathematical transformations (see Subsection 2.1.4.3) if the data were so
transformed. These steps will be done later. Also, do not make any zero

43
 EPA Traceability Protocol for Gaseous Calibration Standards

control, span control, or other adjustments to the analyzer during these

measurements.

The analyst may assay multiple candidate standards during the same assay
session. For example, a single set of measurements may involve a zero gas,
a reference standard and three candidate standards. Criteria that apply to the
assay of one candidate standard apply to the assay of multiple candidate
standards. The analyst should be aware that the effect of any short-term
calibration drift will be greater when multiple candidate standards are
assayed. This greater effect is due to the longer period of time between
reference standard measurements. Unacceptable uncertainties of the
estimated concentrations for the candidate standards may occur as a result of
the longer assay session.

5. Conduct at least two additional sets of measurements, as described in step 4

above. However, for these subsequent sets of measurements, change the
order of the three measurements (e.g., measure reference standard, zero
gas, and candidate standard for the second set and measure zero gas,
candidate standard, and reference standard for the third set). Changing the
order that the gas mixtures are measured helps the analyst to discover any
effect of that one measurement has on subsequent measurements. The
number of sets of measurements will have been determined during analysis
of the multipoint calibration data such that the expanded uncertainty of the
regression-predicted concentration of the candidate standard is ≤1 percent of
the concentration of the reference standard.

6. If any one or more of the measurements of a set of measurements is invalid

or abnormal for any reason, discard all three measurements and repeat the
measurements. Such measurements may be discarded if the analyst can
demonstrate that the experimental conditions were inappropriate during these
measurements. Data cannot be discarded just because they appear to be
outliers but may be discarded if they satisfy statistical criteria for outliers.39 In
the case of outliers, all of the measurements do not have to be discarded.
Only the outliers should be discarded. As part of the laboratory records, the
analyst must record any discarded data and briefly explain in the laboratory's
records why these data were discarded.

7. The spreadsheets described in Appendices A and C (or an EPA-approved

equivalent statistical technique) must be used to calculate an overall
estimated concentration and the expanded uncertainty of the candidate
standard based on data from the assay measurements and from the
multipoint calibration. Record the estimated concentration and the expanded
uncertainty in the laboratory’s records.

If the assay date is different from the calibration date and if there is no
statistical difference between the two sets of data (i.e., the calibration data

44
 EPA Traceability Protocol for Gaseous Calibration Standards

and the zero and span gas check data), entering both sets of data in the Xi
and Yi columns in Step 1 of the Appendix A spreadsheet will produce a
smaller expanded uncertainty for the estimated concentration than if only the
calibration data were entered in these columns. However, a larger expanded
uncertainty will be produced if there is a statistical difference between the two
sets of data and if both are entered in these columns. The analyst should
check which approach yields the smaller value for the expanded uncertainty.

The Appendix A spreadsheet calculates the percentage of the uncertainty that

is due to the multipoint calibration. This percentage is needed for the total
uncertainty calculations in Appendix C when two or more assays fall under
the same multipoint calibration. Record this value in the laboratory’s records.

8. If the multipoint calibration data and the assay data underwent any
mathematical transformations before their statistical analysis, perform the
reverse transformations for the estimated concentration and the expanded
uncertainty. Record the transformed values in the laboratory’s records.

2.2.8 Stability Test for Newly Prepared Candidate Standards

See Subsection 2.1.5.2.

2.2.9 Certification Documentation

See Subsections 2.1.7 and 2.1.8.

2.2.10 Recertification Requirements

See Subsections 2.1.11 and 2.1.12.

45
 EPA Traceability Protocol for Gaseous Calibration Standards

2.3 Procedure G2: Assay and Certification of a

Compressed Gas Calibration Standard using Dilution
2.3.1 Applicability

This procedure may be used to assay the concentration of a diluted

candidate standard, based on the concentration of a diluted reference standard
of the same composition. It allows a specialty gas producer, an end user, or
other laboratory to certify that the assayed concentration for the candidate
standard is traceable to an NIST-traceable gaseous reference standard and to
NIST-certified flow rate, pressure, volume, mass or time reference standards.

Dilution of the candidate and reference standards with zero gas allows
greater flexibility in the range of concentrations of both the candidate and
reference standards that can be assayed. The procedure employs a low-
concentration-range (i.e., ambient air quality level) pollutant gas analyzer to
compare quantitatively diluted gas samples of both the candidate and reference
standards. This analyzer is more likely to have an inherently linear response
than a high-concentration-range analyzer. However, dilution introduces
additional error into the assay which would not be present if the standards were
assayed without dilution. This additional error is measured by an accuracy check
of the assay apparatus which is performed as part of the multipoint calibration.

This procedure may be used for the assay of multiple candidate standards
at the same time. Criteria that apply to the assay of one candidate standard
apply to the assay of multiple candidate standards.

2.3.2 Limitations

This procedure is not intended to be used for the assay and certification of
a GMIS that could be used as an analytical reference material in the assay of
candidate standards. Procedure G1 should be used for GMISs.

The concentration of the diluted candidate standard may be greater than

or lesser than the concentration of the diluted reference standard. However, both
concentrations must lie within the well-characterized region of the analyzer's
multipoint calibration (see Subsection 2.1.4.2).

Additionally, the magnitude of the expanded uncertainty for the estimated

concentration of the candidate standard must be ≤±1 percent of the reference
standard concentration. This criterion may be more restrictive than the
corresponding criterion for the multipoint calibration, but it allows the analyst
greater flexibility in the selection of a reference standard for the assay of a
particular candidate standard. For example, assume that a 70-ppm candidate
standard is being assayed using a 50-ppm reference standard and that the
analyzer’s calibration was found to be well characterized between 20 and 80

47
 EPA Traceability Protocol for Gaseous Calibration Standards

ppm. The expanded uncertainty for the example candidate standard’s estimated
concentration must be ≤±0.5 ppm.

An NIST-traceable system for flow measurement and gas dilution is

required (see Subsections 2.1.3 and 2.3.4).

The balance gas in both the candidate and reference standards must be
identical, unless either a high dilution flow rate ratio (i.e., at least 50 parts zero
gas to 1 part standard) is used for the assay or it has been demonstrated that the
analyzer is insensitive to differences in the balance gas.

2.3.3 Assay Apparatus

Dilution may be accomplished by using dynamic dilution methods or static

volumetric dilution methods. Examples of dynamic methods include capillary
tubes, critical orifices, volumetric piston pumps, and thermal mass-flow
controllers. An example static method involves filling an evacuated, fixed-volume
White (i.e., multipass) gas cell with measured pressures of a candidate or
reference standard and zero gas and then measuring the concentration of
resulting gas mixture using a Fourier transform infrared (FTIR) spectrometer.
This procedure mainly focuses on dynamic methods, but it can be applied to
static methods by substituting NIST-traceable pressure measurements for the
NIST-traceable flow rate measurements that are described in this procedure.

The components of the assay apparatus can be assembled in several

different configurations without diminishing performance. Two possible designs
of the assay apparatus are illustrated in Figures 2-3 and 2-4. The former figure
shows a configuration in which discrete components (i.e., three-way valves, gas
flow controllers, and a mixing chamber) are used to dilute the reference and
candidate standards. The latter figure shows a configuration in which a
commercially-available gas dilution system is used to dilute the standards. Both
designs share the important characteristic that the candidate standard is diluted
by the same components that dilute the reference standard. Inert materials (e.g.,
Teflon®, stainless steel, borosilicate glass or silanized glass) and clean,
noncontaminating components should be used in those portions of the apparatus
that are in contact with the gas mixtures being assayed.

48
 EPA Traceability Protocol for Gaseous Calibration Standards

Figure 2-3 Two designs for assay apparatus with dilution (Procedure G2)

49
 EPA Traceability Protocol for Gaseous Calibration Standards

In the upper diagram in Figure 2-3, either zero gas or a diluted standard
can be routed to the analyzer by rotation of three three-way values (i.e., V1, V2,
and V3). One gas flow controller (i.e., C1) regulates the flow rates of the
reference and candidate standards. These flow rates can be measured by a
single flowmeter connected to an outlet port on valve V2 or by a flowmeter built
into C1. Another gas flow controller (i.e., C2) regulates the flow rate of the zero
gas. This flow rate can be measured by a flowmeter connected to an outlet port
on valve V3 or by a flowmeter built into C2. The gas flow controllers may be
needle valves, capillary tubes, thermal mass flow controllers, or other suitable
devices (see Subsection 2.3.4). If different flow rates are used for the reference
and candidate standards during the assay (see Subsection 2.3.5), separate gas
flow controllers may be used for the two standards. However, the same
flowmeter must be used to measure the two flow rates to minimize error in the
measurement (see Subsection 2.3.4). Flow rates should be controlled and
measured with a relative uncertainty of 1 percent or less. For large dilutions of
the standards, the reference and candidate standard flow rates may be quite
small. Therefore, the internal volume of the tubing and components should be
kept small to minimize the flushing time when valve V1 is rotated. The mixing
chamber combines the two gas streams and should be designed to produce
turbulent flow to ensure thorough mixing of the gas streams. The diluted gas
mixtures are routed to the analyzer through a union tee tube fitting, which vents
excess gas flow. Normally, the excess gas is vented to the atmosphere without
any obstructions in the tubing and the gas entering the analyzer is at near-
atmospheric pressure. However, the excess gas can be routed through an
uncalibrated rotameter by rotation of a three-way valve (i.e., V4). The rotameter
is used to demonstrate that the total gas flow rate exceeds the sample flow rate
of the analyzer and that no room air is being drawn in through the vent line.

The assay apparatus may be modified in several ways that will not
diminish its performance. For example, the three-way valves could be replaced
by solenoid valves. Alternatively, valve V1 could be replaced by a single length
of tubing that is connected manually to the two standards' pressure regulators in
succession (see also Subsection 2.1.2 concerning the acceptability of alternative
designs for the assay apparatus.).

In the lower diagram in Figure 2-3, the reference and candidate standards
are diluted with a gas dilution system. This gas dilution system may use capillary
tubes, positive-displacement pumps, thermal mass flow controllers, or other
suitable devices to dilute the standards. If a gas dilution system is used, it must
have a specified accuracy of not greater than 1.0 percent of the undiluted
reference standard concentration. The analyst must check the accuracy of the
gas dilution system during the multipoint calibration by measurement of a check
standard (see Subsections 2.1.4.2 and 2.3.8.1).

50
 EPA Traceability Protocol for Gaseous Calibration Standards

The gas dilution system used in this procedure need not be certified as
NIST-traceable using the procedures given in Section 4 because this procedure
is adequate for establishing the traceability of candidate standards using dilution.

2.3.4 Flowmeters and Gas Dilution Systems for Dilution of Gas Mixtures

The flow rate/control reference standards or gas dilution systems must be

traceable to NIST primary flow rate, volume, mass or time reference standards
by original calibration by its manufacturer and by annual recertification at a NIST-
accredited state weights and measures laboratory or at a calibration laboratory
that is accredited by the National Voluntary Laboratory Accreditation Program
(NVLAP), the American Association for Laboratory accreditation (A2LA) or by the
International Laboratory Accreditation Conference (ILAC) under ISO/IEC 17025
(General Requirements for the Competence of Testing and Calibration
Laboratories).10,11,12 Alternatively, flow rate reference standards or gas dilution
systems may be traceable to the flow rate, volume, mass or time reference
standards of other national metrology institutes (e.g., KRISS in South Korea, NEL
in the United Kingdom, PTB in Germany) provided that these institutes are
participants in the Working Group for Fluid Flow (WGFF) of the International
Committee for Weights and Measures (CIPM).

The flow rate reference standards may be primary standards or secondary

standards. A primary standard is defined as a device or object used as the
reference in a calibration that is acknowledged to be of the highest metrological
quality and that derives its measurement without reference to some other
standard of the same quantity.

The reference standards' certifications must cover the range of flow rates
that will be used by the gas dilution systems and their uncertainties of the flow
rates must be documented in a calibration certificate. Their uncertainties must
be evaluated and expressed according to accepted metrological practice (see
Reference 9) and must not be greater than plus or minus 1 percent of set-point
for the flow rates that will be used for gas dilution. Uncertainties that are much
less than plus or minus 1 percent of reading are desirable so that the
uncertainties of the gas mixture concentrations are decreased. They must be
stable, repeatable, and linear and have good resolution.

Figure 2-2 shows flow measurement ports on valves V2 and V3. In this
configuration, a single flowmeter can be used to measure both the standard flow
rate and the zero gas flow rate. Such an approach would reduce measurement
errors arising from differences in the calibration of multiple flowmeters.
Alternatively, the reference standard and zero gas flow rates can be measured
separately at the outlet of the mixing chamber or the outlet of the gas dilution
system. In either case, a NIST-traceable volumetric flowmeter such as a wet test
meter, a thermal mass flowmeter, or a soap bubble flowmeter can be used (see
Subsection 2.1.3). Each flow rate must be measured separately while the other

51
 EPA Traceability Protocol for Gaseous Calibration Standards

flow rates are set to zero. Care must be exercised to ensure that each measured
flow rate remains constant when combined with the other flow rate(s) and
between the time of measurement and the time of the assay. Additionally, care
must be taken to ensure that the flowmeter does not cause any back pressure in
the gas stream and any resulting change in the flow rate through the flow
controller.

If in-line flowmeters are mounted directly downstream of the flow

controllers, they may not operate at atmospheric pressure because of back
pressure from downstream components. Also, this back pressure may vary as a
function of the total flow rate. Thus, the flowmeters must compensate for the
variable in-line pressure. Thermal mass flowmeters do not need to be corrected
for pressure effects. Measurements from pressure-sensitive flowmeters such as
rotameters or from volumetric flowmeters such as piston-type positive-
displacement flowmeters, wet test meters or soap-film flowmeters must be
carefully corrected for the actual gas pressure during the flow measurement. An
in-line flowmeter must not contaminate or react with the gas mixture passing
through it.

All volumetric flow-rate measurements must be corrected or referenced to

the same temperature and pressure conditions, such as EPA-standard conditions
(i.e., 760 millimeters of mercury (mm Hg), 25 °C) or the ambient temperature and
pressure conditions prevailing in the laboratory during the assay. Measurements
using wet test meters and soap-film flowmeters also must be corrected for the
saturation of the gas stream with water vapor in the moist interiors of these
flowmeters. The equation to correct the flow rate for temperature, pressure, and
humidity effects is given below:

Volume   PM − PWV   TS 
Flow rate =    
 Time   PS  TM 
where

PM = measured barometric pressure (mm Hg);

PWV = partial pressure of water vapor (mm Hg);
PS = standard pressure (mm Hg);
TS = standard temperature (298.2 K); and
TM = measured ambient temperature (273.2 + °C).

If using a flowmeter that does not use water or a soap bubble for measurement
the term PWV can be ignored.

Measurement of reference and candidate standard flow rates with the

same flowmeter and measurement of both dilution zero gas flow rates with the
same flowmeter tend to reduce measurement errors, associated with the use of
multiple flowmeters. These errors are more pronounced at higher dilution flow
rate ratios. Note that the impact of any flow measurement error is reduced if the

52
 EPA Traceability Protocol for Gaseous Calibration Standards

same dilution ratio can be used for both the reference standard and candidate
standard measurements.

2.3.5 Selection of Gas Dilution Flow Rates or Gas Concentration Settings

The flow rates or settings used for the zero gas, reference standard, and
candidate standard should be selected carefully to provide diluted concentrations
for both the candidate and reference standards that fall in the well-characterized
region of the analyzer's calibration curve. The diluted concentration of the
candidate standard may be greater than or lesser than the diluted concentration
of the reference standard. Any assay error due to the dilution process will be
reduced if the same dilution flow-rate ratio or concentration setting can be used
for both the candidate and reference standards. Select the diluted
concentrations of the reference and candidate standards, and select flow rates or
concentration settings that will produce the highest analyzer responses within the
well-characterized region of the analyzer's calibration curve. If possible, select
flow rates or a flowmeter range such that the flow rates to be measured fall in the
upper half of the range.

If the same dilution flow-rate ratio or concentration setting cannot be used

for both the candidate and reference standards, select different ratios or settings
for the candidate and reference standards to produce concentrations that are
approximately equal and that fall in the well-characterized region of the
analyzer's calibration curve. Select flow rates or settings such that only one of
the apparatus controls must be adjusted when switching from the reference
standard to the candidate standard, or vice versa. Where a choice of analyzer
ranges is available, higher dilution ratios or lower concentration settings will
reduce the consumption of the standards.

2.3.6 Pollutant Gas Analyzer

The pollutant gas analyzer must have a well-characterized calibration

curve and must have a range that is capable of measuring the diluted
concentration of both the candidate and the reference standards (see Subsection
2.1.4.1 concerning general analyzer calibration requirements). It must have good
resolution, good precision, a stable response, and low output signal noise. A
suitable analyzer with acceptable performance specifications may be selected
from the list of EPA-designated reference and equivalent method analyzers.40 In
addition, the analyzer should have, but is not required to have, good specificity
for the pollutant of interest so that it has no detectable response to any other
component or contaminant that may be contained in either the candidate or
reference standards. When selecting an analyzer for the assay of a specific gas
mixture, the analyst should investigate potential analytical inferences between
gas mixture components. If any component in a multiple-component standard
interferes with the assay of any other component, the analyst must conduct an
interference study to determine an interference correction equation. If the

53
 EPA Traceability Protocol for Gaseous Calibration Standards

candidate and reference standards contain dissimilar balance gases (e.g., air
versus nitrogen or different proportions of oxygen in the balance air), either a
high dilution flow-rate ratio (i.e., at least 50 parts zero gas to 1 part standard)
should be used or it must have been demonstrated that the analyzer's response
is not sensitive to differences in the balance gas composition. This
demonstration may be accomplished by showing that no difference exists in an
analyzer's response when measuring a compressed gas calibration standard that
has been diluted with identical flow rates of different balance gases.

The analyzer should be connected to a high-precision data acquisition

system, which must produce an electronic record of the analyzer's response
during the assay. A high-precision digital panel meter, a digital voltmeter, a data
logger or some other data acquisition system with four-digit resolution can be
used to obtain numerical values of the analyzer's response. More precise values
will be obtained if this system has a data-averaging capability. The specialty gas
producer or other laboratory that assayed the candidate standard must maintain
laboratory records for at least the standard's certification period or until such time
as the standard is blown down after being returned by the final user to the
producer.

If the analyzer has not been in continuous operation, turn it on and allow it
to stabilize (e.g., for at least 12 hours) before beginning any measurements.

2.3.7 Assay Gases

2.3.7.1 Candidate Standard—

See Subsections 2.1.5, 2.3.2, and 2.3.5.

2.3.7.2 Reference Standard—

See Subsections 2.1.3, 2.1.12, 2.3.2, and 2.3.5. The reference standard
used for the assay of the candidate standard must be an SRM, an RGM, a PRM,
a CRM, an NTRM or a GMIS. This standard need not be the same as any of the
reference standards used for the span gas check or for the multipoint calibration.
Information concerning the reference standard (e.g., cylinder identification
number, certified concentration, certification expiration date, etc.) must be
recorded in the laboratory’s records.

If the multipoint calibration data have been fitted to a linear (i.e., straight-
line) model, then only a single reference standard need be measured during the
assay of the candidate standard. If these data have been fitted to a quadratic or
higher-order polynomial model, then at least two reference standards must be
measured. One reference standard is adequate to determine the slope of a
linear equation, but additional reference standards are needed to determine the
curvature of quadratic or higher-order polynomial equations. The concentrations
of the additional reference standards should be located at the maximum

54
 EPA Traceability Protocol for Gaseous Calibration Standards

difference between the polynomial curve and the corresponding straight line
between the zero gas and the highest-concentration reference standard.

2.3.7.3 Zero Gas—

The zero gas used for dilutions must be clean, dry, zero-grade air or
nitrogen containing no detectable concentration of the pollutant of interest (see
Subsection 2.1.3.5). Use the same zero gas for dilution of both candidate
standards and reference standards. The use of an NIST-traceable zero air
material for the zero gas check is recommended, but not required (see
Subsection 2.4). The analyst may substitute a low-concentration, NIST-traceable
reference standard for the zero gas in zero gas checks and assays if there is
reason to believe that the zero gas reading may not accurately represent the
zero-intercept of the calibration equation. Information concerning the zero gas
must be recorded in the laboratory's records.

2.3.8. Analyzer Calibration

2.3.8.1 Multipoint Calibration—

See Subsections 2.1.4.2 and 2.1.4.3. Following completion of the
multipoint calibration, the accuracy of the assay apparatus must be checked to
verify that the error associated with the dilution is not excessive. This accuracy
check involves the measurement of an undiluted or diluted check standard. The
check standard must be an SRM, an RGM, a PRM, a CRM, an NTRM or a GMIS
as specified in Subsection 2.1.3.1. It must have a certified concentration that is
different from that of the reference standard used in the multipoint calibration.
Information concerning this standard (e.g., cylinder identification number,
certified concentration) must be recorded in the laboratory's records.

If an undiluted check standard is used, its concentration must fall in the

well-characterized region of the calibration curve. If a diluted check standard is
used, the diluted concentration must fall in the well-characterized region. The
point on the calibration curve at which the dilution-related uncertainty is checked
will depend on the undiluted reference standards that are available to be used for
the accuracy check.

Make three or more discrete measurements of the undiluted or diluted

check standard. "Discrete" means that the analyst must change the gas mixture
being sampled by the analyzer between measurements. For example, the
analyst might alternate between measurements of the check standard and the
zero gas. Record these measurements in the laboratory's records.

Next the analyst must verify that the dilution error is not excessive. For
the check standard measurements, calculate the relative difference (in percent)
between the mean analyzer response to the undiluted or diluted check standard
and the corresponding response that is predicted from the multipoint calibration

55
 EPA Traceability Protocol for Gaseous Calibration Standards

regression equation and the undiluted or diluted check standard concentration.

That is,

Relative Difference = 100 (Mean Analyzer Response - Predicted Response)

(Predicted Response)

If the relative difference is greater than 1.0 percent, the dilution error is
considered to be excessive. The analyst must investigate why the relative
difference is excessive. The problem may be due to errors in the reference
standard and check standard concentrations, errors in assay apparatus or to
some other source. Assays may not be conducted until the relative difference for
a subsequent accuracy check is less than or equal to 1.0 percent.

2.3.8.2 Analyzer Range—

The range of the analyzer must include the concentrations of the zero gas
and of the diluted candidate and reference standards (see Subsection 2.3.5).
The concentrations of the diluted reference and candidate standards must fall
within the well-characterized region of the analyzer's calibration curve. Because
the selection of the dilution ratio or ratios to be used in the assay provides great
flexibility in the choice of concentrations to be measured by the analyzer, the
analyzer range should be selected based on optimum accuracy, stability, and
linearity.

2.3.8.3 Linearity—
The data reduction technique used in this procedure requires that the
analyzer have a well-characterized, but not necessarily linear, calibration curve
(see Subsection 2.1.4.2). Many lower-concentration analyzers of the type that
may be used for this procedure have straight-line calibration curves. If not, they
usually have a predictable nonlinear calibration curve that can be described by a
polynomial equation or can be mathematically transformed to produce a straight-
line calibration curve suitable for use in this procedure. Any such polynomial
equation or mathematical transformation should be verified during the multipoint
calibration. Caution should be exercised in using a transformed calibration curve
because zero or span control adjustments to the analyzer may produce
unexpected effects in the transformed calibration curve.

2.3.8.4 Zero and Span Gas Checks—

See Subsections 2.1.3.2 and 2.1.3.3. Prior to carrying out the assay of
the candidate standard, use zero and span gases to check for calibration drift in
the analyzer since the multipoint calibration. Zero gas and span gas checks
must be performed on any day after the multipoint calibration that candidate
standards are assayed. If multiple assays are being performed on the same
analyzer range, the analyst needs to perform only a single set of zero gas and
span gas checks. However, another set must be performed if the range is
changed.

56
 EPA Traceability Protocol for Gaseous Calibration Standards

The gas mixtures to be used during the zero and span gas checks need
not be the same as any of the reference standards used for the assay of the
diluted candidate standard or for the multipoint calibration. The reference
standard for the span gas check must be traceable to an SRM, an RGM, a PRM,
a CRM, or an NTRM. The zero gas must meet the requirements in Subsection
2.1.3.4. The use of NIST-traceable zero air material for the zero gas check is
recommended, but not required. Information concerning these standards (e.g.,
cylinder identification number, certified concentration, uncertainty, certification
expiration date, etc.) must be recorded in the laboratory's records.

Make three or more discrete measurements of the zero gas and three or
more independent measurements of the diluted reference standard. Record
these measurements in the laboratory's records.

Next, the analyst must verify that the analyzer's precision is acceptable.
Calculate the mean and standard deviation of the analyzer's response to the zero
gas. Repeat these calculations for the diluted reference standard
measurements. These calculations are performed in Step 6 of the spreadsheet
described in Appendix A. Record these calculations in the laboratory's records.
The standard error of the mean for each set of measurements must be less than
or equal to 1.0 percent of the mean response to the diluted reference standard.
That is,
S R
≤ DRS
n 100
where

s = standard deviation of the analyzer's response;

n = the number of measurements of the gas mixture; and
RDRS = the mean analyzer response to the diluted reference standard.

The value of the standard error of the mean can be made smaller by
increasing the number of measurements. This calculation will enable the analyst
to determine how many replicate measurements are needed during the assay of
the diluted candidate standard to obtain acceptable precision. The analyst may
wish to use a data logger or data acquisition system with data averaging
capability to obtain more precise measurements. If the value of the standard
error of the mean is not acceptable, then the analyzer must be repaired or
another analyzer must be used for the assay.

Next the analyst must verify that excessive calibration drift has not
occurred since the multipoint calibration. For the zero gas measurements,
calculate the relative difference (in percent) between the mean analyzer
response during the zero gas check and the corresponding response that is
predicted from the multipoint calibration regression equation. That is,

57
 EPA Traceability Protocol for Gaseous Calibration Standards

Relative Difference = 100 (Current Zero Response - Calibration Zero Response)

(Calibration Response for Diluted Reference Standard)

Note that the relative difference is calculated relative to the zero-corrected

response for the diluted reference standard from the multipoint calibration. If the
diluted reference standard was not measured during the multipoint calibration,
use the regression-predicted response for a concentration equal to that of this
standard.

Repeat this relative difference calculation for the diluted reference

standard measurements. That is,

Relative Difference = 100 (Current Span Response - Calibration Span Response)

(Calibration Response for Diluted Reference Standard)

These relative difference calculations are performed in Step 6 of the

Appendix A spreadsheet. Record these calculations in the laboratory's records.

Then, if the relative differences for the zero and span checks are less than
or equal to 5.0 percent, the analyzer's current calibration is considered to be
approximately the same as during the multipoint calibration and the assay may
be conducted. The zero and span controls need not be adjusted and the assay
may be conducted. The data reduction technique does not require that some
calibration drift has not occurred since the multipoint calibration. Some minor
calibration drift is acceptable because the drift will be corrected for during the
reduction of the assay data.

However, if the relative difference for either the zero or the span gas
checks is greater than 5.0 percent, then the analyzer is considered to be out of
calibration. A new multipoint calibration may be conducted before the candidate
standard is assayed or the analyzer's zero and span controls may be adjusted to
return the analyzer's response to the original calibration levels. For some
analyzers such as nondispersive infrared instruments, daily changes in
environmental variables such as barometric pressure may shift the calibration.
After any adjustment of the controls, the analyst should repeat the zero and span
gas checks and recalculate the relative differences to verify that the analyzer is
sufficiently in calibration. The analyzer will be considered to be out of calibration
if the relative differences remain greater than 5.0 percent.

The zero gas and diluted reference standard measurements that are
performed for the assay of the diluted candidate standard may also be used for
the zero gas and span gas checks.

58
 EPA Traceability Protocol for Gaseous Calibration Standards

2.3.9 Assay Procedure

1. Verify that the assay apparatus is properly configured as shown in Figure 2-3
or Figure 2-4 and as described in Subsection 2.3.3. Inspect the analyzer to
verify that it appears to be operating normally and that all controls are set to
their expected values. Record these values in the laboratory's records.

2. Verify that the calibrations of the flow rate/control reference standards or gas
dilution systems have not drifted significantly since their last calibration or
recertification. This verification need not be done more frequently than
monthly, but that it must have been done within a month before the assay.
Use an NIST-traceable flow rate/control reference standard to check at least
one flow rate setting for each pollutant and dilution gas stream in the assay
apparatus. Record the indicated and actual flow rates in the laboratory's
records. Calculate the relative difference (in percent) between the indicated
and actual volumes or flow rates. That is,

Relative Difference = 100 (Indicated Flow Rate - Actual Flow Rate)

(Actual Flow Rate)

If the relative differences for the pollutant and diluent flow rates are less than
1.0 percent, the calibrations of the flow rate/control reference standards or
gas dilution systems have not drifted significantly since their last calibration or
recertification. The assay can proceed. If the relative differences for either
the pollutant flow rate or the diluent flow rate exceed than 1.0 percent, then
significant drift has occurred and the device should be recertified before the
assay is conducted.

3. Verify that a multipoint calibration of the analyzer has been performed within 1
month prior to the assay date and that it demonstrates acceptable instrument
performance and that the dilution error that is not excessive (see Subsections
2.1.4.2, 2.3.6, and 2.3.8.1). Additionally, verify that the zero and span gas
checks indicate that the analyzer is in calibration (see Subsection 2.3.8.4).
Finally, verify that the concentrations of the diluted reference and candidate
standards fall within the well-characterized region of the analyzer's calibration
curve (see Subsection 2.1.4.2).

4. Determine and establish the flow rates or concentration settings of the gas
mixtures (i.e., reference standard(s), candidate standard, and zero gas) that
will be used for the assay (see Subsections 2.3.4, 2.3.5, and 2.3.8.2). Also
check that the total flow rate coming from the mixing chamber will provide
enough flow for the analyzer and sufficient excess to ensure that no ambient
air will be drawn into the vent line. Changes in the sample pressure may
change the calibration curve. When using the same flow rates for both
candidate and reference standards, carefully set the delivery pressures of the

59
 EPA Traceability Protocol for Gaseous Calibration Standards

two standards' pressure regulators to the same value so that there is no

change in the flow rate when switching from one standard to the other.

Calculate the diluted reference standards' concentration using the following

equation:

Diluted Standard Conc. = (Undiluted Standard Conc.)(Standard Flow Rate)

(Standard Flow Rate + Zero Gas Flow Rate)

Record the measured flow rates and the undiluted and diluted reference
standard concentrations in the laboratory's records.

5. In succession, measure the zero gas, the diluted reference standard(s) and
the diluted candidate standard using the analyzer. For each measurement,
adjust the flow rates, if necessary, to those determined in Step 4, and allow
sufficient time between measurements to completely purge the analyzer of
the previous gas and for the analyzer to achieve a stable reading. If the
reading is not stable, the precision of the measurements will decline and the
candidate standard might not be certifiable under this protocol. Record the
analyzer response for each measurement, using the same response units
(e.g., volts, area counts, etc.) as was used for the multipoint calibration. At
this point, do not convert the data into concentration values using the
calibration equation. Do not perform any mathematical transformations of the
data. These steps will be done later. Do not make any zero control, span
control, or other adjustments to the analyzer during this set of measurements.
Record these analyzer responses in the laboratory's records.

The analyst may assay multiple candidate standards during the same assay
session. For example, a single set of measurements may involve a zero gas,
a diluted reference standard, and three diluted candidate standards. Criteria
that apply to the assay of one candidate standard apply to the assay of
multiple candidate standards. The analyst should be aware that the effect of
any short-term calibration drift will be greater when multiple candidate
standards are assayed. This greater effect is due to the longer period of time
between reference standard measurements. Unacceptable uncertainties of
the estimated concentrations for the diluted candidate standards may occur
as a result of the longer assay session.

6. Conduct at least two additional sets of measurements, as described in step 5

above. However, for these subsequent sets of measurements, change the
order of the three measurements (e.g., measure the reference standard, zero
gas, and candidate standard for the second set and measure the zero gas,
candidate standard, and reference standard for the third set, etc.). Changing
the order that the gas mixtures are measured helps the analyst to discover
any effect that one measurement has on subsequent measurements. The
number of sets of measurements will have been determined during the

60
 EPA Traceability Protocol for Gaseous Calibration Standards

analysis of the multipoint calibration data such that the expanded uncertainty
for the regression-predicted concentration of the candidate standard is ≤1
percent of the concentration of the reference standard.

7. If any one or more of the measurements of a set of measurements is invalid

or abnormal for any reason, discard all three measurements and repeat the
measurements. Such measurements may be discarded if the analyst can
demonstrate that the experimental conditions were inappropriate during these
measurements. Data cannot be discarded just because they appear to be
outliers but may be discarded if they satisfy statistical criteria for outliers.39 In
the case of outliers, all of the measurements do not have to be discarded.
Only the outliers should be discarded. As part of the laboratory records, the
analyst must record any discarded data and briefly explain in the laboratory's
records why these data were discarded.

8. The spreadsheets described in Appendices A and C (or an EPA-approved

equivalent statistical technique) must be used to calculate an overall
estimated concentration and the expanded uncertainty of the candidate
standard based on data from the assay measurements and from the
multipoint calibration. Record the estimated concentration and the expanded
uncertainty in the laboratory’s records.

If the assay date is different from the calibration date and if there is no
statistical difference between the two sets of data (i.e., the calibration data
and the zero and span gas check data), entering both sets of data in the Xi
and Yi columns in Step 1 of the Appendix A spreadsheet will produce a
smaller expanded uncertainty for the estimated concentration than if only the
calibration data were entered in these columns. However, a larger expanded
uncertainty will be produced if there is a statistical difference between the two
sets of data and if both are entered in these columns. The analyst should
check which approach yields the smaller value for the expanded uncertainty.

The Appendix A spreadsheet calculates the percentage of the uncertainty that

is due to the multipoint calibration. This percentage is needed for the total
uncertainty calculations in Appendix C when two or more assays fall under
the same multipoint calibration. Record this value in the laboratory’s records.

The uncertainty associated with the dilution of the reference standard and the
candidate standard needs to be included in the calculation of the expanded
uncertainty in those cases in which the gas dilution apparatus' settings are
different for the reference standard than for the candidate standard.
However, this uncertainty component does not have to be included in the
calculation of the expanded uncertainty if exactly the same settings for the
gas dilution apparatus are used for the analysis of the reference standard and
the candidate standard. In this case, no variability in the results is introduced
by using a constant dilution setting.

61
 EPA Traceability Protocol for Gaseous Calibration Standards

9. If the multipoint calibration data and the assay data underwent any
mathematical transformations before their statistical analysis, perform the
reverse transformations for the estimated concentration and the expanded
uncertainty. Record the transformed values in the laboratory’s records.

10. Finally, the certified undiluted concentration for a candidate standard

containing an unreactive gas mixture and requiring only a single assay can be
calculated from the mean concentration of the diluted candidate standard as
follows:

Certified Undiluted Conc. = (Mean Diluted Conc) (Total Gas Flow Rate)
(Standard Gas Flow Rate)

where Total Gas Flow Rate = Standard Gas Flow Rate + Zero Gas Flow Rate

2.3.10 Stability Test for Newly Prepared Standards

See Subsection 2.1.5.2.

2.3.11 Certification Documentation

See Subsections 2.1.7 and 2.1.8.

2.3.12 Recertification Requirements

See Subsection 2.1.11.

62
 EPA Traceability Protocol for Gaseous Calibration Standards

2.4 Procedure G3: Assay and Certification of a Zero Air

Material as Conforming to 40 CFR Part 72.2
2.4.1 Applicability

This procedure may be used to measure whether the impurity

concentrations of a candidate zero air material are less than the impurity
specifications of 40 CFR Part 72.27 (i.e., a calibration gas whose concentrations
of SO2, NOX, and THC are not greater than 0.1 ppm; whose concentration of CO
is not greater than 1 ppm; and whose concentration of CO2 is not greater than
400 ppm). Under 40 CFR Part 72.2, a zero air material means a calibration gas
certified by a specialty gas producer, ambient air conditioned and purified by a
continuous emission monitoring system (CEMS), conditioned and purified
ambient air provided by a conditioning system concurrently supplying dilution air
to the CEMS or a multicomponent mixture certified by the supplier of the mixture
that the concentration of the component being zeroed is less than or equal to the
applicable concentration specification. This procedure does not determine an
overall estimated concentration and an expanded uncertainty (U) for the zero air
material. Rather, it is a go/no-go check for the analyzer response from a
candidate zero air material. The data from the multipoint calibration and the
assay measurements are used to predict an analyzer response that would be
produced by the maximum allowable impurity concentration. That analyzer
response minus the uncertainty of the measurements establishes a go/no-go
acceptance criterion for the analyzer response that demonstrates conformance
with the specification. For example, if the predicted analyzer response for 1 ppm
CO is 50 with an expanded uncertainty of +/-2, then a candidate zero air material
whose reading is below 48 passes the go/no-go acceptance criterion and
conforms to the CO concentration specification. These statistical calculations are
performed by the Appendix E spreadsheet. See Reference 50 for information
about statistical conformance with specifications.

This procedure allows a specialty gas producer, a standard user or other

analytical laboratory to determine whether a candidate zero air material meets
the Part 72.2 impurity specifications. It is analytically simpler than Procedures
G1 and G2. It employs a pollutant gas analyzer to compare the candidate's and
reference standards' contents by direct measurement. Only a single
measurement of each candidate is required. The go/no-go check allows the
analyst to make an immediate assessment (as soon as a stable reading is
obtained) of the candidate's conformance with the go/no-go acceptance criterion,
rather than having to use the Appendix E spreadsheet to make a separate
calculation for each candidate zero air material being assayed in the same
analytical session.

The reference standards used for assaying a candidate under this

procedure are an NIST-traceable zero air material and an NIST-traceable
upscale reference standard, both of which must be a SRM, an NTRM, an RGM, a

63
 EPA Traceability Protocol for Gaseous Calibration Standards

PRM, a CRM or a GMIS. It employs a pollutant gas analyzer to compare the

candidate's and reference standards' contents by direct measurement.

This procedure may be used for the assay of more than one candidate
during the same assay session. Criteria that apply to the assay of one candidate
apply to the assay of multiple candidates.

At this time, EPA does not require the regulated community to use NIST-
traceable zero air materials for the calibration of ambient air or continuous
emission monitors that are required by 40 CFR Parts 50, 58, 60, and 75.
However, end users may elect to purchase these standards from specialty gas
producers who elect to assay and certify them.

2.4.2 Limitations

This procedure is not intended to be used for the assay and certification of
the impurity concentrations of bulk zero air material such as could be used to fill
individual cylinders that might be sold as NIST-traceable zero air materials.
Rather, it is intended to be used for the assay of the impurity concentrations of
candidate zero air materials in individual cylinders because the concentrations in
these cylinders may differ from those in the bulk gas because of residual
impurities that were in the cylinders before filling.

Because of the simplified analytical and statistical techniques that are

employed in this procedure, zero air materials cannot be certified as containing
specific impurity concentrations using this procedure. Rather, they can be
certified as conforming to the impurity specifications of Part 72.2.

This procedure is not intended to be used for the assay and certification of
a zero air GMIS that could be used as a reference standard in the assay of
candidate zero air materials. Procedure G1 should be used for the assay and
certification of a candidate zero air GMIS.

2.4.3 Assay Apparatus

Figure 2-2 illustrates one possible design of apparatus for the assay of
candidate zero air materials without dilution. See Subsections 2.1.3 and 2.2.3 for
a discussion of this apparatus. The apparatus may be modified in several ways
that will not diminish its performance.

2.4.4 Pollutant Gas Analyzer

See Section 2.2.4 for the requirements for the pollution gas analyzer. Use
the most sensitive operating range that allows measurement of the zero gas, the
candidate zero air material, the upscale reference standard, and the reference
standards used in the multipoint calibration. The detection limit for the analyzer

64
 EPA Traceability Protocol for Gaseous Calibration Standards

must be less than the go/no-go acceptance criterion. If the analyzer has not
been in continuous operation, turn it on and allow it to stabilize (e.g., for at least
12 hours) before beginning any measurements.

2.4.5 Analyzer Calibration

2.4.5.1 Multipoint Calibration—

See Subsections 2.1.4.2, 2.1.4.3, and 2.3.5.1 for the requirements for the
pollution gas analyzer's multipoint calibration. The analyzer used for the assay
must have had a multipoint calibration within one month prior to the assay date
that demonstrates acceptable instrument performance. The reference standards
used in the multipoint calibration must include an NIST-traceable zero air
material. The reference standards may be diluted with an NIST-traceable zero
air material for the multipoint calibration.

Note that the uncertainty of the calibration curve may comprise a larger
fraction of the expanded uncertainty for a candidate zero air material than for a
candidate standard whose concentration is near the upper end of the analyzer's
range. The inclusion of more measurements or more concentrations in the
multipoint calibration will reduce the uncertainty of the calibration curve, which
may allow a clearer determination of whether a candidate zero air material
conforms to the impurity specifications of Part 72.2.

Note that an analyzer whose calibration equation on a less-sensitive range

can be easily fitted to a straight-line regression model may need to be fitted to a
quadratic or higher-order regression model when a more sensitive range is used
to assay candidate zero air materials. The calibration curve may have a hook at
lower concentrations. If an incorrect regression model is used for such data, the
statistical analysis of these data may produce incorrect estimates of impurity
concentrations. To investigate this possibility, include multiple reference
standards at the low end of the range during the multipoint calibration.

2.4.5.3 Linearity—
See Subsection 2.3.5.3 for a discussion of the linearity of the pollution gas
analyzer's calibration curve.

2.4.6 Assay Gases

2.4.6.1 Candidate Zero Air Material—

See Subsections 2.1.6 and 2.2.2.

2.4.6.2 Upscale Reference Standard—

See Subsections 2.1.2 and 2.2.2. The upscale reference standard used
for the assay of the candidate zero air material must be an SRM, an RGM, a
PRM, a CRM, an NTRM or a GMIS. This standard need not be, but can be, one
of the reference standards used for the multipoint calibration. It may be diluted

65
 EPA Traceability Protocol for Gaseous Calibration Standards

with an NIST-traceable zero air material for the assay of the candidate zero air
material. Information concerning the standard (e.g., cylinder identification
number, certified concentration, uncertainty, certification expiration date, etc.)
must be recorded in the laboratory's records.

If the multipoint calibration data have been fitted to a linear (i.e., straight-
line) regression model, then only a single upscale reference standard need be
measured during the assay of the candidate zero air material. If these data have
been fitted to a quadratic or higher-order regression model, then at least two
upscale reference standards must be measured during the assay. One upscale
reference standard is adequate to determine the slope of a linear equation, but
additional upscale reference standards are needed to determine the curvature of
quadratic and higher-order polynomial equations. The concentrations of the
additional upscale reference standards should be located at the maximum
difference between the polynomial curve and the corresponding straight line
between the zero gas and the highest-concentration upscale reference standard.

2.4.6.3 Zero Gas—

The use of a NIST-traceable zero air material is required (see Section
2.1.3.3). It must be an SRM, an RGM, a PRM, a CRM, an NTRM or a GMIS. It
need not be, but can be, the zero gas that is used for the multipoint calibration.
Information concerning the zero gas (e.g., cylinder identification number, certified
concentration, uncertainty, etc.) must be recorded in the laboratory's records.

2.4.7 Assay Procedure

1. Verify that the assay apparatus is properly configured, as described in

Subsection 2.2.3 and shown in Figure 2-1. Inspect the analyzer to verify that
it appears to be operating normally and that all controls are set to their
expected values. Record these control values in the laboratory's records.

2. Verify that a multipoint calibration has been performed within 1 month prior to
the assay date and that demonstrates acceptable instrument performance
(see Subsections 2.1.4.2 and 2.1.4.3). Additionally, verify that the
concentrations of the candidate zero air material and the reference standards
fall within the well-characterized region of the analyzer's calibration curve.

3. Measure and adjust the flow rates of the gas mixtures (i.e., upscale reference
standard(s), candidate zero air material, and zero gas) to approximately the
same value that will provide enough flow for the analyzer and sufficient
excess to assure that no ambient air will be drawn into the vent line.

4. Make three or more discrete measurements of the zero gas, three or more
discrete measurements of the upscale reference standard, and at least one
measurement of the candidate zero air material(s). Use Valves V1 and V2 to
select each of the gases for measurement. "Discrete" means that the analyst

66
 EPA Traceability Protocol for Gaseous Calibration Standards

must change the gas being sampled by the analyzer between measurements.
The analyst may assay multiple zero air materials during the same assay
session. For example, the analyst might cycle through measurements of the
upscale reference standard, the zero gas, the first candidate zero air material,
and then repeat this cycle substituting in second and third candidate zero air
materials. Allow sufficient time between measurements to completely purge
the analyzer of the previous gas and for the analyzer to achieve a stable
reading. Changing the order that the gas mixtures are measured helps the
analyst to discover any effect of that one measurement may have on
subsequent measurements. For each discrete measurement, allow ample
time for the analyzer to achieve a stable response. If the response for each
measurement is not stable, the precision of the measurements will decline
and the candidate standard may not be certifiable under this protocol.

Criteria that apply to the assay of one candidate standard apply to the assay
of multiple candidate standards. The analyst should be aware that the effect
of any short-term calibration drift will be greater when multiple candidate zero
air materials are assayed. This greater effect is due to the longer period of
time between reference standard measurements.

5. Record the analyzer response for each measurement in the laboratory's

records, using the same response units (e.g., volts, area counts, etc.) as was
used for the multipoint calibration. At this point, do not convert these data into
concentration values using the calibration equation. Do not perform any
necessary mathematical transformation at this point. The conversion and
transformation will be done later. Do not make any zero control, span control,
or other adjustments to the analyzer during these measurements.

6. If any one or more of the measurements is invalid or abnormal for any reason,
discard all three measurements and repeat the measurements. Such
measurements may be discarded if the analyst can demonstrate that the
experimental conditions were inappropriate during these measurements.
Data cannot be discarded just because they appear to be outliers but may be
discarded if they satisfy statistical criteria for testing outliers.39 The analyst
must record any discarded data and a brief explanation as to why these data
were discarded in the laboratory's records.

7. Verify that the analyzer's precision is acceptable. Calculate the mean and
standard deviation of the analyzer's responses to the zero gas. Repeat the
calculations for the analyzer's responses to upscale reference standard.
These calculations are performed in Step 5 of the spreadsheet described in
Appendix E. Record these calculations in the laboratory's records. The
standard error of the mean for each set of measurements must be less than
or equal to 1.0 percent of the mean response to the upscale reference
standard. That is,

67
 EPA Traceability Protocol for Gaseous Calibration Standards

S RURS
≤
n 100
where
s = standard deviation of the analyzer's response;
n = the number of measurements of the zero gas; and
RURS = the zero-corrected mean analyzer response to the upscale
reference standard.

If the value of the standard error of the mean is not acceptable, then the
analyzer must be repaired or another analyzer must be used for the assay.

8. Verify that excessive calibration drift has not occurred since the multipoint
calibration. For the zero gas measurements, calculate the relative difference
(in percent) between the current mean analyzer response during the assay
and the corresponding response that is predicted from the multipoint
calibration regression equation. That is,

Relative Difference = 100 (Assay Zero Response - Calibration Zero Response)

(Zero-Corrected Upscale Reference Standard Response)

This calculation is performed in Step 5 of the Appendix E spreadsheet. Note

that the relative difference is calculated relative to the zero-corrected
response for the upscale reference standard from the multipoint calibration. If
this standard was not measured during the multipoint calibration, use the
regression-predicted response for a concentration equal to that of this
standard.

Repeat this relative difference calculation for the upscale reference standard
measurements. Record these calculations in the laboratory's records.

Then, if the relative differences for the zero gas and upscale reference
standard checks are less than or equal to 5.0 percent, the analyzer's current
calibration is considered to be approximately the same as during the
multipoint calibration and the multipoint calibration does not have to be
repeated. The data reduction technique does not require that no calibration
drift has occurred since the multipoint calibration. Some minor calibration drift
is acceptable because the effect of any drift will be corrected during the
reduction of the assay data.

However, if the relative difference for either the zero gas or the upscale
reference standard check is greater than 5.0 percent, then the analyzer is out
of calibration. Either a new multipoint calibration must be conducted or the
analyzer's zero and span controls must be adjusted to return the analyzer's
response to the original calibration levels. For some analyzers such as

68
 EPA Traceability Protocol for Gaseous Calibration Standards

nondispersive infrared instruments, daily changes in environmental variables

such as barometric pressure may shift the calibration.

After any new multipoint calibration, repeat the measurements of the zero
gas, the candidate zero air material, and the upscale reference standard.
Record these measurements in the laboratory's records.

After any adjustment of controls, repeat the measurements of the zero gas,
the candidate zero air material, and the upscale reference standard. Record
these measurements in the laboratory's records. Recalculate the relative
differences to verify that the analyzer has returned to being in calibration.
The analyzer will be considered to be still out of calibration if the relative
differences remain greater than 5.0 percent. In such case, the analyzer may
need to be repaired before the assay is repeated.

9. The spreadsheet described in Appendix E or an EPA-approved equivalent

statistical technique must be used to determine whether the candidate zero
air material meets the Part 72.2 impurity specifications.

The spreadsheet also calculates the percentage of the uncertainty that is due
to the multipoint calibration. This percentage is needed for the total
uncertainty calculations when two or more assays fall under the same
multipoint calibration. Record this value in the laboratory’s records.

The analyst should investigate any of the measurements that appear to be

outliers. Such data may be discarded if the analyst can demonstrate that the
experimental conditions were inappropriate during these measurements.
Data cannot be discarded just because they appear to be outliers, but may be
discarded if they satisfy statistical criteria for testing outliers.39 The analyst
must record any discarded data and a brief summary of the investigation in
the laboratory’s records.

10. If the multipoint calibration data and the assay data underwent any
mathematical transformations before their statistical analysis, perform the
reverse transformations for the estimated concentration and the expanded
uncertainty. Record the transformed values in the laboratory’s records.

11. The specialty gas producer or other laboratory that assayed the candidate
zero air material must maintain laboratory records until such time as the
cylinder is blown down after being returned by the end user to the producer.

2.4.12 Recertification Requirements

Zero air materials contained in compressed gas cylinders do not need to be

recertified, but those that are generated by an ambient air purification system
must be recertified on a yearly basis or after system maintenance.

69
 EPA Traceability Protocol for Gaseous Calibration Standards

Section 3 - EPA Traceability Protocol for Assay and

Certification of Permeation Device Calibration Standards
3.1 General Information
3.1.1 Purpose and Scope of the Protocol

This protocol describes two procedures for assaying the permeation rate
of a candidate permeation device calibration standard and for certifying that the
assayed permeation rate is traceable to National Institute of Standards and
Technology (NIST) reference standards. This protocol is mandatory for certifying
the permeation device calibration standards used for the pollutant monitoring that
is required by the regulations of the Code of Federal Regulations, Chapter 40,
Parts 50 and 582,3 for the calibration and audit of ambient air quality analyzers.
This protocol covers the assay and certification of sulfur dioxide (SO2) and
nitrogen dioxide (NO2) permeation device calibration standards. It may be used
by permeation device producers, standard users, or other analytical laboratories.
The assay procedure may involve the comparison of the candidate standards to
compressed gas reference standards (i.e., Procedure P2), or to mass reference
standards (i.e., Procedure P3).

Earlier versions of this protocol included a procedure (i.e., Procedure P1)

for comparison of a candidate standard to permeation device reference
standards. This procedure has been eliminated because NIST no longer certifies
SRM permeation device reference standards.

3.1.2 Reference Standards

The compressed gas reference standards that may be used under this
protocol are NIST Standard Reference Materials (SRMs), Van Swinden
Laboratorium (VSL, the National Metrology Institute of the Netherlands) Primary
Reference Materials (PRMs) that are equivalent to SRMs, NIST-Traceable
Reference Materials (NTRMs), or gas manufacturer's intermediate standards
(GMISs). These standards are described in Subsection 2.1.2 of this report.

NIST's reported uncertainty of SRMs, NTRMs, and PRMs is expressed as

the expanded uncertainty (U), which is the combined uncertainty (uC) multiplied
by a coverage factor (k) almost always equal to 2.9 It includes allowances for the
uncertainties of known sources of systematic error as well as the random error of
measurement. A value of one-half of U for these reference standards should be
used in calculating the total analytical uncertainty of standards that are certified
under this protocol using the Appendix C spreadsheet (or an EPA-approved
equivalent statistical technique, see Subsection 2.1.6).

Mass reference standards must be traceable to NIST mass standards.41-43

Additionally, they must have an individual tolerance of no more than 0.05

71
 EPA Traceability Protocol for Gaseous Calibration Standards

milligrams (mg). Examples of mass reference standards that meet these

specifications are American National Standards Institute/American Society for
Testing and Materials (ANSI/ASTM) Classes 1, 3, and 4. The mass reference
standards must be recertified on a regular basis (e.g., yearly) at a NIST-
accredited State weights and measures laboratory or at a calibration laboratory
that is accredited by the National Voluntary Laboratory Accreditation Program
(NVLAP)10,11, which is administered by NIST, or by the American Association for
Laboratory Accreditation (A2LA) or by the International Laboratory Accreditation
Conference (ILAC) under ISO/IEC 17025 (General Requirements for the
Competence of Testing and Calibration Laboratories).10,11,12 Alternatively, mass
reference standards may be traceable to mass reference standards of other
national metrology institutes (e.g., KRISS in South Korea, NEL in the United
Kingdom, PTB in Germany). The recalibration frequency is to be determined
from records of previous recalibrations of these standards.

Two separate sets of mass reference standards are recommended.

Working calibration standards should be used for routine permeation device
weighings and should be kept next to the analytical balance in a protective
container. Laboratory primary standards are kept in a protective container,
handled very carefully when in use and kept in a secure location when not in use.
The working standards should be compared to the laboratory primary standards
every 3 or 6 months to check for mass shifts associated with handling or
contamination. The current masses of the working standards as traced to the
laboratory primary standards should be recorded in the laboratory's records and
should be used to check the calibration of the analytical balance.

Always use smooth, nonmetallic forceps for handling mass reference

standards. The standards are handled only with these forceps, which are not
used for any other purpose. Mark these forceps to distinguish them from the
forceps that are used to handle permeation devices. Handle the standards
carefully to avoid damage that may alter their masses.

The temperature sensor must have an uncertainty of no more than 0.05

degrees Celsius (°C). It must have a manufacturer's calibration certificate
showing traceability to NIST standards. It must be recertified on a regular
schedule (e.g., yearly) according to NIST guidelines44,45 by the user, a NIST-
accredited State weights and measures laboratory or at a calibration laboratory
that is accredited by NVLAP, A2LA or by ILAC under ISO/IEC 17025.
Alternatively, temperature sensor calibrations may be traceable to temperature
standards of other national metrology institutes.

3.1.3 Selecting a Procedure

Procedure P2 is applicable to the assay and certification of candidate

permeation device calibration standards using an ambient air quality analyzer. It
provides for the assay to be referenced to a compressed gas reference standard.

72
 EPA Traceability Protocol for Gaseous Calibration Standards

Procedure P3 is applicable to the assay and certification of candidate

standards using an analytical balance. It provides for the assay to be referenced
to a mass reference standard.

3.1.4 Using the Protocol

The assay/certification protocol described here is designed to minimize

both systematic and random errors in the assay process. Therefore, the protocol
should be carried out exactly as it is described. The assay procedures in this
protocol include one possible design for the assay apparatus. The analyst is not
required to use this design and may use alternative components and
configurations that produce equivalent-quality measurements. The protocol is
like a basic recipe that cooks will follow while substituting different ingredients.

The accuracy of the concentration of gas mixtures that are produced by

permeation devices may decrease due to physical or chemical sorption of the
permeated gas in the permeation system. This sorption will have a larger effect
on the inaccuracy as the concentration decreases. Nonreactive materials (e.g.,
Teflon®, stainless steel, borosilicate glass or silanized glass) and clean,
noncontaminating components should be used in those portions of the
permeation system that are in contact with the gas mixtures being assayed.

3.1.5 Certification Documentation

Each certified permeation device calibration standard must be

documented in a written certification report containing the following information:

1. Permeation device identification number;

2. The contents of the permeation device;

3. Certified permeation rate [in nanograms (ng) per minute];

4. Overall uncertainty estimate associated with the candidate standard. This

estimate must include the uncertainty associated with the assay and the
uncertainty associated with the reference standard.

5. The certification temperature (in °C to the nearest 0.1°C);

6. The dilution gas (air or nitrogen) used in the assay (for Procedure P2);

7. Date of the assay/certification;

8. Identification of the reference standards used in the assay: producer,

cylinder identification number, certified concentration, estimated

73
 EPA Traceability Protocol for Gaseous Calibration Standards

uncertainty, and identification of the reference standard as being an SRM,

an NTRM, an RGM, a PRM, a CRM or a GMIS; manufacturer, model
number and serial number for a mass or temperature reference standard.
The certification documentation must identify the type of reference
standard (i.e., gaseous or mass) used in the assay;

9. Statement that the assay/certification was performed according to this

protocol and that lists the assay procedure (e.g., Procedure P2) used;

10. The analytical method that was used in the assay;

11. Identification of the laboratory where the standard was assayed;

12. Chronological record of all certifications for the standard by the laboratory;

13. Statement that the standard will retain its certification only as long as 5
percent of the original liquid weight or a visible amount of liquid remains in
it; and

14. Any environmental exposure conditions (e.g., temperature and moisture)

that will invalidate the certification.

This certification documentation must be given to the purchaser of the standard.

The permeation device producer must maintain laboratory records and
certification documentation for 3 years after the standard's certification date. A
permeation device producer or other vendor may redocument an assayed and
certified standard that it has purchased from another permeation device producer
and that it wishes to sell to a third party. However, the new certification
documentation must clearly list the permeation device producer or other
laboratory where the standard was assayed.

3.1.6 Certification Label

The standard must be labeled with its identification number.

3.1.7 Assay/Certification of Candidate Permeation Device Calibration Standards

3.1.7.1 Permeation Device Design—

Permeation devices are designed and constructed in various ways, but all
devices consist of a sealed chamber containing liquified gas and a permeable
area through which the gas is allowed to permeate. The permeated gas is swept
and diluted with a measured volumetric flow rate of dry air or nitrogen to create a
quantitative concentration of the pollutant gas.

74
 EPA Traceability Protocol for Gaseous Calibration Standards

3.1.7.2 Precautions for Use and Storage of Permeation Devices—

The permeation rate of all permeation devices is critically dependent on
temperature; a permeation device is useful as a concentration standard only
when its temperature is precisely controlled and accurately measured and when
an accurately metered dilution gas flow rate is provided.

The reproducibility of the certified permeation rate of a permeation device

may be adversely affected by exposure of the device to temperatures greater
than the specified operating or storage temperature range for the device or by
exposure to excessive moisture. NO2 permeation devices must be stored under
dry conditions and preferably at a temperature between 20 and 35 °C, or as
otherwise recommended by the manufacturer. SO2 permeation devices may be
refrigerated for storage.

A series of NIST SO2 permeation tubes SRMs were studied at NIST for
their long-term stability at temperatures of 20, 25 and 30° C 46.47. Contrary to
previous reports, their steady-state permeation rates were not constant, but
showed a very slight downward trend with time. The amount of decay in the
permeation rate is less than 2% of its initial rate over the life of the tube. In
normal day-to-day use, permeation rate changes occurring over the entire
steady-state life of the tube would not be discernible. The NIST S02 permeation
tube SRMs had been certified to an uncertainty of ± 2% which is well beyond the
variation in permeation rate usually observed during short-term use.

There is a limited temperature range at which NO2 permeation devices can

be used as standards. This temperature range is conservatively given as 20 to
35 °C.48, 49 Low or high temperature storage of NO2 permeation devices is not
recommended.

NO2 permeation devices must be stored in and used in dry dilution gas.
One study showed that NO2 permeation rates were significantly lower in
moderately humid air (i.e., 30 to 40 percent relative humidity) than in dry air on
the preceding day.47 Furthermore, the permeation rates did not return to the
original levels after dry air had passed over the device for 24 hours. Another
study found that NO2 concentrations from a permeation device declined by about
one-third as relative humidity levels increased from 0 to 100 percent.50

Candidate standards being certified under Procedure P3 must be stored

under constant temperature conditions between assays. A storage container for
this application is described in the procedure.

When stored at a temperature other than the assay temperature, some

permeation devices require an equilibration period at the assay temperature to
reach thermal equilibrium and a stable, accurate permeation rate. When
transferred from a different storage temperature, thin-walled permeation devices
should be maintained at the assay temperature with a fixed dilution flow rate for

75
 EPA Traceability Protocol for Gaseous Calibration Standards

at least 48 hours before use or before certification. Temperature changes of >10

°C may require equilibration periods of up to 15 days for NO2 permeation devices
to attain a stable permeation rate.48,51 Upon return to the original temperature,
some devices may not return to the same permeation rate as before the
temperature change. Other types of permeation devices may require longer
equilibration periods. Observe any manufacturer's recommendations for
equilibration and use.

3.1.7.3 Equilibration of Newly Prepared Permeation Devices—

A newly prepared permeation device must be equilibrated for at least 48
hours at the assay temperature before being assayed for the first time. NIST
found that permeation tubes made of TFE and FEP Teflon® containing inorganic
gases must be conditioned at a temperature of above 40°C for a minimum of 24
hours to assure stable and reproducible operation47. Once conditioned, the
equilibration time from a storage temperature at -15°C to an operation
temperature of 30°C for sulfur dioxide and other inorganic gas permeation tubes
range from 24 hours to 48 hours. For smaller step changes in temperature, the
permeation rate reaches 95 % of the new value in approximately 30 minutes.
The equilibration period may be 100 hours or longer for some permeation
devices.51 This period will vary as a function of the permeating compound, the
material and the thickness of the permeating surface and the temperature.

3.1.7.4 Certification Conditions for Permeation Devices—

A standard will retain its certification only as long as 5 percent of the
original liquid weight or a visible amount of liquid remains in it. It loses its
certification if it is exposed for prolonged periods of time to excessive moisture or
to temperatures greater than 15 °C above its certification temperature. A
decertified standard must be reassayed before it can be certified for further use.

3.1.8 Technical Variances

Permeation device producers, standard users, and other analytical

laboratories may petition the U.S. EPA for technical variances to the assay
procedures in this protocol. A technical variance allows the use of a specific
alternative assay procedure for candidate standards, which can be certified
under this protocol. The petitioner must send a written request with a detailed
description of the alternative assay procedure and supporting analytical data to
EPA Traceability Protocol Project, Technical Services Branch, U.S. EPA, Mail
Code E343-03, Research Triangle Park, NC 27711. The supporting analytical
data must demonstrate the equivalence of the alternative assay procedure with
the procedures given in this protocol. Schuirmann's TOST13-18 is the
recommended statistical technique to demonstrate such equivalence if
technically possible. Technical variances may also be given for alternative
temperature ranges of certifying or storing permeation devices provided that
supporting analytical data are provided with the written request.

76
 EPA Traceability Protocol for Gaseous Calibration Standards

Permeation device producers, standard users, and other analytical

laboratories may petition the U.S. EPA to allow the assay and certification of
permeation devices that contain gases or liquified gases other than SO2 and
NO2. The petitioner must send a written request with a detailed description of the
permeation device and supporting analytical data to the EPA Traceability
Protocol Project at the address given above. The supporting analytical data must
demonstrate that the permeation rate for the proposed device can be accurately
determined, that only the specified compound is permeating, that the rate is
stable over the lifetime of the device, and that the rate is not changed by
temperature and humidity effects.

77
 EPA Traceability Protocol for Gaseous Calibration Standards

3.2 Procedure P2: Assay and Certification of Permeation

Device Calibration Standards based on a Compressed Gas
Reference Standard
3.2.1 Applicability

This procedure may be used to assay the permeation rate of a candidate

SO2 and NO2 permeation device calibration standard, based on the concentration
of a compressed gas reference standard of the same pollutant compound, and to
certify that the assayed permeation rate is traceable to the reference standard.
The procedure employs a low-concentration range (i.e., ambient air quality level)
pollutant gas analyzer to compare quantitatively diluted concentrations from the
permeation device calibration standard with quantitatively diluted concentrations
from the compressed gas reference standard. This procedure may be used for
the assay of multiple candidate standards during the same assay session.
Criteria that apply to the assay of one candidate standard apply to the assay of
multiple candidate standards. This procedure may be used by permeation
device producers, standard users or other analytical laboratories.

3.2.2 Limitations

1. The concentration of the diluted candidate standard may be greater than

or less than the concentration of the diluted reference standard. However,
the diluted concentrations from both standards must lie within the well-
characterized region of the analyzer's calibration curve (see Subsection
2.1.4.2). Additionally, U for the regression-predicted concentration of the
diluted candidate standard must be ≤1.0 percent of the concentration of
the diluted reference standard. This uncertainty estimate is obtained from
the statistical analysis of the multipoint calibration data using the
spreadsheet described in Appendix A or using an EPA-approved
equivalent statistical technique (see Subsection 2.1.6). This criterion
means that the uncertainty associated with the multipoint calibration
determines the concentration range over which a diluted candidate
standard may be assayed.

2. An accurate dilution and flow measurement system is required.

3. A source of clean, dry zero gas is required (see Subsection 2.1.3.5). It

may come from compressed gas cylinders or from zero gas generators.

4. This procedure is designed to assay the permeation rate of a candidate

standard that is mounted in a specially designed assay dilution system.
The procedure does not accommodate the certification of a candidate
standard that is mounted in its own self-contained dilution/flow
measurement system.

79
 EPA Traceability Protocol for Gaseous Calibration Standards

3.2.3 Assay Apparatus

Figure 3-1 illustrates the components and configuration of one possible

design for the assay apparatus, including a common dilution system for both the
reference and candidate standards. The configuration is designed to allow
convenient routing of zero gas and diluted concentrations of the reference
standard and the candidate standard, in turn, to the analyzer for measurement,
as selected by valves V1, V2, and V3. Three gas flow controllers (i.e., C1, C2,
and C3) regulate the total dilution flow rate for the candidate standard, the purge
gas flow rate, and the reference standard flow rate. These gas flow controllers
may be needle valves, capillary tubes, thermal mass flow controllers, or other
suitable devices. The flow rates should be controlled to within 1.0 percent
variation during the assay. The dilution flow rates for the reference and
candidate standards is measured by a single, common flowmeter (i.e., M1). The
reference standard and purge gas flow rates may be measured at the vent port of
valve V2 or by flowmeters M2 and M3 that are mounted in the two gas streams.

When the candidate standard is being measured, valve V1 directs a

portion of the total dilution flow through the candidate standard chamber. This
sweep flow rate is regulated by gas flow controller C1 and is measured by gas
flowmeter M2. This flowmeter need not be accurately calibrated because only
the total dilution flow rate, measured by flowmeter M1, is used in the dilution
calculations. When the reference standard is being measured, valve V1 directs
the purge gas through the candidate standard chamber. The purge gas prevents
the buildup of high pollutant concentrations in the chamber. It is vented through
valve V2 and is not measured by flowmeter M1.

If it is necessary to use different dilution flow rates for the reference

standard and the candidate standard (see Subsection 3.2.6), separate flow
controllers for the two dilution flow rates may be used. However, the same
flowmeter should be used to measure both dilution flow rates to help reduce
systematic flow measurement errors.

The mixing chamber combines the gas streams and should provide
turbulence to ensure thorough mixing of the two gas streams. The diluted gas
mixtures are routed to the analyzer through a union tee tube fitting, which vents
excess gas flow. Normally, the excess gas is vented to the atmosphere without
any obstructions in the tubing and the gas entering the analyzer is at near-
atmospheric pressure. However, the excess gas can be routed through an
uncalibrated rotameter by rotation of a three-way valve (i.e., V4). The rotameter
is used to demonstrate that the total gas flow rate exceeds the sample flow rate
of the analyzer and that no room air is being drawn in through the vent line.
Check the apparatus carefully for leaks and correct all leaks before use.

80
 EPA Traceability Protocol for Gaseous Calibration Standards

Figure 3-1. One possible design of the apparatus for the assay of permeation
device calibration standards referenced to a compressed gas reference standard

The assay apparatus illustrated in Figure 3-1 may be modified by the

addition of multiple chambers, which may be set to different temperatures.

The mean temperature of the candidate standard chamber must be

controlled to within 0.05 °C of the setpoint with a temperature stability of ±0.05
°C. It must be measured with a NIST-traceable temperature sensor having a
measurement uncertainty of not more than 0.05 °C (see Subsection 3.1.2).

3.2.4 Pollutant Gas Analyzer

See Subsection 2.2.4. The pollutant gas analyzer must have a well-
characterized calibration curve and a range capable of measuring the diluted
concentrations of both the candidate and reference standards. It must have
good resolution, good precision, a stable response, and low output signal noise.
It must have good specificity for the pollutant of interest so that it has no
detectable response to any contaminant that may be contained in the standards.
A suitable analyzer with acceptable performance specifications may be selected
from the list of EPA-designated reference and equivalent method analyzers37. If

81
 EPA Traceability Protocol for Gaseous Calibration Standards

the balance gas of the reference standard must be different from the zero gas
used for dilution (e.g., air versus nitrogen or different proportions of oxygen),
either a high dilution ratio (i.e., at least 50 parts zero gas to 1 part standard)
should be used, or the analyzer must be proven to be not sensitive to differences
in the balance gas composition. The latter may be demonstrated by showing no
difference in an analyzer's response when measuring a calibration standard that
has been diluted with identical flow rates of the different balance gases.

The analyzer should be connected to a high-precision data acquisition

system (DAS), which must produce numeric values and a graphic record of the
analyzer response during the assay. More precise values will be obtained if the
DAS has some data averaging capability. The numeric values and the graphic
record must be maintained for 3 years after the standard’s certification date.

If the analyzer has not been in continuous operation, turn it on and allow it
to stabilize (e.g., for at least 12 hours) before beginning any measurements.

3.2.5 Analyzer Calibration

3.2.5.1 Multipoint Calibration—

See Subsections 2.1.4.2 and 2.1.4.3 for information about the multipoint
calibration. If an undiluted check standard is used, its concentration must fall in
the well-characterized region of the calibration curve. If a diluted check standard
is used, the diluted concentration must fall in the well-characterized region.
Following completion of the multipoint calibration, the accuracy of the assay
apparatus must be checked to verify that the error associated with the dilution is
not excessive. This accuracy check involves the measurement of an undiluted or
diluted check standard. The check standard must be a NIST SRM, a SRM-
equivalent PRM, an NTRM, or a GMIS as specified in Subsection 2.1.3. It must
have a certified concentration that is different from that of the reference standard
used in the multipoint calibration. Information concerning this standard (e.g.,
cylinder identification number, certified concentration) must be recorded in the
laboratory's records.

Make three or more discrete measurements of the undiluted or diluted

check standard. "Discrete" means that the analyst must change the gas mixture
being sampled by the analyzer between measurements. For example, the
analyst might alternate between measurements of the check standard and the
zero gas. Record these measurements in the laboratory's records.

Next the analyst must verify that the dilution error is not excessive. For
the check standard measurements, calculate the relative difference (in percent)
between the mean analyzer response and the corresponding response that is
predicted from the multipoint calibration regression equation and the undiluted or
diluted check standard concentration. That is,

82
 EPA Traceability Protocol for Gaseous Calibration Standards

Relative Difference = 100 [ Mean Analyzer Response - Predicted Response]

[Predicted Response]

If the relative difference is >1.0 percent, the dilution error is considered to be

excessive. The analyst must investigate why it is excessive (e.g., errors in the
reference standard and check standard concentrations, errors in assay
apparatus or to some other source). Assays may not be conducted until the
relative difference for a subsequent accuracy check is ≤ 1.0 percent.

3.2.5.2 Analyzer Range—

See Subsection 2.2.6.2 for information about analyzer range.

3.2.5.3 Linearity—
See Subsection 2.2.6.3 for information about linearity.

3.2.5.4 Zero and Span Gas Checks—

See Subsection 2.2.6.4 for information about zero and span gas checks.

3.2.6 Selection of Gas Dilution Flow Rates

The dilution flow rates used for the reference standard and the candidate
standard should be selected carefully to provide diluted concentrations for both
standards that fall in the well-characterized region of the analyzer's calibration
curve. Additionally, the magnitude of U for the estimated concentration of the
diluted candidate standard must be ≤ ±1.0 percent of the concentration of the
diluted reference standard. It is desirable to measure both dilution flow rates with
the same flowmeter (i.e., M1). This practice reduces measurement errors
associated with the use of multiple flowmeters.

3.2.7 Flowmeter Type and Flowmeter Calibration

Flowmeters M1 and M3, shown in Figure 3-1, measure in-line flow rates
and do not operate at atmospheric pressure because of backpressure from
downstream components. Also, this backpressure is variable, depending on the
total dilution and reference standard flow rates. Thus, the flowmeters must
compensate for the variable in-line pressure. Thermal mass flowmeters do not
need to be corrected for pressure effects. Measurements from pressure-
sensitive flowmeters such as rotameters or wet test meters must be carefully
corrected for the actual in-line pressure during the flow rate measurements.

Alternatively, the flow rates can be measured at the outlet of the dilution
apparatus, with the excess gas flow vent temporarily plugged. In this case, a
volumetric meter such as a piston-type positive-displacement flowmeter, a wet
test meter or a soap-film flowmeter can be used, and flow measurements may be
conveniently referenced to atmospheric pressure. Each flow rate must be
measured independently while the other flow rate is set to zero. Great care must

83
 EPA Traceability Protocol for Gaseous Calibration Standards

be exercised to ensure that each measured flow rate remains constant when
combined with the other flow rate and between the time of measurement and the
time of the assay.

The flowmeters used should be stable, repeatable, linear, and have good
resolution. The flowmeters must not contaminate or react with the gas mixture
passing through them. If possible, select flow rates or flowmeter ranges such
that the measured flow rates fall in the upper half of the flowmeters' ranges. The
flowmeters should be carefully calibrated at several flow rates to prove linearity.
The calibration of the zero gas flowmeter should be accurate to ≤ 1.0 percent,
referenced to an accurate flow or volume standard traceable to a NIST primary
standard (see Subsection 3.1.2). This flowmeter calibration should be checked
and recertified on a regular schedule (e.g., yearly). The recertification frequency
is to be determined from stability information such as a chronological control
chart of calibration data.

All volumetric flow-rate measurements must be corrected or referenced to

the same temperature and pressure conditions, such as EPA-standard conditions
(i.e., 760 millimeters of mercury (mm Hg) and 25 °C) or the ambient temperature
and pressure conditions prevailing in the laboratory during the assay.
Measurements using wet test meters and soap-film flowmeters also must be
corrected for the saturation of the gas stream with water vapor in the moist
interiors of these flowmeters. The equation to correct the flow rate for
temperature, pressure, and humidity effects is given below:

Flow Rate = (Volume/Time)(PM – PWV/PS)(TS/TM)

where

PM = measured barometric pressure (mm Hg);

PWV = partial pressure of water vapor (mm Hg);
PS = standard pressure (mm Hg);
TS = standard temperature (298.2 K); and
TM = measured ambient temperature (273.2 + EC).

3.2.8 Candidate Standard

See Subsections 3.1.7.1 and 3.2.2. Follow the manufacturer's instructions

for equilibration and for use of the candidate standard and for selecting the
temperature at which it is to be assayed and certified. The candidate standard
should be assayed at the same temperature at which it will be subsequently
used. The mean operating temperature of the candidate standard chamber must
be controlled to within 0.05 °C of the setpoint with a temperature stability of ±0.05
°C. This temperature must be measured with a NIST-traceable temperature
sensor with a measurement uncertainty ±0.05 °C or less (see Subsection 3.1.2).

84
 EPA Traceability Protocol for Gaseous Calibration Standards

3.2.9 Reference Standard

See Subsections 2.1.3 and 3.1.2 for information about the reference
standard.

3.2.10 Zero Gas

See Subsection 2.1.3.6 for information about the zero gas. If possible, the
zero gas should be the same as the balance gas of the reference standard.

3.2.11 Assay Procedure

1. Verify that the assay apparatus is properly configured as shown in Figure

3-1 and described in Subsection 3.2.3. Inspect the analyzer to verify that
it is operating normally and that all controls are set to their appropriate
values. Record these control values in the laboratory's records.

2. Determine and establish the operating temperature for the candidate

standard in its temperature-controlled chamber. Install the candidate
standard, start the purge gas flow, and allow ample time for the device to
equilibrate (see Subsection 3.1.7.3). Record the temperature in the
laboratory's records.

3. Verify that the flowmeters are properly calibrated (see Subsection 3.2.7).

4. Verify that a multipoint calibration of the analyzer has been performed

within the past month demonstrating acceptable instrument performance
(see Subsection 2.1.4.2). Additionally, verify that the zero and span gas
checks indicate that the analyzer is in calibration (see Subsection 2.1.4.4).

5. Determine and establish the reference standard flow rate and the dilution
flow rates and diluted concentrations for the reference standard and the
candidate standard that will be used for the assay (see Subsection 3.2.7).
Ensure that the diluted reference standard and diluted candidate
standards concentrations are within the well-characterized region of the
analyzer's calibration curve (see Subsection 2.1.4.2). Also check that both
dilution flow rates will provide enough flow for the analyzer, with sufficient
excess to ensure that no ambient air will be drawn into the vent line. Also
adjust the flow rate of the portion of the dilution flow that passes over the
candidate standard (i.e., flow controller C3), and adjust the purge flow rate
(i.e., flow controller C2) to approximately the same value.

Calculate the diluted concentration using the following equation:

Diluted Conc. = (Undiluted Standard Conc.)(Standard Flow Rate)

(Standard Flow Rate + Zero Gas Flow Rate)

85
 EPA Traceability Protocol for Gaseous Calibration Standards

Calculate the diluted candidate standard concentration (in ppm) using the
following equation:

Diluted Standard Conc. = (10-3)(MV/MW)(Perm Rate/Dilution Flow Rate)

where

MV = molar volume of the dilution gas (liters/mole);

= (0.08206) TM
MW = molecular weight of the dilution gas (grams/mole);
the permeation rate is given in nanograms/minute; and
the dilution flow rate is given in liters/minute.

Use an estimated permeation rate for the candidate standard in this

calculation.

Record the measured flow rates and the undiluted and diluted reference
standard concentrations in the laboratory's records.

6. In succession, measure the zero gas, the diluted reference standard and
the diluted candidate standard using the analyzer. Use valves V1, V2,
and V3 to select each of the three gas mixtures for measurement. For
each measurement, adjust the flow rates, if necessary, to those
determined in Step 5, and allow ample time for the analyzer to achieve a
stable reading. If the reading for each measurement is not stable, the
precision of the measurements will decline and the candidate standard
might not be certifiable under this protocol. Record the analyzer response
for each measurement, using the same response units (e.g., volts, area
counts, etc.) as was used for the multipoint calibration. At this point, do
not convert the data into concentration values using the calibration
equation. Do not perform any mathematical transformations of the data.
These steps will be done later. Do not make any zero control, span
control, or other adjustments to the analyzer during this set of
measurements. Record the responses in the laboratory's records.

The analyst may assay multiple candidate standards during the same
assay session. For example, a single set of measurements may involve a
zero gas, a diluted reference standard, and three diluted candidate
standards. Criteria that apply to the assay of one candidate standard
apply to the assay of multiple candidate standards. The analyst should be
aware that the effect of any short-term calibration drift will be greater when
multiple candidate standards are assayed. This greater effect is due to
the longer period of time between reference standard measurements.
Unacceptable uncertainties of the estimated concentrations for the diluted
candidate standards may occur as a result of the longer assay session.

86
 EPA Traceability Protocol for Gaseous Calibration Standards

7. Conduct at least two additional sets of measurements, as described in

step 6 above. However, for these subsequent sets of measurements,
change the order of the three measurements (e.g., measure the reference
standard, zero gas, and candidate standard for the second set and
measure the zero gas, candidate standard, and reference standard for the
third set, etc.). Changing the order that the gas mixtures are measured
helps the analyst to discover any effect that one measurement has on
subsequent measurements. The number of sets of measurements will
have been determined during the analysis of the multipoint calibration data
such that U of the regression-predicted concentration of the candidate
standard ≤1 percent of the concentration of the reference standard.

8. If any one or more of the measurements of a set of measurements is

invalid or abnormal for any reason, discard all three measurements and
repeat the measurements. Such measurements may be discarded if the
analyst can demonstrate that the experimental conditions were
inappropriate during these measurements. Data cannot be discarded just
because they appear to be outliers but may be discarded if they satisfy
statistical criteria for outliers.39 In the case of outliers, all of the
measurements do not have to be discarded. Only the outliers should be
discarded. As part of the laboratory records, the analyst must record any
discarded data and briefly explain in the laboratory's records why these
data were discarded.

9. The spreadsheet described in Appendix A or an EPA-approved equivalent

statistical technique (see Subsection 2.1.6) must be used to calculate the
estimated concentration and U for the diluted candidate standard based
on data from the assay measurements and from the multipoint calibration.
The use of both sets of data in the statistical analysis produces an
estimated concentration with smaller uncertainty while correcting for any
minor calibration drift that may have occurred since the multipoint
calibration. Record the estimated concentration and U in the laboratory’s
records.

The spreadsheet also calculated the percentage of the uncertainty that is

due to the multipoint calibration. This percentage is needed for the total
uncertainty calculations when two or more assays fall under the same
multipoint calibration. Record this value in the laboratory’s records.

10. If the multipoint calibration data and the assay data underwent any
mathematical transformations before their statistical analysis, the analyst
must perform to reverse transformations for the estimated concentration
and U. Record the transformed values in the laboratory’s records.

11. Finally, calculate the certified permeation rate (in nanograms/minute) and
uncertainty for the candidate standard using the equations below:

87
 EPA Traceability Protocol for Gaseous Calibration Standards

Perm Rate = (103)(MW/MV)(Estimated Conc.)(Dilution Flow Rate)

Uncertainty = (103)(MW/MV)(U)(Dilution Flow Rate)

3.2.12 Equilibration Test for Newly Prepared Permeation Devices

A permeation device that has not been previously assayed must be tested
for a stable permeation rate as follows: Reassay the permeation rate at least 24
hours after the first assay and compare the two assayed concentrations. The
Appendix C spreadsheet or an EPA-approved equivalent statistical technique
(see Subsection 2.1.6) must be used to evaluate the stability of the permeation
rate by comparison of the two assays. If the mean estimated concentrations
from the assays differ by less than 1.0 percent from the overall mean estimated
concentration using Schuirmann's two one-sided test (TOST)13-18, the permeation
rate can be considered to be stable and the candidate standard may be certified
for use. Otherwise, equilibrate the device longer at the operating temperature
and repeat the test, using the second and third assays as if they were the first
and second. This process may be repeated until the rate stabilizes. Permeation
devices that are not stable may not be used for calibration or audit purposes.
Candidate standards that fail the initial and the repeat stability tests are unstable
and are disqualified for further use under this protocol.

3.2.13 Certification Documentation

See Subsections 3.1.5 and 3.1.6.

88
 EPA Traceability Protocol for Gaseous Calibration Standards

3.3 Procedure P3: Assay and Certification of Permeation Device

Calibration Standards based on a Mass Reference Standard
3.3.1 Applicability

This procedure may be used to assay the permeation rate of a candidate

SO2 or NO2 permeation device calibration standard based on mass reference
standards, and to certify that the assayed permeation rate is traceable to the
reference standard. The procedure employs an analytical balance to measure
the weight loss in the candidate standard. It may be used for the assay of
multiple candidate standards during the same assay session. Criteria that apply
to the assay of one candidate standard apply to the assay of multiple candidate
standards. This procedure may be used by permeation device producers,
standard users, or other analytical laboratories.

3.3.2 Limitations

This procedure is intended only for the assay of candidate standards

containing SO2 or NO2. These liquid compounds must be anhydrous grade (i.e.,
minimum purity 99.99 percent) or phosphorous pentoxide-dried commercial-
purity grade (i.e., minimum purity 99.5 percent).

An accurate analytical balance with a NIST-traceable calibration is

required to weigh the candidate standard.

A temperature-controlled chamber for maintaining the candidate standard

at a constant, NIST-traceable temperature between weighings is required.

A source of clean, dry zero gas is required (see Subsection 2.1.3.5).

3.3.3 Assay Apparatus

3.3.3.1 Analytical Balance—

Choose a balance with adequate vibration-stabilization control and
appropriate specifications for total weighing capacity, accuracy, precision, and
readability. The balance should be chosen such that the manufacturer's
specified uncertainty (i.e., three times the standard deviation or "reproducibility")
of the balance divided by the weight of the candidate standard does not exceed
0.001. The balance must be calibrated annually using NIST-traceable mass
reference standards by the manufacturer or a manufacturer's representative (see
Subsection 3.1.2).

If possible, locate the balance in a climate-controlled, draft-free room,

preferably dedicated to the use of balances. If this is not possible, the general
guidelines listed below are to be followed to control environmental factors that
may affect balance performance:

89
 EPA Traceability Protocol for Gaseous Calibration Standards

- Locate the balance away from potential sources of drafts such as doors,
windows, aisles with frequent traffic, ventilation ducts, and equipment with
fans or moving parts.

- Locate the balance out of direct sunlight and away from local heating or
cooling sources such as open flames, hot plates, water baths, ventilation
ducts, windows, and heat-producing lamps.

- Locate the balance on a sturdy base (ideally, a stone weighing table) and
away from any equipment that produces vibrations. If this is not possible,
isolate the balance from such equipment by placing a stabilizing slab
under the balance or composite damping-pads under the balance legs.

- Ensure that the balance-support is sufficiently level to permit leveling of

the balance according to the manufacturer's instructions.

3.3.3.2 Temperature-Controlled Chamber—

A temperature-controlled chamber is required for storing the candidate
standard between weighings. One possible design for the chamber is depicted in
Figure 3-248. Clean, dry zero gas enters the chamber at the bottom after passing
through the heat exchanger tubing (i.e., several turns of copper tubing). The
zero gas' flow rate must be sufficient to purge the chamber thoroughly. The
chamber and the heat exchanger are immersed in a thermostatted bath to the
level shown in the figure. The bath must control the mean temperature of the
chamber to within 0.05 °C of the setpoint with a temperature stability of ±0.05 °C.
The temperature of the bath or the chamber must be measured and recorded in
the laboratory's records on at least a daily basis. A NIST-traceable temperature
sensor must be used for this measurement (see Subsection 3.1.2). It must be
calibrated annually using NIST-traceable temperature reference standards and
must have an uncertainty similar to that of these reference standards. The
output of a temperature-sensing device may be recorded by a DAS.

3.3.3.3 Electrostatic Charge Neutralization—

Electrostatic charge buildup will prevent an analytical balance from
operating properly. Static charge is the accumulation of electrical charges on the
surface of a nonconductive material, which could be the permeation device or
part of the analytical balance. Common symptoms of this problem include noisy
readout, drift, and sudden readout shifts.

90
EPA Traceability Protocol for Gaseous Calibration Standards

Figure 3-2. Chamber for storing permeation tubes48.

91
 EPA Traceability Protocol for Gaseous Calibration Standards

To reduce static charge within the balance, it may be necessary to place a

radioactive antistatic strip containing a very small amount (i.e., 500 picocuries) of
Polonium-210 (Po210) in the weighing chamber. It may also be necessary to put
each permeation device on a strip before it is weighed. Po210 antistatic strips are
used to reduce electrostatic charge buildup in the analytical balance's weighing
chamber and on individual permeation devices by charge neutralization. They will
neutralize electrostatic charges on items placed within an inch of them. These
strips are safe, commonly available, and very inexpensive. Po210 has a half-life of
138 days. Change the strips semiannually and dispose of the old strips
according to the manufacturer's recommendations.

Antistatic solutions are available for coating (and recoating at appropriate

and relatively infrequent intervals) the interior and exterior nonmetallic surfaces
of the chamber. This coating facilitates the draining of electrostatic charges from
these surfaces to a common electrical ground to which the metallic conductive
surfaces are connected. Earth-grounded conductive mats placed on the weighing
table surface and under the analyst's shoes are used to reduce electrostatic
charge buildup. Do not assume that the electrical grounding of the analytical
balance eliminates all buildup because the ground may not be perfect.

Even though a permeation device's weight might stabilize within 60

seconds and no weight drift is observed during that period, the balance may still
be influenced by electrostatic charge buildup. It may still be necessary to repeat
the neutralization procedure and to use antistatic strips inside the weighing
chamber. One may reduce the effect of buildup on permeation devices by
putting them in an aluminum foil boat on the balance pan during weighings.

Charge neutralization times may need to be longer than 60 seconds.

Electrostatic charge buildup becomes greater as the air becomes drier. A 60-
second neutralization may work sufficiently in ambient indoor air conditioned to
37 percent relative humidity and 23 °C, but not in zero nitrogen. This latter
environment may require that the permeation device sit for more time on the
antistatic strip. The neutralization may have to be done inside the weighing
chamber or a second small chamber, which is used just for charge neutralization.

3.3.4 Weighing Interval

The minimum time period between weighings of the candidate standard is

a function of the expected permeation rate, the specified uncertainty for the rate,
and the analytical balance's readability (i.e., the smallest scale division). The
following equation is based on a ±1 percent uncertainty specification for the
permeation rate:

100(readability)
Minimum weighing interval =
(expected permeation rate)

92
 EPA Traceability Protocol for Gaseous Calibration Standards

where the weighing interval is in minutes; the readability is in grams; and the
expected permeation rate is in grams per minute.

3.3.5 Assay Procedure

1. Turn on the balance and allow it to warm up for the period specified in the
operator's manual. To ensure maximum stability, it is recommended to
keep the balance turned on at all times. This procedure enables the
balance to be operational at all times and eliminates the need for a
warmup period before analyses. Newer balances are always turned on
(except for their displays) when they are plugged in.

2. Check the balance level and, if necessary, adjust the level according to
the manufacturer's instructions.

3. Ensure that the balance room temperature is between 15 and 30 °C or, if

given, within the balance manufacturer's specifications and that the
balance and mass reference standards are equilibrated to the balance
room temperature. Record the temperature in the laboratory's records.

4. Zero (i.e., tare) and calibrate the balance according to the manufacturer’s
directions. Record the tare reading in the laboratory’s records. Many
newer balances calibrate themselves automatically or only require a key to
be touched to calibrate themselves.

5. On each day that the candidate standard is to be assayed, verify the

balance's calibration using an NIST-traceable mass reference standard.
Use smooth, nonmetallic forceps to handle the standards. This standard
must have a mass that is similar to that of the candidate standard. Record
the date, balance identification, standard's identification, certified weight of
the standard, and the measured weight of the standard in the laboratory's
records. Calculate the relative difference (in percent) between the
standard's certified and measured weights as follows:

100 (measured weight - certified weight)

Relative difference =
(certified weight)

Record the relative difference in the laboratory's records. If the relative

difference is >0.1 percent, the balance cannot be used under this protocol
until it has been recalibrated or repaired and until a subsequent
verification has a relative difference of ≤0.1 percent.

6. Review the recorded bath or chamber temperature readings since the

most recent weighing of the candidate standard, or since the standard was
first put into the temperature-controlled chamber. Record the minimum
and maximum temperatures in the laboratory's records. The minimum

93
 EPA Traceability Protocol for Gaseous Calibration Standards

and maximum temperatures must not have deviated from the setpoint by
more than 0.1 °C. If these temperatures deviate by more than this
amount, the current assay and all previous assays are invalidated.

7. Record the current chamber temperature in the laboratory's records.

8. Verify that the candidate standard has been in the temperature-controlled

chamber for a long enough time for its permeation rate to have stabilized.

9. Remove the candidate standard from the temperature-controlled chamber

and place it on the balance's pan using stainless steel forceps or a similar
noncontaminating device. Note that Teflon® permeation tubes may have
an electrostatic charge buildup due to the passage of the dry gas over
them between weighings. Such charges should be removed from the
candidate standard before weighing by Po210 antistatic strips or similar
devices. Note that electronic force balances may require that candidate
standards be thermally equilibrated before they can be weighed.

10. Record the date, time and the candidate standard's identification number
and current weight in the laboratory's records.

11. Return the candidate standard to the temperature-controlled chamber.

The standard should be outside the chamber only for a long enough time
to weigh it using reasonable laboratory technique.

3.3.6 Number of Weighings of the Candidate Standard

The candidate standard must be weighed at least six times after its
permeation rate has stabilized at the certification temperature. After the six or
more weighings, the analyst may assess the stability and uncertainty of the
permeation rate by using the spreadsheet described in Appendix B or an EPA-
approved equivalent statistical technique (see Subsection 2.1.6). The analyst
may calculate a provisional permeation rate from the measured weights and the
time between weighings using the following equation:

(previous weight - current weight)

Provisional permeation rate =
elapsed time between weighings

Based on this data analysis, the analyst may perform additional weighings to
reduce the uncertainty or to replace data that are discarded because they were
obtained before the permeation rate stabilized.

3.3.7 Calculation of Certified Permeation Rate

The certified permeation rate for the candidate standard is the slope of the
least squares regression line for data from at least six weighings of the candidate

94
 EPA Traceability Protocol for Gaseous Calibration Standards

standard after the permeation rate has stabilized. This statistical analysis
technique produces permeation rate estimates that are more precise than those
calculated from weight differences between individual weighings. Although the
minimum number of weighings is six, more precise estimates will be obtained for
more weighings.

Calculate the certified permeation rate and its uncertainty using the
spreadsheet given in Appendix B or using an EPA-approved equivalent statistical
technique (see Subsection 2.1.6). Figure 3-3 presents an example of the
spreadsheet's graphic output.

Permeation Tube Calibration Curve, with 99.999% Confidence

Bands for Estimation (red) and Prediction (blue)
3.425

3.420

3.415

3.410

3.405
Weight (grams)

3.400

3.395

3.390

3.385

3.380

3.375

3.370
314 316 318 320 322 324 326

Julian Date

Figure 3-3. Example of Appendix B spreadsheet graphic output

The permeation rate is the slope (b1) of the least squares regression line
where the x-values are the elapsed times from the initial weighing and the y-
values are the measured weights of the candidate standard. The spreadsheet
also calculates the predicted initial weight (b0) of the permeation device and 95-
percent confidence limits (CLs) for b0 and b1.

After the data from the six or more weighings have been entered in the
spreadsheet, examine the 95-percent CLs for b0. If the measured weight from
the initial weighing falls outside of these limits, the candidate standard may not
have been completely equilibrated at the initial weighing. The analyst may elect
to discard the data from the initial weighing to reduce the uncertainty of the
certified permeation rate.

95
 EPA Traceability Protocol for Gaseous Calibration Standards

Examine the upper and lower 95-percent CLs for b1. They should differ
from b1 by no more than 1 percent of its value. That is,

Upper CL(b1) - b1 ≤ (b1)/1000

b1 - lower CL(b1) ≤ (b1)/1000

If these two criteria are met, the candidate standard can be certified with a
permeation rate equal to b1 and an uncertainty equal to the larger of the two
values. If the criteria are not met, the analyst must make additional weighings of
the candidate standard as described in Subsection 3.3.5. These additional
measurements will be pooled with the previously collected measurements. The
pooled data will be used to obtain new estimates of the b1 and its uncertainty.
When an acceptable value for the uncertainty is obtained, record it and the slope
in the laboratory's records. If an acceptable value is not obtained, the candidate
standard cannot be certified under this protocol.

The analyst should investigate any of the measurements that appear to be

outliers. Such data may be discarded if the analyst can demonstrate that the
experimental conditions were inappropriate during these measurements. Data
cannot be discarded just because they appear to be outliers but may be
discarded if they satisfy statistical criteria for testing outliers39. The analyst must
record any discarded data and a brief summary of the investigation in the
laboratory's records.

3.3.8 Uncertainty of Certified Permeation Rate for Candidate Standard

The total analytical uncertainty of the certified permeation rate includes the
uncertainty of regression slope and the uncertainty of the mass reference
standard that was used to verify the balance’s calibration. The two components
are combined using the following equation for the propagation of errors:

1/2
Uncertainty (Total)  Uncertainty (Slope) 2  Uncertainty (Mass) 2 
=   +  
Permeation Rate  Slope   Mass  

3.3.9 Certification Documentation

See Subsections 3.1.5 and 3.1.6.

96
 EPA Traceability Protocol for Gaseous Calibration Standards

Section 4 - EPA Traceability Protocol for Assay and

Certification of Dynamic Gas Dilution Systems
4.1 Procedure D1: Assay and Certification of a Dynamic Gas
Dilution System
4.1.1 Applicability

This protocol describes a procedure for assaying gas dilution systems and
for certifying that the concentrations in the pollutant gas mixtures that they
generate are traceable to NIST-certified gaseous, flow rate, volume, mass or
time reference standards. In general, a pollutant gas mixture containing a known
concentration is prepared by diluting an NIST-traceable gaseous reference
standard with a zero gas. The flow rates of the reference standard and the zero
gas must be traceable to NIST-traceable flow rate, volume, mass or time
reference standards. Dynamic gas dilution systems employ various dilution
methods to prepare pollutant gas mixtures. Examples of these methods include
capillary tubes, critical orifices, volumetric piston pumps, and thermal mass-flow
controllers. This protocol may be used by the manufacturers of gas dilution
systems, other laboratories, and end users.

This protocol and its associated spreadsheet (see Appendix F) allow an

analyst to estimate the uncertainty of concentrations of pollutant gas mixtures
that are generated by gas dilution systems being certified. This estimate
includes the uncertainty of the gas dilution and the uncertainty of the NIST-
traceable gaseous reference standard being diluted. The uncertainty of the gas
dilution is determined from annual multipoint flow rate calibrations and the
uncertainty of the NIST-traceable flow rate, volume, mass or time reference
standards. The uncertainty of the gaseous reference standard is determined
from its certificate of analysis. The accuracy of the annual multipoint flow rate
calibration is verified by an annual calibration check using nonreactive gaseous
reference standards and a pollutant gas analyzer.

An annual reactive gas calibration check is required for gas dilution

systems that involve reactive gases, such as ammonia (NH3), hydrogen chloride
(HCl), hydrogen sulfide (H2S), oxides of nitrogen (NOX), and sulfur dioxide (SO2).

The gas mixtures generated by gas dilution systems certified under this
protocol may be used for calibrations and audits of ambient air quality analyzers,
including both field and laboratory based instrumentation, and continuous
emission monitors for pollutant monitoring that is required by the regulations in
40 CFR Parts 50, 58, 60, and 753-6.

Note that EPA Method 20552 may also be used for the verification of gas
dilution systems for field instrument calibrations of EPA Methods 3A, 6C, 7E, 10,

97
 EPA Traceability Protocol for Gaseous Calibration Standards

15, 16, 20, 25A and 25B that are required by the regulations in 40 CFR Part 60
as well as performance specifications according to 40 CFR Part 60, Appendix F.

At this time, EPA does not require the regulated community to use NIST-
traceable dynamic gas dilution systems for the calibration of ambient air or
continuous emission monitors that are required by 40 CFR Parts 50, 58, 60, and
75. However, end users may elect to use them for calibrations.

4.1.2 Limitations

This protocol is not intended for the blanket certification of multiple gas
dilution systems that may be identical in operating principle, construction, and
materials. It is intended for the certification of individual gas dilution systems.
This protocol is not intended for the certification of static gas dilution systems.

4.1.3 Assay Apparatus

Figure 4-1 illustrates one possible design of a gas dilution system for the
generation of pollutant gas mixtures using dynamic dilution methods. This
apparatus is designed to allow the quantitative mixing of a pollutant gas with a
zero gas using gas flow control devices. It also is designed to allow the
measurement of the pollutant gas flow rate and the zero gas flow rate. Inert
materials (e.g., Teflon®, stainless steel, borosilicate glass or silanized glass) and
clean, noncontaminating components are to be used in those portions of the
apparatus that are in contact with the pollutant gas and the gas mixture being
generated. The pathways of the gas streams through the gas dilution system are
controlled by rotation of two three-way valves. Pressure regulators, shutoff
valves, and gas flow control devices regulate the flow from the cylinders. The
gas flow control devices may be needle values, capillary tubes, volumetric piston
pumps, thermal mass flow controllers, or other flow control devices. The
pollutant gas and zero gas flow rates are measured at ports that are connected
to the three-way valves. The pollutant gas stream and the zero gas stream must
be mixed in a chamber whose design ensures that the resulting gas mixture is
thoroughly homogenized. The gas mixture output is connected to the sample
line of the pollutant gas analyzer.

Figure 4-1 shows separate flow measurement ports for the pollutant gas
stream and the zero gas stream. In this configuration, a single flow rate
reference standard can be used to measure either the pollutant gas flow rate or
the zero gas flow rate. Such an approach reduces measurement errors arising
from differences in the calibration of multiple reference standards. Alternatively,
the flow rates can be measured at the outlet port of the gas dilution system with
one of the gas streams shut off. The gas dilution system's design must ensure
that the flow rate in one gas stream remains constant when combined with the
other gas stream.

98
 EPA Traceability Protocol for Gaseous Calibration Standards

Pollutant Flow
3-Way Measurement
Gas Valve Port Pollutant Flowrate
Flow Reference
Control Standard
Shutoff
Valve
Pollutant
Gas
Cylinder Diluted Gas
Output Port
Pollutant Gas
Mixing
Analyzer to be
Chamber
Calibrated

Dilution Flow
Measurement
Gas Port Dilution Flowrate
Flow Reference
Control Standard
Shutoff 3-Way
Valve Valve
Zero
Gas
Cylinder Gas Dilution System

Figure 4-1. One possible design of a gas dilution system for the generation
of pollutant gas mixtures using dynamic dilution methods

99
 EPA Traceability Protocol for Gaseous Calibration Standards

If a flow rate reference standard is mounted directly downstream of a flow

control device, it causes back pressure in the gas dilution system and a resulting
change in the flow rate through the flow control device or a resulting increase in
the gas pressure entering the standard. This back pressure may vary as a
function of the flow rate. Thus, the reference standard may need to compensate
for the variable inlet pressure. Thermal mass flow controllers do not need to be
corrected for such pressure effects, but measurements by inlet-pressure-
sensitive, volumetric-flow-rate reference standards must be corrected.

The apparatus may be modified in several ways that will not diminish its
performance. For example, the calibration could be performed using a flow rate
reference standard that is mounted in-line upstream of the flow control device.

4.1.4 Pollutant Gas Analyzer

The pollutant gas analyzer being calibrated must have a well-

characterized calibration curve and must be capable of directly measuring the
concentration of the pollutant gas mixture that is generated by the gas dilution
system. It must have good resolution, good precision, a stable response, and
low output signal noise. An analyzer with acceptable performance specifications
may be selected from the list of EPA-designated reference and equivalent
method analyzers.40 In addition, the analyzer should have good specificity for the
pollutant of interest so that it has no detectable response to any other component
or contaminant that may be contained in the pollutant gas mixture.

If any component in a multiple-component pollutant gas mixture interferes

with the assay of any other component, the analyst must conduct an interference
study to determine an interference correction equation. If multiple balance gases
are used in the gas dilution system (e.g., air versus nitrogen or different
percentages of oxygen in air), it must have been demonstrated that the
analyzer's response is not sensitive to differences in balance gas composition.
This demonstration can be accomplished by showing that no difference exists in
the analyzer's response when measuring a compressed gas calibration standard
that has been diluted with identical flow rates of different balance gases.

The analyzer must be connected to a high-precision data acquisition

system, which must produce an electronic record of the analyzer's response
during the assay. A high-precision digital panel meter, a digital voltmeter, a data
logger or some other data acquisition system with four-digit resolution can be
used to obtain numerical values of the analyzer's response. More precise values
will be obtained if this system has a data-averaging capability.

If the analyzer has not been in continuous operation, turn it on and allow it
to stabilize (e.g., for at least 12 hours) before beginning the measurements.

100
 EPA Traceability Protocol for Gaseous Calibration Standards

4.1.5 Reference Standards

The EPA monitoring regulations define a "traceable" gaseous calibration

standard as one that has been compared and certified, either directly or via not
more than one intermediate standard, to a primary standard such as an SRM or
a CRM.3,4 The monitoring regulations require that calibration gases used for
calibration and audit of ambient air quality analyzers and continuous emission
monitors be traceable to NIST reference standards. Gaseous reference
standards used for assaying and certifying a gas dilution system must be SRMs,
NTRMs, RGMs, PRMs, CRMs or GMISs. See Subsection 2.1.3 for a description
of these standards. This procedure will use a reference standard that is diluted
and a check standard that is not diluted in the annual gas calibration checks.

Zero gas used in gas dilution systems should be clean, dry, zero-grade air
or nitrogen containing no detectable concentration of the pollutant of interest. It
may come from compressed gas cylinders or from zero gas generators. The use
of NIST-traceable zero air material is recommended, but is not required. The
zero gas should match the balance gas in the gaseous reference standard,
unless it has been demonstrated that the pollutant gas analyzer being calibrated
is insensitive to differences in the balance gas composition. It also should
contain no contaminant that causes a detectable response in the analyzer or that
suppresses or enhances the analyzer's response to the pollutant. The oxygen
content of zero air should be approximately that of ambient air, unless it has
been demonstrated that varying the oxygen content does not suppress or
enhance the analyzer's response. The water vapor concentration in the zero gas
should be less than 5 ppm. Information concerning the zero gas (e.g., cylinder
identification number, impurity levels, cylinder pressure, etc.) must be recorded in
the laboratory's records.

The flow rate reference standards for calibrating gas dilution systems
must be traceable to NIST primary flow rate, volume, mass or time reference
standards 54 by original calibration by their manufacturer and by annual
recertification at a NIST-accredited state weights and measures laboratory or at a
calibration laboratory that is accredited by the National Voluntary Laboratory
Accreditation Program (NVLAP), the American Association for Laboratory
accreditation (A2LA) or by the International Laboratory Accreditation Conference
(ILAC) under ISO/IEC 17025 (General Requirements for the Competence of
Testing and Calibration Laboratories).10-12 Their certifications must cover the
range of flow rates that will be used by the gas dilution systems. The expanded
uncertainties (U) of the standards must be documented in a calibration certificate
or they must be available through the manufacturer or calibration facility. Their
expanded uncertainties must not be greater than plus or minus 1 percent of
reading for the flow rates that will be used for gas dilution. Expanded
uncertainties that are much less than plus or minus 1 percent of reading are
desirable so that the uncertainties of the gas mixture concentrations are
decreased. They must be stable, precise, linear, and have good resolution.

101
 EPA Traceability Protocol for Gaseous Calibration Standards

Alternatively, flow rate reference standards or gas dilution systems may

be traceable to the flow rate, volume, mass or time reference standards of other
national metrology institutes (e.g., KRISS in South Korea, NEL in the United
Kingdom, PTB in Germany) provided that these institutes are participants in the
Working Group for Fluid Flow (WGFF) of the International Committee for Weights
and Measures (CIPM).

4.1.6 Calibrations of the Dynamic Gas Dilution System

On an annual basis, three aspects of a gas dilution system's performance

must be evaluated. First, the flow-rate performance must be determined by a
multipoint calibration with NIST-traceable flow rate reference standards. Second,
the multipoint calibration must be verified using a pollutant gas analyzer by direct
comparison of an undiluted, nonreactive gaseous check standard with a diluted,
nonreactive pollutant gas mixture generated by the gas dilution system. For
applications in which the dynamic gas dilution system will be used with reactive
gas mixtures, the inertness of the wetted interior surfaces of the dynamic gas
dilution system is checked using a pollutant gas analyzer by direct comparison of
an undiluted, reactive gaseous check standard with a diluted reactive pollutant
gas mixture generated by the gas dilution system.

On a monthly basis, the calibration of the gas dilution system is checked

for drift by a single-point flow rate check using the same flow rate reference
standards that were used during the annual multipoint flow rate calibration.

4.1.6.1 Annual Multipoint Flow Rate Calibration—

The dynamic gas dilution system must have had a multipoint flow rate
calibration demonstrating acceptable performance within 1 year prior to it being
used to generate pollutant gas mixtures. This calibration will be performed
separately on the pollutant gas and zero gas streams. It will cover the full range
of flow rates that will be used for gas dilution.

The multipoint calibration must consist of one or more measurements of

the dynamic gas dilution system's pollutant gas and zero gas streams at five or
more different flow rates. Each gas stream must be measured separately.
Measurement of each gas stream with the same flow-rate reference standard
reduces the uncertainty that associated with the use of multiple standards. The
flow rates should be approximately evenly spaced over the operating range. The
calibration is valid only for the range lying between the largest and smallest
measured flow rates. Record the reference standard's measurements and the
gas dilution system's flow rate set-point values in the laboratory's records.

If the gas dilution system has multiple flow rate ranges or multiple flow
control devices, a multipoint calibration must be done for all ranges and flow
control devices that will be used later for generating pollutant gas mixtures.

102
 EPA Traceability Protocol for Gaseous Calibration Standards

All volumetric-flow-rate measurements must be corrected for or referenced

to the same temperature and pressure conditions, such as EPA-standard
conditions (i.e., 760 millimeters of mercury (mm Hg), 25 °C) or the ambient
temperature and pressure conditions prevailing in the laboratory during the
calibration. The equation to correct the flow rate for temperature, pressure, and
humidity effects is given below:

Volume   PM − PWV   TS 
Flow rate =    
 Time   PS  TM 
where

PM = measured barometric pressure (mm Hg);

PWV = partial pressure of water vapor (mm Hg);
PS = standard pressure (mm Hg);
TS = standard temperature (298.2 K); and
TM = measured ambient temperature (273.2 + °C).

The data reduction technique used in this procedure is based on the

assumption that the gas dilution system has a well-characterized calibration
curve. The accuracy of the concentration of a diluted pollutant gas mixture is
dependent upon this assumption. The analyst cannot assume that the calibration
curve is a straight line. The analyst must determine the calibration equation and
the uncertainty for its predicted concentrations by statistical analysis of the
measurements obtained during the multipoint calibration (see Appendix F).

Data from the multipoint calibration must be evaluated using least-

squares regression analysis.19 This analysis technique will be used to determine
the flow rate calibration curve and to characterize the uncertainty associated with
the flow rate. The reference standard measurements are the dependent (i.e., Y)
values in the analysis and their units may be liters per minute or any other
appropriate unit. The gas dilution system's settings are the independent (i.e., X)
values in the analysis and their units may be liters per minute or any other
appropriate unit. The measurements and the set-points should have a resolution
that is finer than 1 percent of the maximum measurement or set-point.

Calculate the least-squares regression coefficients using the

spreadsheets described in Appendix F or using an EPA-approved equivalent
statistical technique (see Subsection 2.1.6). If an EPA-approved equivalent
statistical technique is used, it must be identified as such in the certificate
documentation and it must be described in media readily accessible to end
users. Record the regression coefficients and other statistical results in the
laboratory's records.

The spreadsheets allow the multipoint calibration data to be fitted to

straight-line or quadratic linear regression models. EPA discourages the use of

103
 EPA Traceability Protocol for Gaseous Calibration Standards

the cubic and quartic models and believes that better fits of the data can be
obtained by performing multipoint calibrations over more limited concentration
ranges and by using straight-line or quadratic models. Additionally, a multipoint
calibration should not change regression model orders from one year to the next.

The spreadsheet described in Appendix F will suggest the best regression

model for the multipoint calibration data, but the analyst should choose the model
that best fits the measurement process on theoretical grounds.

Because a flow rate reference standard's measurement process has a

random error component, repeated measurements of the same flow controller
setting will not produce identical measured values. The analyst may wish to
investigate the precision by making replicate flow rate measurements at several
different settings. Least-squares regression analysis normally assumes that the
precision is the same at all flow rates. However, this statistical assumption may
not be true for some reference standards and the analyst may need to use
alternate statistical procedures to analyze the multipoint calibration data.

4.1.6.2 Uncertainty of the Flow-Rate Calibration Curve—

The total uncertainty of the pollutant and zero gas flow rates is composed
of several components. The first component is the uncertainty of the flow-rate
reference standards that are used in the annual multipoint flow-rate calibration. It
is obtained from the calibration certificate for the standards. It is normally
presented as the expanded uncertainty (U) and as a percentage of the reading.

The second uncertainty component is the precision of the measurements

of the flow-rate reference standard. It is minimized by increasing the number of
replicate measurements at each flow rate or by increasing the number of different
flow rates used in the multipoint calibration. Additionally, precision can be
improved by using an averaged reference standard response, rather than an
instantaneous response, for each measurement.

The third uncertainty component is the uncertainty associated with the

flow rates that are predicted from the calibration curve. It is a measure of how
well the calibration data fit an equation which the analyst assumes is the "true"
calibration equation for the gas dilution system. The form of the component that
is most directly useful to the analyst is the value of the expanded uncertainty (U)
for a regression-predicted flow rate given one or more measurements of the flow
rate reference standard. See Subsection 2.1.6 for a discussion of U. This value
may be calculated using the spreadsheet described in Appendix F or using an
EPA-approved equivalent statistical technique (see Subsection 2.1.6).
Comparison of values of U from straight-line and quadratic equations permits the
analyst to select the equation that best represents the data.

Two points should be noted about the third uncertainty component. First,
its magnitude decreases as n increases where n is the number of measurements

104
 EPA Traceability Protocol for Gaseous Calibration Standards

in the multipoint flow rate calibration. Second, its magnitude increases as the
mean measured response ( Y ' ) for the flow-rate reference standard diverges from
the overall mean measured response ( Y ) for the multipoint calibration. That is,
the absolute uncertainty of the regression predicted concentration is larger at the
extremes of the calibrated concentration range than at the middle of the range.

A multipoint calibration will be considered to be well-characterized for all

flow rates that are within the range of the multipoint calibration measurements
and for which U for the regression-predicted analyzer response is ≤±1 percent of
the measured response for the largest flow rate in the multipoint calibration. For
example, assume that a multipoint calibration was conducted between 0 and 10
liters per minute. In this example, the calibration is well-characterized for all flow
rates for which U is ≤±0.1 liters per minute, which is equal to ±1 percent of 10
liters per minute.

Note that Cells D29, H27, B44, and H44 of the day-of-use worksheet in
the Appendix F spreadsheet allows the analyst to enter pollutant gas and zero
gas flow controller settings and to obtain the regression-predicted flow rates and
the corresponding relative standard uncertainties (i.e., uc/flow rate) associated
solely with the calibration. To determine whether these predicted flow rates are
in the well-characterized region of the calibration curve, multiply these
uncertainties by the quantity 2 x (largest calibration flow rate/predicted flow rate)
and compare the resulting value to 0.01, which is the acceptance criterion for the
well-characterized region.

A multipoint calibration may fail to meet this uncertainty acceptance

criterion for several possible reasons:

• inadequate analytical precision;

• inaccuracy of the reference standards; or
• excessive uncertainty due to incorrect assumptions about the form of
the calibration equation.

For analyzers having an inherently nonlinear, but precise response, the

effect of excessive uncertainty in a straight-line calibration equation may be
eliminated by using a quadratic calibration equation or by transforming the
nonlinear calibration data mathematically so that they may be fitted to a straight
line regression equation. See Reference 19 for a discussion of such linearizing
transformations of data. The analyst may need to try several different
transformations before the optimum transformation is determined. Using
appropriately transformed calibration data, a calibration equation can be
calculated with an acceptable value of U for the regression-predicted
concentration. Subsequently, data obtained from the annual gas calibration
checks, the monthly flow check, and the day-of-use must be similarly
transformed to obtain regression-predicted flow rates.

105
 EPA Traceability Protocol for Gaseous Calibration Standards

A quantitative assessment of the calibration's uncertainty is needed to

allow the analyst to determine whether the multipoint calibration data adequately
characterizes the "true" calibration curve of the flow controllers. The statistical
parameter to be used to characterize the uncertainty of the multipoint calibration
is U for a flow rate predicted from the regression line using measured flow rates.
This parameter can be calculated using the spreadsheet described in Appendix
F. Record the uncertainty calculations in the laboratory's records.

4.1.6.3 Annual Nonreactive Gas Calibration Check—

Following completion of the multipoint calibration, the accuracy of the
multipoint calibration must be checked to verify that the error associated with the
dilution is not excessive. This calibration check involves the measurement of a
diluted nonreactive reference standard and an undiluted nonreactive check
standard, both of which must be an SRM, an RGM, a PRM, a CRM, an NTRM or
a GMIS as specified in Subsection 2.1.3.1. Both standards must have the same
reactive gas composition. The diluent gas must be the same as the balance gas
for both standards. The check standard must have a certified concentration that
is at least an order of magnitude (i.e., a factor of 10) less than the certified
concentration of the reference standard that is diluted by the gas dilution system.
Its concentration must fall in the well-characterized region of the pollutant gas
analyzer's calibration curve. Information concerning this check standard (e.g.,
cylinder identification number, certified concentration, uncertainty, traceability to
NIST) must be recorded in the laboratory's records.

The reference standard and a zero gas are to be connected to the inlet of
the gas dilution system. The pollutant and zero gas flow controllers' settings
must be adjusted such that the concentration of the diluted reference standard is
within +/- 1.0 percent of the concentration of the undiluted check standard. Note
that the annual gas check worksheet in the Appendix F spreadsheet or an EPA-
approved equivalent statistical technique (see Subsection 2.1.6) allows the
analyst to enter the reference standard concentration, the check standard
concentration, and various pollutant and zero gas flow rates to match the two
concentrations. The worksheet also determines the correct settings for the
pollutant and zero gas flow controllers to match the two concentrations and the
relative concentration difference (RCD) between the two concentrations. That is

RCD = 100 (Diluted Concentration - Undiluted Concentration)

(Undiluted Concentration)

Make three or more discrete measurements of both the undiluted check

standard and the diluted reference standard using the pollutant gas analyzer.
"Discrete" means that the analyst must change the gas mixture being sampled by
the analyzer between measurements. For example, the analyst might alternate
between measurements of the undiluted check standard and the diluted
reference standard. Document these measurements in the laboratory's records.

106
 EPA Traceability Protocol for Gaseous Calibration Standards

Calculate the mean and standard deviation of both sets of measurements using
the annual gas check worksheet in the Appendix F spreadsheet.

Next the analyst must verify that the systematic error (bias) of the dilution
is not excessive by comparing the mean analyzer responses for the diluted
reference standard and the undiluted check standard. Calculate the relative
agreement between the two responses using the annual gas check worksheet in
the Appendix F spreadsheet. That is,

Relative Agreement = (Mean Diluted Response - Mean Undiluted Response)

(Mean Undiluted Response)

The acceptance criterion for the relative agreement includes the relative
standard uncertainties for the concentrations of the reference and check
standards, the uncertainties of the pollutant and zero gas flow rates for the
reference standard, the relative standard deviations for the individual
measurements of the diluted reference standard and undiluted check standard,
and the RCD. That is,

2 2 2
 uref   ucheck   u flow1   u flow 2   sconc1   sconc 2
2 2 2

Criterion =
RCD + 2  +
   + +
  +
   + 
 ref   check   flow1   flow2   meanconc1   meanconc 2 

This calculation is performed using the annual gas check worksheet in the
Appendix F spreadsheet.

Document the RCD, the relative agreement, and the acceptance criterion
in the laboratory's records. If the agreement is greater than the acceptance
criterion, the dilution error is considered to be excessive. The analyst must
investigate why the relative difference is excessive. The problem may be due to
errors in the reference standard and check standard concentrations, errors in
assay apparatus or to some other source. Pollutant gas analyzer calibrations
may not be conducted until the agreement for a subsequent gas check is less
than or equal to the criterion for that gas check.

4.1.6.4 Annual Reactive Gas Calibration Check—

For applications in which the dynamic gas dilution system will be used
with reactive gas mixtures [including ammonia (NH3), hydrogen chloride (HCl),
hydrogen sulfide (H2S), nitric oxide (NO), oxides of nitrogen (NOX), and sulfur
dioxide (SO2)], the annual gas calibration check must be repeated for each of the
reactive gas mixtures. This additional calibration check is needed because
reactive gases may be lost to the interior walls of a dynamic gas dilution system
that does not cause such wall losses for nonreactive gas mixtures55. It involves
the measurement of a diluted reactive reference standard and an undiluted
reactive check standard, both of which must be SRMs, RGMs, PRMs, CRMs,
NTRMs or GMISs as specified in Subsection 2.1.3.1. Both standards must have
the same reactive gas composition. The diluent gas must be the same as the

107
 EPA Traceability Protocol for Gaseous Calibration Standards

balance gas for both reactive gas standards. The undiluted reactive check
standard must have a certified concentration that is at least an order of
magnitude less (i.e., a factor of 10) than the concentration of the reactive
reference standard that is diluted. Information concerning both standards (e.g.,
cylinder identification number, certified concentration, uncertainty, traceability to
NIST) and the diluent gas must be recorded in the laboratory's records.

The reference standard and a zero gas are to be connected to the inlet of
the gas dilution system. The pollutant and zero gas flow controllers' settings
must be adjusted such that the concentration of the diluted reference standard is
within +/- 1.0 percent of the concentration of the undiluted check standard. Note
that the annual gas check worksheet in the Appendix F spreadsheet allows the
analyst to enter the reference standard concentration, the check standard
concentration, and various pollutant and zero gas flow rates to match the two
concentrations. The worksheet also determines the correct settings for the
pollutant and zero gas flow controllers to match the two concentrations and the
relative concentration difference (RCD) between the two concentrations. That is

RCD = 100 (Diluted Concentration - Undiluted Concentration)

(Undiluted Concentration)

Make three or more discrete measurements of the diluted reactive

reference standard and of the undiluted reactive check standard. Document
these measurements in the laboratory's records. Calculate the mean and
standard deviation of both sets of measurements using the annual gas check
worksheet in the Appendix F spreadsheet.

Next the analyst must verify that the dilution-plus-wall-loss error is not
excessive by comparing the mean analyzer responses for the diluted reference
standard and the undiluted check standard. Calculate the relative agreement
between the two responses using the annual gas check worksheet in the
Appendix F spreadsheet or using an EPA-approved equivalent statistical
technique (see Subsection 2.1.6). That is,

Relative Agreement = (Mean Diluted Response - Mean Undiluted Response)

(Mean Undiluted Response)

The acceptance criterion for the relative agreement includes the relative
standard uncertainties for the concentrations of the reference and check
standards, the uncertainties of the pollutant and zero gas flow rates for the
reference standard, the relative standard deviations for the individual
measurements of the diluted reference standard and undiluted check standard,
and the RCD. That is,

2 2 2
 uref   ucheck   u flow1   u flow 2   sconc1   sconc 2
2 2 2

Criterion =
RCD + 2  +
   + +
  +
   + 
 ref   check   flow1   flow2   meanconc1   meanconc 2 

108
 EPA Traceability Protocol for Gaseous Calibration Standards

This calculation is performed using the annual gas check worksheet in the
Appendix F spreadsheet.

Document the RCD, the relative agreement, and the acceptance criterion
in the laboratory's records. If the agreement is greater than the acceptance
criterion, the dilution error is considered to be excessive. The analyst must
investigate why the relative difference is excessive. The problem may be due to
errors in the reference standard and check standard concentrations, errors in
assay apparatus or to some other source. Pollutant gas analyzer calibrations
may not be conducted until the agreement for a subsequent gas check is less
than or equal to the criterion for that gas check.

4.1.6.5 Monthly Flow Rate Quality Control Check—

The dynamic gas dilution system must have had a single-point flow rate
quality control (QC) check demonstrating acceptable performance within 1 month
prior to the day that it is used to generate pollutant gas mixtures to calibrate
pollutant gas analyzers. This QC check need not be done every month, but only
in those months in which the system will be used. It will be performed separately
on the pollutant gas and zero gas streams. The QC check is valid only for the
flow-rate range that was measured during the annual multipoint calibration and
not other flow-rate ranges. It will be performed at a single flow rate that lies in
the middle of the range that was measured during the annual multipoint
calibration.

If the dynamic gas dilution system has multiple flow-rate ranges or

multiple flow control devices, a single-point flow rate QC check must be done for
all ranges and flow control devices that will be used later for generating pollutant
gas mixtures.

The single-point flow rate QC check must consist of one or more

measurements of the gas dilution system's pollutant gas and zero gas streams.
Each gas stream must be measured separately. Measurement of each gas
stream with the same flow-rate reference standard reduces measurement errors
that are associated with the use of multiple standards. The zero gas can be used
in the place of the pollutant gas for this QC check if the analyst wished to
conserve the pollutant gas. Record the reference standard's measurements and
the gas dilution system's flow rate set-point values in the laboratory's records.

Next the analyst must verify that the flow-rate shift is not excessive. For
the annual multipoint calibration measurements and the monthly QC check
measurements, calculate the relative difference (in percent) between the mean
QC check flow rate and the corresponding flow rate that is predicted from the
annual multipoint calibration regression equation. That is,

Relative Difference = 100 (Mean Analyzer Response - Predicted Response)

(Predicted Response)

109
 EPA Traceability Protocol for Gaseous Calibration Standards

This relative difference is calculated using the monthly flow check

worksheet in the Appendix F spreadsheet. Document the relative difference in
the laboratory's records. If the relative difference is greater than 1.0 percent, the
calibration error is considered to be excessive. The analyst must investigate why
the relative difference is excessive. The problem may be due to errors in the
reference standard and check standard concentrations, errors in assay
apparatus or to some other source. Assays may not be conducted until the
relative difference for a subsequent check is less than or equal to 1.0 percent.

4.1.7 Day-of-Use Calculations

After the annual flow-rate calibrations, the annual nonreactive and

reactive gas checks, and the monthly flow-rate checks have been successfully
completed, the day-of-use worksheet in the Appendix F worksheet or an EPA-
approved equivalent statistical technique (see Subsection 2.1.6) can be used to
determine the correct flow-controller settings that are needed to obtain a desired
pollutant gas concentration and to calculate the expanded uncertainty of this
concentration. Because data from the annual multipoint calibration already has
been entered in the spreadsheet, the linear regression parameters obtained in
the calibration will be used in these calculations. The pollutant gas that is used
may differ from that used in the annual flow-rate calibrations and the annual
nonreactive and reactive gas checks, but the pollutant gas and zero gas flow
rates that are selected must fall within the well characterized region of the
respective calibration curves.

Note that Cells D29, H27, B44, and H44 of the day-of-use worksheet in
the Appendix F spreadsheet allows the analyst to enter pollutant gas and zero
gas flow controller settings and to obtain the regression-predicted flow rates and
the corresponding relative standard uncertainties (i.e., uc/flow rate) associated
solely with the calibration. To determine whether these predicted flow rates are
in the well-characterized region of the calibration curve, multiply these
uncertainties by the quantity 2 x (largest calibration flow rate/predicted flow rate)
and compare the resulting value to 0.01, which is the acceptance criterion for the
well-characterized region.

4.1.8 Expanded Uncertainty of the Concentration at the Output of the System

The expanded uncertainty (U) of the diluted concentration at the output of

the gas dilution system is the combined standard uncertainty (uC) multiplied by a
coverage factor (k), which is equal to 2.8,9 It is due to many different error
sources, including the uncertainty in the flow-rate reference standard(s),
uncertainty in the annual multipoint calibration, uncertainty in the annual
nonreactive or reactive gas calibration check, uncertainty in monthly flow-rate QC
checks, and random measurement error. The spreadsheet in Appendix F
combines these uncertainty components and calculates U. There may be
additional uncertainty sources that cannot be assessed with the limited data that

110
 EPA Traceability Protocol for Gaseous Calibration Standards

are produced when implementing this protocol. The calculated value of U must
be given in the certification documentation. The use of an uncalculated or
blanket estimate of U (e.g. +/- 1 percent) is not acceptable.

As a rule-of-thumb, U should be rounded to one significant figure unless

the leading figure is a 1 in which case two significant figures should be reported8.

The estimate of U (e.g., +/- 20 ppm) may be supplemented, but not

replaced) by a statement of the equivalent fractional uncertainty (e.g., +/- 1
percent), if desired, using the following equation:

UFRACTIONAL = 100(U/ Concentration of the System Output)

The same rule-of-thumb regarding significant figures applies to the

fractional uncertainty.

If an analyst wishes to use an equivalent statistical technique to calculate

the expanded uncertainty, the analyst must first submit a detailed description of
the technique and any explanatory software to EPA Traceability Protocol Project,
Technical Services Branch, U.S. EPA, Mail Code E343-03, Research Triangle
Park, NC 27711 for statistical evaluation and approval. The description must be
in sufficient detail to demonstrate equivalence to the spreadsheets described in
Appendix F. Example calculations demonstrating equivalence must be included.
Any information about the technique that is submitted to EPA will be treated as
confidential business information. Upon approval by EPA, this equivalent
statistical technique must be identified as such in the certificate documentation
and it must be described in media readily accessible to end users.

4.1.9 Certification Documentation for the Dynamic Gas Dilution System

For each dynamic gas dilution system that is assayed using this protocol,
the assay results must be documented in a written report, which contains at least
the following information:

1. A statement that the assay/certification was performed according to this

protocol;

2. The manufacturer, the model, and the serial number of the dynamic gas
dilution system;

3. The flow-rate ranges for which the dynamic gas dilution system is certified
based on the annual multipoint calibration;

4. The dynamic gas dilution system's calibration equation based on the annual
multipoint calibration. The calibration equation can be supplemented by a
graphical representation of the calibration curve if desirable;

111
 EPA Traceability Protocol for Gaseous Calibration Standards

5. A calculated estimate of U for the dynamic gas dilution system. The use of an
uncalculated, blanket value for U (e.g., +/- 1 percent) is not acceptable. The
estimate is expressed as a 95-percent confidence interval, which is the
combined standard uncertainty (uc) multiplied by a coverage factor (k), which
is equal to 2.8,9 It must include the uncertainty in the flow-rate reference
standard(s), uncertainty in the annual multipoint calibration, uncertainty in the
annual nonreactive or reactive gas calibration check, uncertainty in monthly
flow-rate QC checks, and random measurement error. Use the spreadsheet
described in Appendix F or an EPA-approved equivalent statistical technique
to calculate U. If an EPA-approved equivalent statistical technique is used
rather that the Appendix F spreadsheet, it must be identified as such in the
certificate documentation and must be described in media readily accessible
to end users. As a rule-of-thumb, U should be rounded to one significant
figure unless the leading figure is a 1 in which case two significant figures
should be reported12. The estimate of U (e.g., +/- 20 milliliters per minute)
may be supplemented by a statement of the equivalent fractional uncertainty
(e.g., +/- 1.0 percent) if desired (see Section 4.1.5);

6. Dates of the annual multipoint calibration;

7. Certification expiration date (i.e., the multipoint calibration date plus one year
plus one day);

8. Information about the flow-rate reference standard(s) (i.e., the manufacturer,

the model, the serial number, and measurement principle) that was used in
the multipoint calibration and the QA checks, its traceability to NIST, and the
date of its most recent certification;

9. The relative difference(s) that was calculated from the annual nonreactive
(and reactive) gas calibration check(s);

10. Information about the undiluted reference standard and the diluted check
standard that were used in the annual nonreactive (and reactive) gas
calibration check(s): NIST SRM number, NIST sample number, cylinder
identification number, certified concentration, expanded uncertainty, and
certification expiration date for an SRM; cylinder identification number,
certified concentration, expanded uncertainty, and certification expiration date
for an NTRM, an RGM, a PRM, a CRM or a GMIS. The certification
documentation must identify the reference standard as being an SRM, an
NTRM, an RGM, a PRM, a CRM or a GMIS. For a GMIS, the reference
standard that was used for its assay must be identified in the documentation
for the candidate standard being certified.

11. Identification of the dynamic gas dilution system manufacturer or other

laboratory (i.e., laboratory name, laboratory location, and PGVP vendor ID

112
 EPA Traceability Protocol for Gaseous Calibration Standards

issued by EPA for the production location) where the dynamic gas dilution
system was calibrated, even if another organization will sell it to an end user.
This identification must be given in the same or larger font as the other
required information in the report.

This certification documentation must be given to the end user of the

standard. The dynamic gas dilution system manufacturer or other laboratory that
calibrated the system must maintain laboratory records and certification
documentation for at least one year after the annual multipoint calibration. A
distributor of dynamic gas dilution systems may redocument a system that it has
purchased from another manufacturer and that it wishes to sell to a third party.
However, the new certification documentation must clearly identify the
manufacturer or other laboratory (i.e., name and location) where the system was
calibrated and present all the information that is contained in the original report.

4.1.10 Certification Label

A label or tag must be attached to the dynamic gas dilution system

bearing the information described in Items 1-5, 7, and 11 of Subsection 4.1.6.

4.1.11 Certification Periods for the Dynamic Gas Dilution System

The certification of a dynamic gas dilution is valid for only one year
following its annual multipoint calibration.

4.1.12 Technical Variances

Gas dilution system producers, standard users, and other analytical

laboratories may petition the U.S. EPA for technical variances to the assay
procedures in this protocol. A technical variance allows the use of a specific
alternative assay procedure for gas dilution systems, which can be certified
under this protocol. The petitioner must send a written request with a detailed
description of the alternative assay procedure and supporting analytical data to
EPA Traceability Protocol Project, Technical Services Branch, U.S. EPA, Mail
Code E343-03, Research Triangle Park, NC 27711. The supporting analytical
data must demonstrate the equivalence of the alternative assay procedure with
the procedures given in this protocol. Schuirmann's TOST13-18 is the
recommended statistical technique to demonstrate such equivalence if
technically possible.

113
 EPA Traceability Protocol for Gaseous Calibration Standards

Section 5 - References

1. U.S. Environmental Protection Agency. "EPA Traceability Protocol for

Assay and Certification of Gaseous Calibration Standards." EPA/600/R-
97/121, Research Triangle Park, NC. 1997. 52 pp.

2. U.S. Environmental Protection Agency. Code of Federal Regulations.

Title 40, Chapter I, Subchapter C, Part 50. "National Primary and
Secondary Ambient Air Quality Standards." Washington, DC. Office of
the Federal Register. July 1, 2010.

3. U.S. Environmental Protection Agency. Code of Federal Regulations.

Title 40, Chapter I, Subchapter C, Part 58. "Ambient Air Quality
Surveillance." Washington, DC. Office of the Federal Register. July 1,
2010.

4. U.S. Environmental Protection Agency. Code of Federal Regulations.

Title 40, Chapter I, Subchapter C, Part 60. "Standards of Performance for
New Stationary Sources." Washington, DC. Office of the Federal
Register. July 1, 2010.

5. U.S. Environmental Protection Agency. Code of Federal Regulations.

Title 40, Chapter I, Subchapter C, Part 75. "Continuous Emission
Monitoring." Washington, DC. Office of the Federal Register. July 1,
2010.

6. F.R. Guenther, W.D. Dorko, W.R. Miller, and G.C. Rhoderick. The NIST
Traceable Reference Program for Gas Standards. National Institute of
Standards and Technology. Special Publication 260-126. 1996. 40 pp.

7. U.S. Environmental Protection Agency. Code of Federal Regulations.

Title 40, Chapter I, Subchapter C, Part 72. "Acid Rain Program General
Provisions." Washington, DC. Office of the Federal Register. July 1,
2010.

8. American National Standards Institute (ANSI) and National Conference of

Standards Laboratories (NCSL), U.S. Guide to Expression of Uncertainty
in Measurement, ANSI NCSL Z540.2 1997 R2007, 1997.

9. B.N. Taylor and C.E. Kuyatt. Guidelines for Evaluating and Expressing the
Uncertainty of NIST Measurement Results: 1994 Edition. National
Institute of Standards and Technology. Technical Note 1297. 1994.
(http://physics.nist.gov/pubs/guidelines/TN1297/tn1297.pdf, accessed
April18, 2011).

115
 EPA Traceability Protocol for Gaseous Calibration Standards

10. G.L. Harris. State Weights and Measures Laboratories: State Standards
Program Description. National Institute of Standards and Technology.
Special Publication 791. 1994. 130 pp.

11. V.R. White. National Voluntary Laboratory Accreditation Program: 1997

Directory. National Institute of Standards and Technology. Special
Publication 810. 1997. 225 pp.

12. International Organization for Standardization (ISO) and International

Electrotechnical Commission (IEC). General Requirements for the
Competence of Testing and Calibration Laboratories, Second Edition,
ISO/IEG 17025:2005(E), Geneva, Switzerland, 2005.

13. D.J. Schuirmann. "A Comparison of the Two One-Sided Tests Procedure
and the Power Approach for Assessing the EQuivalence of Average
Bioavailability", Journal of Pharmokinetics and Biopharmaceutics. 15(6):
657-680. 1987.

14. S.J. Richter and C. Richter. "A Method for Determining Equivalence in
Industrial Applications", Quality Engineering, 14(3):375-380. 2002.

15. H. Rita and P. Ekholm. "Showing Similarity of Results given by Two

Methods: A Commentary", Environmental Pollution, 145(2):383-386. 2007.

16. G.B. Limentani et al. "Beyond the t-Test: Statistical Equivalence Testing",
Analytical Chemistry. 77(11):221A-226A. 2005.

17. D. Chambers et al. "Analytical Method Equivalency: An Acceptable

Analytical Practice", Pharmaceutical Technology. 29(9):64-80. 2005.

18. K.R.Lung et al. "Statistical Method for the determination of Equivalence of

Automated Test Procedures", Journal of Automated Methods and
Management in Chemistry. 25(6):123-127. 2003.

19. M.G. Natrella. Experimental Statistics, National Bureau of Standards

Handbook No. 91, U.S. Government Printing Office. Washington, DC.
1963. pp. 5-1 to 5-46 (reprinted by Dover Publications under the same
title, 2005).

20. J. R. Taylor. Introduction to Error Analysis: The Study of Uncertainties in

Physical Measurements, Second Edition, University Science Books,
Sausalito, CA, 1997.

21. W.D. Dorko. personal communication, 2011.

116
 EPA Traceability Protocol for Gaseous Calibration Standards

22. R.C. Shores, F. Smith, and D.J. von Lehmden. "Stability Evaluation of
Sulfur Dioxide, Nitric Oxide and Carbon Monoxide Gases in Cylinders."
EPA/600/4-86, U.S. Environmental Protection Agency, Research Triangle
Park, NC. 1984. 52 pp.

23. S.G. Wechter. "Preparation of Stable Pollution Gas Standards Using

Treated Aluminum Cylinders." In Calibration in Air Monitoring, ASTM STP
598, American Society for Testing and Materials. Philadelphia, PA. 1976.
pp. 40-54.

24. National Council of the Paper Industry for Air and Stream Improvement,
Inc. "An Investigation of the Stability of H2S in Air Cylinder Gases."
NCASI Special Report No. 90-09, New York, NY. 1990. 13 pp.

25. W.R. Miller and G.C. Rhoderick. "Stability of compressed gas mixtures
containing low level volatile organic compounds in aluminum cylinders,"
Fresenius J Anal Chem. 351(2-3):221-229. 1995.

26. G.C. Rhoderick. " Long-term stability of hydrocarbons in NIST gas

standard reference material (SRM) 1800," Anal Bioanal Chem. 383(1):98-
106. 2005.

27. G.C. Rhoderick. "Development of a 100 nmol mol−1 propane-in-air SRM

for automobile-exhaust testing for new low-emission requirements," Anal
Bioanal Chem. 387(7):2425–2432. 2007.

28. W.R. Miller and F.R. Guenther. "The certification of a nominal 20 μmol/mol
H2S/N2 standard reference material using two independent methods," Anal
Bioanal Chem. 387(7):2433–2439. 2007.

29. K. Nishi. "Preparation of Stable Hydrogen Chloride Standard Gases at

Low Concentrations," presented at 8th APMP/TCQM Gas CRM Workshop.
2010 (see www.nmij.jp/~nmijclub/gas/docimgs/A0-
01_Agenda_APMP_GAS_CRM_WS_NMIJ_2010.pdf, accessed August 1, 2011)

30. R.S. Wright et al. Performance Audits of EPA Protocol Gases and
Inspection and Maintenance Calibration Gases, Journal of the Air Pollution
Control Association, 37(4):384-385, 1987.

31. R.S. Wright et al. Accuracy Assessment of EPA Protocol Gases in 1988,
Journal of the Air and Waste Management Association, 39(9):1225-1227,
1989.

32. E.A. Coppedge et al. Accuracy Assessment of EPA Protocol Gases

Purchased in 1991, Journal of the Air and Waste Management
Association, 42(12):1617-1619, 1992.

117
 EPA Traceability Protocol for Gaseous Calibration Standards

33. R.C. Shores et al. "Stability Evaluation of Multicomponent EPA Protocol

Gases", Measurement of Toxic and Related Air Pollutants, Air and Waste
Management Association Publication No. VIP-39, U.S. EPA Publication
No. EPA-600/R-94/136, 1994.

34. A.P. Hines, Analysis of Protocol Gases: An On-Going Quality Assurance

Audit (Results through January 1998),
http://www.epa.gov/ttn/emc/ord/ana-pro.pdf, accessed April 11, 2011.

35. U.S. Environmental Protection Agency. EPA Needs an Oversight

Program for Protocol Gases. Office of Inspector General Report No. 09-
P-0235, 2009. (http://www.epa.gov/oig/reports.htm, accessed April 18,
2011).

36. U.S. Environmental Protection Agency. U.S. EPA Ambient Air Monitoring
Protocol Gas Verification Program, Annual Report, CY 2010, Office of Air
Quality Planning and Standards, 2011
(www.epa.gov/ttn/amtic/files/ambient/qaqc/aagvp2010report.pdf,
accessed April 18, 2011).

37. National Institute of Standards and Technology. Environmental Protection

Agency Blind Audit 2010, NIST Report of Analysis 639.03-11-026a, 2011
(http://www.epa.gov/airmarkets/emissions/docs/Final%202010%20Audit%
20Report.pdf, accessed April 11, 2011).

38. U.S. Environmental Protection Agency. Protocol Gas Verification Program

and Minimum Competency Requirements for Air Emission Testing,
Federal Register, Vol. 76, No. 59, Pages 17288-17325, Monday, March
28, 2011

39. American Society for Testing and Materials. Standard Practice for Dealing
with Outlying Observations. ASTM Standard Practice E 178-80, 1980.

40. U.S. Environmental Protection Agency. List of Designated Reference and

Equivalent Methods. Current edition available from U.S. Environmental
Protection Agency, National Exposure Research Laboratory, Human
Exposure and Atmospheric Sciences Division (MD-D205-03),, Research
Triangle Park, NC 27711 or from the Ambient Monitoring Technology
Information Center (AMTIC) at http://www.epa.gov/ttn/amtic/criteria.html.

41. American National Standards Institute (ANSI) and American Society for
Testing and Materials (ASTM). Standard Specification for Laboratory
Weights and Precision Mass Standards. ANSI/ASTM Standard E 617-91,
1991.

118
 EPA Traceability Protocol for Gaseous Calibration Standards

42. W. Kupper. "High Accuracy Mass Measurements, From Micrograms to

Tons," Instrument Society of America Transactions. 29(4). 1990.

43. G. Harris. "Ensuring Accuracy and Traceability of Weighing Instruments,"

ASTM Standardization News. 21(4):44-51. 1993.

44. J.A. Wise. NIST Measurement Services: Liquid-in-Glass Thermometer

Calibration Service. National Institute of Standards and Technology
Special Publication 250-23. 1988.

45. J.A. Wise. A Procedure for the Effective Recalibration of Liquid-in-Glass

Thermometers. National Institute of Standards and Technology Special
Publication 819. 1991.

46. G. D. Mitchell, W. D. Dorko and P. A. Johnson. "Long-term stability of

sulfur dioxide permeation tube standard reference materials", Fresenius'
Journal of Analytical Chemistry, 344(6):229-233. 1992.

47. G. D. Mitchell. “A Review of Permeation Tubes and Permeators”,

Separation and Purification Methods, 29(1), 2000, 119-128.

48. E.E. Hughes et al. "Performance of a Nitrogen Dioxide Permeation

Device." Analytical Chemistry. 49(12):1823-1829. 1977.

49. W.J. Mitchell et al. “Simple Systems for Calibrating and Auditing SO2
Monitors at Remote Sites.” Atmospheric Environment. 26A(1):191-194.
1992.

50. F.P. Scaringelli et al. “Preparation of Known Concentrations of Gases and

Vapors with Permeation Devices Calibrated Gravimetrically.” Analytical
Chemistry. 42(8):871-876. 1970.

51. D.L. Williams. "Permeation Tube Equilibration Times and Long-Term

Stability." Calibration in Air Monitoring. ASTM Publication STP 598.
American Society for Testing and Materials. 1976. pp. 183-197.

52. U.S. Environmental Protection Agency. Method 205 - Verification of Gas

Dilution Systems for Field Instrument Calibrations
(www.epa.gov/ttn/emc/methods/method205.html, accessed August 1,
2011).

53. EURACHEM/CITAC Guide Use of Uncertainty Information in Compliance

Assessment, First Edition. A Williams, S L R Ellison, M Roesslein (eds.)
ASIN: B002AD0U9Q. 2007. Available from the Eurachem Secretariat (see
http://www.eurachem.org, accessed August 1, 2011).

119
 EPA Traceability Protocol for Gaseous Calibration Standards

54. J.D. Wright. "Laboratory Primary Standards", Flow Measurement, Second

Edition. 2001. D. W. Spitzer, Editor. ISA - The Instrumentation, Systems,
and Automation Society, Research Triangle Park, NC. ISBN 1-55617-
736-4.

55. R.S. Wright and R.W. Murdoch. "Labratory Evaluation of Gas Dilution
Systems for Analyzer Calibration and Calibration Gas Analysis", Air &
Waste, 44(4):428-430. 1994.

120
 EPA Traceability Protocol for Gaseous Calibration Standards

Appendix A - Statistical Spreadsheet for

Procedures G1 and G2
1. ReadMe Worksheet

Purpose

This worksheet supports linear, quadratic, cubic, and quartic models:

Linear: y = 0 + 1x+ 
Quadratic:y = 0 + 1x + 2 x2 + 
Cubic: y = 0 + 1 x + 2 x2 + 3x3 + 
Quartic: y = 0 + 1 x + 2 x2 + 3 x3 + 4x4 + 

The inputs are:

x = the concentration of the certified calibration standard

y = the measured instrument response
 = a random effect due to variation of influence quantities

It is assumed that the concentrations of the calibration standards (x) have

negligible uncertainties.

The output quantities are estimated parameters (s) and related uncertainties.
The workbook then helps the user perform the following functions:

• determine which model (linear, quadratic, etc.) is better

• determine the replication of unknowns needed for uncertainty control
• determine whether zero and span responses are acceptable
• estimate the concentration and expanded uncertainty of candidate
standards analyzed on the same day as the initial calibration or a
subsequent day.

Warning

EPA discourages the use of cubic and quartic models and believes that better fits
of the data can be obtained by performing multipoint calibrations over more
limited concentration ranges and by using straight-line or quadratic models.
Inclusion of cubic and quartic models in the spreadsheet is for experimental use
or for situations in which there is a theoretical basis for the use of such higher-
order models. Analysts should be aware that apparent higher-order calibration
curves may be caused by artifacts such as inaccurate reference standards or
leaks in a gas dilution system. They should not use higher-order regression
models to fit multipoint calibration data that have inadequate precision and that
should be fitted to lower-order regression models. If analysts suspect that

121
 EPA Traceability Protocol for Gaseous Calibration Standards

precision is inadequate, they should make replicate measurements at each

different concentration. Additionally, a multipoint calibration should not change
regression model orders from one month to the next.

Organization

The workbook consists of several worksheets, which are displayed as tabs at the
bottom of the screen. The functions of these worksheets are described below:

ReadMe describes the workbook, explaining how to use the

worksheets
Measurement Data allows for user input of calibration and other analytical
data and includes statistical calculations for
polynomial regression
Curves 1 displays the calibration data, the best-fit line, and its
confidence bands
Residuals 1 displays the difference between the observed
responses and those estimated by the best-fit
calibration line
Curves 2 displays the calibration data, the best-fit quadratic
curve, and its confidence bands
Residuals 2 displays the difference between the observed
responses and those estimated by the quadratic
regression line
Curves 3 displays the calibration data, the best-fit cubic curve,
and its confidence bands
Residuals 3 displays the difference between the observed
responses and those estimated by the best fit cubic
regression line
Curves 4 displays the calibration data, the best-fit cubic curve,
and its confidence bands
Residuals 4 displays the difference between the observed
responses and those estimated by the best-fit quartic
regression line
Chart Data includes the data used to create the Curves
and Residuals charts.

Conventions

The Measurement Data worksheet includes instructions that guide the user
through the steps in its use. The worksheet is also color coded to simplify use.
Shaded cells that are bordered in blue lines are for input of data. These cells are
unprotected, but all other cells of the Measurement Data worksheet are
protected. The only other unprotected cell in the workbook is cell F4 of the Chart
Data worksheet. That cell controls the width of the confidence bands plotted in
the Curves 1 and Curves 2 charts.

122
 EPA Traceability Protocol for Gaseous Calibration Standards

Derived values and statements are colored red. These cells contain formulas
and are protected to prevent alteration.

Spreadsheet background colors indicate the order of the polynomial supported

by the calculations in the area.

Light green is used for the linear model.

Yellow is used for the quadratic model.
Gray is used for the cubic model
Light blue is used for the quartic model.

Use

The Measurement Data worksheet guides the user through six steps.

STEP 1 Enter Calibration Data

In this step, up to 50 calibration points may be entered. Each calibration point

has two parts– the certified concentration of the calibration gas standard and the
instrument response when testing the standard. These values are entered in two
columns. The spreadsheet performs computations in Columns I through P
(linear), Q through X (quadratic), Y through AZ (cubic), and BA and above
(quartic).

STEP 2 Review the Parameter Estimates

In this step, the user reviews the estimates of the intercepts (0), slopes (1) and
other coefficients (2, 3 , and 4 ) for the four models, examines their confidence
intervals and the residual error variances (s2). The result of an F-test indicates
which of the models is best. The linear model is recommended unless the
quadratic or higher-order model significantly reduces the residual error.

STEP 3 Review the Charts

In this step, the user reviews the charts named Curves 1, Residuals 1, Curves 2,
Residuals 2, etc. These charts help the user understand why one model
performs better than the other.

STEP 4 Assess Magnitude of Uncertainty

In this step, the user enters the assumed concentration of a candidate standard
and selects a replication number, r. Based on the calibration results, the
worksheet estimates the expanded uncertainty associated with an approximate
level of confidence of 95% that would result from measuring such a standard r

123
 EPA Traceability Protocol for Gaseous Calibration Standards

times. The user can use this as a guide for deciding whether to proceed with
analysis, producing additional calibration points, or taking some corrective action.

STEP 4a Check Uncertainty Associated with Dilution

Following completion of the multipoint calibration, the accuracy of the assay

apparatus must be checked to verify that the error associated with the dilution is
not excessive. This accuracy check involves the measurement of an undiluted or
diluted check standard.

Make 3 or more discrete measurements of the undiluted or diluted check

standard. Enter the response values in the step 4a section of the Measurement
Data worksheet. The spreadsheet calculates the relative difference between the
mean analyzer response and the corresponding response from the multipoint
calibration regression equation and the undiluted or diluted check standard
concentration. If the relative difference is greater than 1.0%, the dilution error is
considered to be excessive. Assays may not be conducted until the relative
difference for a subsequent accuracy check is less than or equal to 1.0%.

STEP 5 Assay Candidate Standard on Same Day

In this step, the user enters the responses to a candidate standard that is tested
on the same day as the calibration of STEP 1. The worksheet provides an
estimate of the candidate's concentration and its 95% uncertainty. The
worksheet also indicates whether the variability in responses is larger than
expected (unacceptable).

STEP 6 Assay Candidate Standard on Different Day from Initial Calibration

In this step, the user enters the responses to a candidate standard that is tested
on a different day from the calibration of STEP 1. The worksheet first assesses
the zero and span responses. If the zero and span responses are acceptable,
the user proceeds to enter the results from testing a candidate standard. The
results include those for zero and nonzero reference standards. The table below
specifies the minimum number of different reference standard concentrations
needed for each model, in addition to the zero standard. Failure to meet these
requirements will result in serious calculation errors and invalid results.

Model Minimum number of different nonzero reference standards required

Linear 1
Quadratic 2
Cubic 3
Quartic 4

Whenever possible the concentrations of the additional reference standards

should be located at points which correspond to large differences between the

124
 EPA Traceability Protocol for Gaseous Calibration Standards

polynomial calibration curve and the corresponding straight line between the zero
gas and the highest-concentration standard.

The spreadsheet determines whether the regression curve has changed since
the initial calibration. The data are corrected for any change and the estimated
concentration of the candidate standard is provided together with its expanded
uncertainty associated with an approximate level of confidence of 95%.

The spreadsheet also determines whether the relative combined standard

uncertainty of the mean response is acceptable (less than 1% of the mean
response). This additional check is meant to guard against hysteresis or other
errors that are not corrected by the spreadsheet's adjustments.

2. Measurement Data Worksheet

This spreadsheet is for the EPA protocol gases calculation.

STEP 1 Enter Calibration Data

Enter the calibration data in the shaded spaces below. The first column (I)
simply counts the calibration points that you enter. The second column (X) is for
the certified concentrations of the calibration gas standards. The third column
(Y) is for the instrument responses corresponding to the calibration standards.
The number of points cannot exceed 50.

i Xi Yi Color Code
1 0.000 0.2194
2 0.500 0.7141 red = derived value (protected)
3 1.000 1.2885
4 1.500 1.9132 blue = entered value (unprotected)
5 2.000 2.5910
6 2.500 3.2866 black = fixed text (protected)
7 3.000 4.1078
8 3.500 4.9446
9 4.000 5.8145
10 4.500 6.7230
11 5.000 7.7284
12 5.500 8.7566
13 6.000 9.8013
14 6.500 10.8818
15 7.000 12.0982
16 7.500 13.3122
17 8.000 14.5840
18 8.500 15.9238
19 9.000 17.3271

125
 EPA Traceability Protocol for Gaseous Calibration Standards

STEP 2 Review the Parameter Estimates

Review the estimates of the coefficients (b0, (b1,...) for the linear and quadratic
models, their confidence, and the residual error variances (s2).

Linear Model Expanded Uncertainty

Parameter Estimate Lower Upper
b0 = -1.0778 -1.7351 -0.4204
b1 = 1.9005 1.7757 2.0253
s2 = 0.4986 0.2807 1.1205
s= 0.7061 0.5298 1.0585
df = 17
t= 2.1098

Quadratic Model Expanded Uncertainty

Parameter Estimate Lower Upper
b0 = 0.1964 0.1960 0.1968
b1 = 1.0011 1.0010 1.0012
b2 = 0.0999 0.0999 0.0999
s2 = 0.0005 0.0003 0.0011
s= 0.0220 0.0164 0.0335
df = 16
t= 2.1199

The approximate level of confidence associated with the expanded uncertainties

is 95%.

The corresponding intervals take the form estimate +/- t*u, where: estimate is the
quantity of interest, u is its uncertainty, and t is a coverage factor.

For each model, the coverage factor, t, is determined from the t-distribution with
appropriate degrees of freedom (df).

Note that the uncertainties are not displayed, but can easily be derived from the
estimate, expanded uncertainty, and the coverage factor.

Comparing the two models:

F ratio = 1026.76
F critical = 2.3167 (5% significance level)

The quadratic model produces a significantly smaller error variance. The

quadratic model appears to be the better choice.

If cubic or quartic models are supported by compelling scientific theory or data,

then review the following estimates for those models. Otherwise, go to Step 3.

126
 EPA Traceability Protocol for Gaseous Calibration Standards

Cubic Model Expanded Uncertainty

Parameter Estimate Lower Upper
b0 = 0.1952 0.1593 0.2310
b1 = 1.0030 0.9676 1.0385
b2 = 0.0994 0.0901 0.1087
b3 = 0.0000 -0.0006 0.0007
s2 = 0.0005 0.0003 0.0012
s = 0.0227 0.0168 0.0352
df = 15
t= 2.1315

See above notes on calculation of expanded uncertainties.

Comparing quadratic and cubic models:

F ratio = 0.9385
F critical = 2.3849 (5% significance level)

The error variances are not significantly different at the 5% level. The quadratic
model appears to be a better choice than cubic.

Quartic Model Expanded Uncertainty

Parameter Estimate Lower Upper
b0 = 0.2206 0.1900 0.2512
b1 = 0.9285 0.8786 0.9783
b2 = 0.1390 0.1156 0.1625
b3 = -0.0069 -0.0109 -0.0030
b4 = 0.0004 0.0002 0.0006
s2 = 0.0003 0.0001 0.0007
s= 0.0165 0.0121 0.0261
df = 14
t= 2.1448

Comparing quadratic and cubic models:

F ratio = 1.8954
F critical = 2.4630 (5% significance level)

The error variances are not significantly different at the 5% level. The cubic
model appears to be a better choice than quartic.

STEP 3 Review the Charts

View the charts named Curves 1 and Residuals 1. Curves 1 shows bands
illustrating expanded uncertainties for the estimated regression. Compare these
bands with those of the quadratic regression, Curves 2. (Note: You can change

127
 EPA Traceability Protocol for Gaseous Calibration Standards

the associated approximate confidence level by changing the "p-value" in cell F4

of the worksheet named Chart Data.) Residuals 1 shows how the calibration
points deviated from the calibration line. Look for a simple pattern (such as a
quadratic curve) in the chart. If such a pattern appears, the quadratic model may
be better. View Residual 2, the deviations from the best-fit quadratic curve. If
Residual 2 effectively removes the simple pattern observed in Residual 1 and if
the magnitude of the deviations has been significantly reduced (as evidenced by
a reduction in the uncertainty u2), then the quadratic model is superior. An F-test
can be run to determine if the two overall uncertainties are significantly different.

F =1026.764 Prob. of greater F = 4.51E-21

The quadratic model produces a significantly smaller error variance. The

quadratic model appears to be the better choice.

If cubic or quartic models are supported by compelling scientific theory or data,

then view their Curves and Residuals charts.

STEP 4 Assess Magnitude of Uncertainty

Enter the concentration at which you would like to evaluate the uncertainty of
estimation and prediction. Also enter r, the number of assays to be performed.
Increasing r reduces the prediction uncertainty, but with diminishing effect.

Concentration = 6
r= 3

Review the estimated mean response (estimate that only takes into account the
calibration uncertainty), and the confidence intervals. Review the predicted
mean response and its confidence intervals. To satisfy the EPA protocol
requirements, the 95% confidence limits for the concentration should be 1% of
the concentration.

Estimates below are based on the quadratic model.

Tab-Right to view estimates based on the other model.

Expanded Uncertainty
Estimate Lower Upper
Instrument Response = 9.8006 9.7858 9.8155
Expanded Uncertainty
Prediction Lower Upper
Instrument Response = 9.8006 9.7698 9.8314
Concentration = 6.0000 5.9860 6.0140
95% uncertainty in prediction = 0.23%

128
 EPA Traceability Protocol for Gaseous Calibration Standards

STEP 4a Check Uncertainty Associated with Dilution

This step checks the accuracy of the apparatus to verify that the error associated
with dilution is not excessive. If no certified standards are diluted in performing
the multipoint calibration, skip to Step 5.

Check Standard Response

6.000 9.800
6.000 9.850
6.000 9.900

mean = 9.850
Cal. Response = 9.8006332
RD = 0.50%
Relative difference is ok

STEP 5 Assay Candidate Standard on Same Day

Proceed with the analysis of candidate standards if their 95% uncertainties, as

estimated above, are <1%. Enter the responses for up to 10 repeated analyses
of an individual candidate standard in the spaces provided below.

Note: This step applies only to candidate standards that are assayed on the
same day as the calibration.

Analysis Estimated
Number Response Concentration
1 4.500 3.2466
2 4.501 3.2473
3 4.499 3.2460

NOTE: For Cubic and Quartic Model estimates, view the Calculations in the
spreadsheet’s shaded regions

mean = 3.2466
standard deviation = 0.0006
df = 2
F= 0.0008
F sig? = FALSE (The sample variance is acceptable.)
Pr(>F) = 0.9992
95% Uncertainty = 0.55% (This is the relative expanded uncertainty.)
23.97% = = Portion of Variance Due to Calibration

Estimated Concentration of Candidate Standard 3.2466

129
 EPA Traceability Protocol for Gaseous Calibration Standards

Portion of Variance Due to

Rel. Expanded Uncertainty Calibration Uncertainty
0.55% 23.97%

STEP 6 Assay Candidate Standard on Different Day from Initial Calibration

This step applies to candidate standards that are assayed on a different day than
the initial calibration. Before candidate standards are run, the measurement
system is challenged with zero and span checks. Three or more discrete checks
of the zero gas and three or more checks of the span gas are made. Enter the
results below:

Response to Response
Zero gas Span conc. to Span
0.000 9.000 16.010
0.001 9.000 16.000
-0.001 9.000 15.990
n= 3 n= 3
mean = 0.000 mean = 16.000
s= 0.001 s= 0.010
Cal. 0.196 Cal. Response = 17.301
Resp.=

s = experimental standard deviation

Zero Gas Results Span Gas Results

Std. Error = s/sqrt(n) = 0.0006 0.0058
Rrs/100= 0.1600 0.1600
Std. Error is okay Std. Error is okay
Relative Difference (RD) 1.14% -7.52%
= RD is okay RD is excessive

RD calculations above are based on linear/quadratic model. Tab right to see

results for cubic/quartic. Following successful completion of the zero and span
checks, the candidate standard is measured together with reference standards.
While the candidate standard is normally interspersed with the reference
standards, the analysis conducted in this sheet requires that the results be
entered separately. There are two ways to do this. One way is to enter an
analysis set (one candidate standard response and the responses from its zero
and nonzero standard analyses) below. Another approach is to enter all of the
responses (multiple sets) below. Enter zero and reference standard responses
in the area to the left and enter the responses to a single candidate standard in
the area to the right, below.

130
 EPA Traceability Protocol for Gaseous Calibration Standards

STEP 6 (Continued)

See table in Step 6 of the Instructions for the minimum number of different
nonzero reference standard concentrations required.

Estimates below are based on the quadratic model.

Tab-Right to view estimates based on the other models.

Reference Standards (Enter 0 for

Zero Concentraton) Candidate Standard
Observed Estimated
Conc. Response Conc.2 Conc.3 Conc.4 Response Conc.
0.000 0.218 0.000 0.000 0.000 4.010 2.9400
0.000 0.219 0.000 0.000 0.000 4.000 2.9337
0.000 1.220 0.000 0.000 0.000 3.990 2.9274
4.500 6.693 20.250 91.125 410.063
4.500 6.723 20.250 91.125 410.063 nnn = 3
4.500 6.773 20.250 91.125 410.063 mean = 4 2.9337
9.000 17.317 81.000 729.000 6561.000 Exp. Stdev = 0.0063
9.000 17.327 81.000 729.000 6561.000 std error = 0.12%
9.000 17.337 81.000 729.000 6561.000 df = 2
F = 0.077954
F sig? = FALSE
Pr{>F} 0.925262
The relative standard
uncertainty is okay.
nn= 9

Coefficients are not significantly different.

Consider including the new data as part of original calibration (Step 1).

Estimated Concentration of Candidate Standard

2.9337

Rel. Expanded Uncertainty Portion of uncertainty2 due to calibration

uncertainty
0.76% 45.66%

Expanded Uncertainty for Candidate Standard Concentration

Lower Upper
2.9114 2.9560

These upper and lower limits are compared with the corresponding limits
estimated on different assay dates to establish that the candidate standard has
not drifted.

131
EPA Traceability Protocol for Gaseous Calibration Standards

132
 EPA Traceability Protocol for Gaseous Calibration Standards

Appendix B - Statistical Spreadsheet for

Procedure P3
1. ReadMe Spreadsheet

What this Workbook Is All About

This workbook enables the user to estimate the rate at which the weight of a
permeation tube decreases. A linear relationship between the tube's weight and
elapsed time is established. If the estimated weight at time zero is significantly
different from the actual weight at time zero, then at least the earliest data pair
should be removed from the analysis. Uncertainty of the slope estimate (the rate
of weight loss or drift) will be determined. The traceability protocol requires that
this estimate have a relative uncertainty of less than 1%.

How the Workbook Is Organized

The workbook consists of several worksheets, which are displayed as tabs on

the bottom of the screen. Each worksheet performs a distinct function as
described below.

ReadMe describes the workbook and explains how to use the

worksheets
Data allows for user input of the calibration data (elapsed
time and weight)
ANOVA performs analysis of variance and determines
whether the intercept term is needed
Calibration Results calculates the drift and its uncertainty
Curve graphically displays the drift line together with its
confidence bands
Residual graphically displays the vertical difference between
the observed and estimated weights
Report summarizes the assay results for a permeation device
Chart Data includes the data used to create the curve and
residual charts.

How the Worksheets Are Set up

Each worksheet contains instructions that guide the user through the steps in
usingthe worksheet. The worksheets are also color-coded to simplify use.
Shaded cells that are bordered in blue lines are cells whose contents you can
change (i.e., enter data). In other sheets you can change the following variables:

133
 EPA Traceability Protocol for Gaseous Calibration Standards

Sheet Variable Location Current Value

Data Unit of Time H22 m
Data Unit of Weight H24 g
Report Device ID F5 test data
Chart Data significance level D2 1.00E-05

Derived values are colored red. These cells contain formulas that should not be
changed. The cells are protected to prevent alteration.

How to Use the Worksheets

Step 1: Enter the elapsed times (all in the same units) and corresponding
tube weights in the Data worksheet. The worksheet will compute
the total weight loss for each observation.

Step 2: Select the significance level (alpha) to be used in producing

confidence limits for the estimated slope and intercept. Then
review the results of the F-test and t-test to determine whether the
intercept term is needed. If the intercept term is significant, then
determine which of the early data points should be removed.
Removing those data, and correcting the elapsed times, repeat
Steps 1 and 2.

Step 3: Examine the corresponding Curve chart. You may need to adjust
the chart's axis scaling. The points should appear to fall virtually on
top of the black line. The black line should be very close to its
confidence bands (colored red and blue).

Step 4: Examine the corresponding Residual worksheet. The residuals

should appear to be random in both magnitude and direction. If
they appear to follow a regular pattern, then the simple linear model
is not appropriate. The device does not have a constant rate of
weight loss. More time may be required to establish and measure
a linear relationship. Observations taken before the linear
relationship is established should be discarded and not used in the
statistical analysis.

Step 5: Print the one-page report provided in the Report sheet. The report
summarizes the assay data and indicates the uncertainty of the
estimate.

134
 EPA Traceability Protocol for Gaseous Calibration Standards

2. Data Entry Worksheet

Elapsed
Time Weight Enter the data in the blue-bordered spaces. The
first column (X) is for the elapsed time. The time of
I Xi Yi the first entry should be zero. The second column
(Y) is for the tube weights.
0 0 4.354206
1 8641 4.33745 n = number of weighing. This can’t exceed 50.
2 18722 4.316766 n=7
3 40322 4.273494
4 64802 4.224514
5 74882 4.20378
6 84962 4.18439 No data entry is required for derived values, which
are colored red, such as n and I. These values are
tabulated automatically and their cells are protected
from alteration.

Multiple weighings at a single point in time requires

multiple entries in each column. Reenter the time in
column X and enter the corresponding weight in
column Y.
Enter the time and weight units in the spaces
below:
Unit of Time = m
Unit of Weight = g

3. Regression Worksheet

WORKSHEET MODEL: LINEAR WITH NON-ZERO INTERCEPT

This sheet derives the regression equation in the form: y = b0 + b1 x + . The

intercept and slope are estimated. The sheet determines whether the intercept
(weight estimated for time zero) is significantly different from the observed weight
at time zero. It also estimates the uncertainty in the slope estimate and
compares this uncertainty with EPA's 1% limit.

STEP 1

Review the estimates of the intercept (b0), slope of the regression line (b1), and
their confidence limits along with the estimates of variance-covariance matrix (V)
and the residual error variance (Var).

Derivation of the estimated intercept (b0) and slope (b1) of the regression line

135
 EPA Traceability Protocol for Gaseous Calibration Standards

X'X = 7 292331 Y'Y = 127.6972 95% Confidence Limits

292331 1.91E+10 df = 5 b0 lower 4.353872
limit =
(X'X)-1 = 0.396793 -6.1E-06 t(0.95, df) 2.570578 b0 upper 4.354935
= limit =
-6.1E-06 1.46E-10
det(X'X) = 4.81E+10 b0 = 4.354403 b1 lower -2E-06
limit =
X'Y = 29.8946 b1 = -2E-06 b1 lower -2E-06
limit =
1234673

Derivation of the error variance (Var) and variance-covariance matrix (V)

b'X'Y = 127.6972 SS(model), 2df

b'X'Y - sum(Y)2 / n = 0.027619 SS(regression) 1df
(Y'Y - b'X'Y) = 5.38E-07 SS(residual)
Var = (Y'Y - b'X'Y) / df = 1.08E-07 MS(residual), n-2 df
V = Var * (X'X)-1 = 4.27E-08 -6.5E-13
-6.5E-13 1.57E-17

STEP 2

Examine the upper and lower limits for the intercept, b0.

b0 lower limit = 4.353872

b0 upper limit = 4.354935
y0 = 4.354206 y0 is within the confidence limits for the intercept

If y1 is within the confidence limits for the intercept, proceed to STEP 3.

Otherwise, consider removing the first observed weight from the analysis. Re-
enter the times and weights. Remember that the first time (X0) should be zero.
This will require adjustment of the other elapsed times. After entering the data,
return to STEP 1, above.

STEP 3

Examine the upper and lower limits for the slope, b1. The limits should differ from
the estimate by no more than 1% of the estimated slope.

(b1 upper - b1) / |b1| = 0.51% Conclusion: Uncertainty is acceptable.

(b1 lower - b1) / |b1| = -0.51%

If the uncertainty is unacceptable, consider collecting additional data. Also, view

the Curve and Residual plots. They may reveal a nonlinear relationship for a

136
 EPA Traceability Protocol for Gaseous Calibration Standards

portion of the data. The initial measurements may not align with subsequent
measurements if the device was in the process of stabilizing or equilibrating
during those times. If this is the case, the initial points of the Residual chart
would appear to be outliers. The residuals with the same sign (all positive or all
negative) and their magnitude will likely be greater than the magnitude of
subsequent residuals. If this is the case, consider removing the initial points from
the computations and re-enter the remaining times and weights with the times
adjusted so the first entry has time zero.

If the uncertainty is acceptable, print the Report spreadsheet and include it with
the certification documentation.

4. Report Worksheet

Assay Results for Permeation Device

This sheet provides calibration information and assay results, including

uncertainty estimates for a permeation device identified as: test data

Reference: Appendix B, EPA Traceability Protocol Assay and Certification of

Gaseous Calibration Standards, USEPA, 1997.

Test Results

Intercept (b0), slope (b1), and their confidence limits

X'X = 7 292331 Y'Y = 127.6972 95% Confidence Limits

292331 1.91E+10 df = 5 b0 lower 4.353872
limit =
(X'X)-1 = 0.396793 -6.1E-06 t(0.95, df) 2.570578 b0 upper 4.354935
= limit =
-6.1E-06 1.46E-10
det(X'X) = 4.81E+10 b0 = 4.354403 b1 lower -2E-06
limit =
X'Y = 29.8946 b1 = -2E-06 b1 lower -2E-06
limit =
1234673

Error variance (Var) and variance-covariance matrix (V).

b'X'Y = 127.6972 SS(model), 2df

b'X'Y - sum(Y)2 / n = 0.027619 SS(regression) 1df
(Y'Y - b'X'Y) = 5.38E-07 SS(residual)
Var = (Y'Y - b'X'Y) / df = 1.08E-07 MS(residual), n-2 df
V = Var * (X'X)-1 = 4.27E-08 -6.5451E-13
-6.5E-13 1.56725E-17

137
 EPA Traceability Protocol for Gaseous Calibration Standards

Upper and lower limits for the intercept, b0:

b0 lower limit = 4.3538722

b0 upper limit = 4.3549347
y0 = 4.354206
Upper and lower limits for the slope, b1:

(b1 upper - b1) / |b1| = 0.51%

(b1 lower - b1) / |b1| = -0.51%

Estimated rate of weight loss, b1 = 2.005E-06 g/m

138
 EPA Traceability Protocol for Gaseous Calibration Standards

Appendix C - Statistical Spreadsheet for Stability

Determination
ASSAY RESULTS

In this sheet the results of two or three assays are entered. Calibration dates are
entered so Assays having the same calibration uncertainty may be treated
correctly. (Assays having a common calibration share the same calibration
uncertainty.)

Enter the results for up to three separate assays in chronological order below.

ASSAY 1

500= estimated concentration

0.10%= Expanded uncertainty (as percentage of estimated concentration)
0.5= portion of expanded uncertainty2 due to calibration
5= number of measurements
5/1/1997= date of prior calibration

ASSAY 2

501= estimated concentration

0.10%= Expanded uncertainty (as percentage of estimated concentration)
0.5= portion of expanded uncertainty2 due to calibration
3= number of measurements
6/1/1997= date of prior calibration

ASSAY 3 (if applicable)

502= estimated concentration

0.10%= Expanded uncertainty (as percentage of estimated concentration)
0.5= portion of expanded uncertainty2 due to calibration
3= number of measurements
6/8/1997= date of prior calibration

Number of different calibrations represented by the above data:

N= 3 (If this value seems to be incorrect, check the dates
entered for the three assays. The earliest data should
be for Assay 1. Trailing spaces may cause N's formula
to interpret identical dates as different.)

139
 EPA Traceability Protocol for Gaseous Calibration Standards

COMPARISONS

Assay 1 vs Assay 2
Two One-Sided Equivalence
with Assay 1 Variance Components
Lower Upper Lower Test Upper Test Calibration Imprecision
Calibration Assay ConfidenceConfidence Signficance Signficance Total
1 1 ---------------------- ------------------------- 0.125 0.125 0.25
2 2 0.644989 1.355011 TRUE TRUE 0.125501 0.125501 0.251001

"FALSE" indicates an inconsistency where the observed confidence interval of

the difference is beyond the tolerance level.

Assay 1 vs Assay 3
Two One-Sided Equivalence
with Assay 1 Variance Components
Lower Upper Lower Test Upper Test Calibration Imprecision
Calibration Assay ConfidenceConfidence Signficance Signficance Total
1 1 ---------------------- ------------------------- 0.125 0.125 0.25
3 3 1.644752 2.355248 TRUE TRUE 0.126002 0.126002 0.252004

"FALSE" indicates an inconsistency where the observed confidence interval of

the difference is beyond the tolerance level. Nothing will appear here if no data
have been entered for Assay 3.

Assay 2 vs Assay 3
Two One-Sided Equivalence
with Assay 1 Variance Components
Lower Upper Lower Test Upper Test Calibration Imprecision
Calibration Assay ConfidenceConfidence Signficance Signficance Total
2 2 ---------------------- ------------------------- 0.125501 0.125501 0.251001
3 3 0.563533 1.436467 TRUE TRUE 0.126002 0.126002 0.252004

"FALSE" indicates an inconsistency where the observed confidence interval of

the difference is beyond the tolerance level. Nothing will appear here if no data
have been entered for Assay 3.

OVERALL ESTIMATE

Note: Calibration Case = 18 (right click to see comment)

Case Cal. Cal. Cal. No.

No. No.
4* = 1 1 ---
6* = 1 2 ---
9 = 1 1 1
12 = 1 1 2
15 = 1 2 2
18 = 1 2 3

140
 EPA Traceability Protocol for Gaseous Calibration Standards

*4 and 6 are cases where there is no 3rd assay. In case 4, the two assays share
a common calibration. In case 6, the two assays have different calibrations.

The standard error of the estimate produced in an assay is equal to

approximately ½ of the "95% uncertainty." The inverse of the square of the
standard error is the (raw) weighting factor used in producing an overall estimate
of the concentration. The raw weights are adjusted (Adj. Wt.) so their sum is
1.00.

Variance
Expanded Wt. of Wt.
Calibration Estimate Uncert. Raw Wt. Adj. Wt. *Conc. *Est.
1 500 0.100% 100000 0.333 166.667 0.02778
2 500 0.100% 100000 0.333 167.00 0.02789
3 502 0.100% 100000 0.333 167.333 0.02800

502 = overall estimate of the candidate standard's concentration

0.2892529 = expanded uncertainty (concentration units)
0.058% = expanded relative uncertainty

The standard error and expanded uncertainty displayed above do not account for
uncertainty in the reference standards used to calibrate the analytical instrument,
or for the uncertainty due to the dilution flow rate. In the first space below, enter
the expanded uncertainty (typically 2 times the standard error) of the reference
standards. If different calibration standards had different uncertainties, enter the
largest.

Example: If NIST SRMs were used in the calibration and their certified
concentrations were 100 +/- 1 ppm, 200 +/- 1 ppm, 300 +/- 2 ppm, 400 +/- 3 ppm
and 500 +/- 4 ppm, then the largest 95% uncertainty is for the 100 ppm standard:
1/100 = 0.01 or 1%.

(SRM uncertainties are expressed as two-sigma limits which are 95% confidence
intervals.)

1.00%= Expanded uncertainty (the two-sigma uncertainty) of the

reference standard
0.00%= Expanded uncertainty (the two-sigma uncertainty) of the dilution
system flow rate (This only applies when the gas settings changed
between the assy of the reference standard and the assay of the
candidate standard are changed, otherwise it should be 0).
0.00%= Expanded uncertainty (the two-sigma uncertainty) for
interference corrections when measuring multiple gases
simultaneously. This must be provided by a Statistician.
1.002% =Expanded uncertainty of the candidate standard (including the
contribution of the reference standard)

141
 EPA Traceability Protocol for Gaseous Calibration Standards

Appendix D - Matrix Notation

1. Matrix Notation

Matrix notation is used to simplify the presentation of calculations that are

performed in the linear regression. A matrix is a rectangular array of numbers.
Boldface capital letters represent matrices, and lower case letters with subscripts
represent individual numbers in the matrices. X, below, is a 10 by 3 matrix. It
has 11 rows and 3 columns. The rows are numbered 0, 1, 2,...10 and columns
are numbered 0, 1, and 2. (Other texts may begin numbering with 1.)

1 1.002 1.0040
1 0.902 0.8136
1 0.802 0.6432
1 0.701 0.4914
X= 1 0.601 0.3612
1 0.501 0.2510
1 0.401 0.1608
1 0.301 0.0906
1 0.200 0.0400
1 0.100 0.0100
1 0.000 0.0000

Xi,j denotes the number that is found in the ith row and the jth column. X0,1 =
1.002. The first row and column are numbered zero.

A matrix that has only one column is called a column vector, and a matrix that
has only one row is called a row vector.

0.999
0.915
0.828
0.738
y= 0.644 is a column vector.
0.549
0.448
0.346
0.237
0.122
0.001

Subscripts following vector names denote the row or column of the vector. For
example, y1 is the number in the second row of y, 0.915. (Remember that we
begin counting rows with zero.)

Matrix operations that come into play for calibration include multiplication,

142
 EPA Traceability Protocol for Gaseous Calibration Standards

transposition, and inversion. The rules for these operations can be found in any
introduction to matrices. We will use the following notation for these operations:

X' denotes the transpose of X (the ith column of X becomes the ith row of X')

For the matrices X and Y above,

1 1 1 1 ... 1
X' = 1.002 0.902 0.802 0.701 ... 0.000
1.0040 0.8136 0.6432 0.4914 ... 0.0000

X'Y denotes multiplication of matrices X' and Y. X' must have the same number
of columns as Y has rows. For the matrix X above,

11 5.511 3.8658 5.827

X'X = 5.511 3.8658 3.0438 and X'Y = 4.0132
3.8658 3.0438 2.5544 3.1272

det(X'X) = 1.0521 (the determinant of X'X)

(X'X)-1 denotes the inverse of the product of X' and X

0.5800 -2.1962 1.7392

-1
(X'X) = -2.196 12.5026 -11.5744
1.7392 -11.5744 11.5513

2. Calibration by Linear Regression Using Matrix Notation - Example

The linear regression approach is illustrated below for the simple quadratic curve.

The starting point for regression analysis will be a matrix named X. This matrix
will have 3 columns (one for each coefficient to be determined). The number of
rows will be the same as the number of calibration measurements that are
performed by the measurement system. The first column is a vector of 1s. The
second column contains the certified concentrations of the calibration standards.
The third column contains the squares of the values appearing in the second
column. When this matrix is multiplied by the vector of coefficients [b0, b1, b2],
the result is a vector of responses, so that:

responsei = 1 * b0 + concentrationi * b1 + concentrationi2 * b2

or, letting y represent response and x represent concentration,

yi = b0 + b1 xi + b2 xi2

143
 EPA Traceability Protocol for Gaseous Calibration Standards

Now, we're interested in estimating the the coefficients b0, b1, and b2, and we're
also interested in computing how much error is involved when we use the
information to estimate the concentration in an "unknown."

3. Determining the Calibration Equation

The coefficients of the calibration equation or curve are found by matrix

multiplication and inversion:

b = (X'X)-1 X'Y = [b0, b1, b2]

Example

1 1.002 1.0040 0.999 0.9967

1 0.902 0.8136 0.915 0.9151
1 0.802 0.6432 0.828 0.8297
1 0.701 0.4914 0.738 0.7394
X =1 0.601 0.3612 y= 0.644 b'x = 0.6462
1 0.501 0.2510 0.549 0.5491
1 0.401 0.1608 0.448 0.4482
1 0.301 0.0906 0.346 0.3434
1 0.200 0.0400 0.237 0.2336
1 0.100 0.0100 0.122 0.1210
1 0.000 0.0000 0.001 0.0046

0.0046
b = 1.1837 b' = 0.0046 1.1837 -0.1932
-0.1932

The quadratic calibration curve is: response = 0.0046 + 1.1837 C + -0.1932 * C2

4. Determining the Estimation and Prediction Error

One assumption that underlies the regression approach is that random error is
constant across the measurement range. Sometimes it may be necessary to
apply a transformation in order to achieve this characteristic, called homogeneity
of variance. An estimate of this variance is obtained using matrix operations:

Var = residual sum of squares / degrees of freedom = Y'Y - b'X'Y / df

This estimate's "degrees of freedom" (df) is the number of calibration points less
the number of coefficients estimated for the calibration equation.

144
 EPA Traceability Protocol for Gaseous Calibration Standards

An important output of the regression analysis is the "variance-covariance"

matrix, V:

V = Var * (X'X)-1

The variance of each coefficient is found in the principal diagonal of V. For

example, the variance of b0 is V0,0. Covariances are found as off-diagonal
elements of V.

Hypothesis tests can be performed and confidence intervals can be estimated for
each coefficient using the coefficient's estimate, the coefficient's variance
(contained in V), and the degrees of freedom, df.

Continuing our example

Var = (Y'Y - b'X'Y) / df = 5.91E-06

3.43E-06 -1.3E-05 1.03E-05

V= -1.3E-05 7.39E-05 -6.8E-05 df = 8
1.03E-05 -6.8E-05 6.83E-05 (df = degrees of
freedom)

95% Confidence Interval for b0= b0 +/- t(0.05,df) * sqrt(V0,0)

95% CI for b0 = 0.000324 to 0.008865

t(0.05, df) = 2.306006

95% Confidence Interval for b1 = b1 +/- t(0.05,df) * sqrt(V1,1)

95% CI for b1 = 1.163855 to 1.203512

95% Confidence Interval for b2= b2 +/- t(0.975,df) * sqrt(V2,2)

95% CI for b2 = -0.21224 to -0.17413

Another use of V is in computing the uncertainty in a regression predicted

concentration of an individual unknown. The analyzer is subjected to the
unknown, and a mean response, R, is produced. A solution for C is found. This
is the estimated concentration of the unknown.

Deriving the confidence intervals for this estimate requires finding two alternative
concentrations, one higher and one lower than the estimate, such that the
probability of having produced a lesser or greater average response is sufficiently
small. For a 95% confidence interval, the lower bound is a concentration whose
response would be less than the observed response with 97.5% probability; the

145
 EPA Traceability Protocol for Gaseous Calibration Standards

upper bound is a concentration whose response would be less than the observed
response with 97.5% probability.

Unfortunately, for quadratic curves, this derivation is not so simple.

R measurements of an unknown produce an average response resp:

R= 6
resp = 0.601

The estimated concentration is found by solving the following quadratic equation:

0.601 = b0 + b1 C + b2 C2
(b0 - 0.601) + b1 C + b2 C2 = 0

The potential solutions are found using the quadratic formula:

C = 0.553935 and 5.573267 (only the first of these is reasonable)

Now, if the concentration really had been at this value, the 95% confidence
interval for the mean response of six measurements would be symmetric about
the observed response:

As above, t = 2.306006
x = 1 0.553935 0.306843 = [1, resp, resp2]
xb = 0.601 (check)

var(predicted mean response for x) = [var/R + x' V x]

x'V = -6.09E-07 6.97E-06 -6.7E-06

x'Vx = 1.2E-06
var/6 = 9.86E-07

var(predicted mean response for x) = 2.19E-06

95% confidence interval for predicted response = 0.597588 to 0.604412

This is the observed response -/+: 0.003412 and 0.003412

Solving for concentration, the interval is no longer perfectly symmetric:

0.550418 to 0.557456

This is the estimated concentration -/+: 0.003516 and 0.003521

As a percentage of the concentration, this is -/+: 0.006348 and 0.0063569

146
 EPA Traceability Protocol for Gaseous Calibration Standards

Fortunately, even with the quadratic calibration curve, with good precision, the
confidence intervals will be within a small enough region that the curve is close to
linear and the interval will be very nearly symmetric. The uncertainty criterion for
multipoint calibration requires the 95% confidence interval's half-width to be less
than 1%. The calibrated range of the analyzer extends across all concentrations
for which the criterion is satisfied.

Continuing our example

Estimated 95% conf. interval 95% conf. interval % error for

Concentration Response for response for concentration concentration
1.002 0.9967 0.9924 1.0010 0.9966 1.0074 -0.53 0.54
0.902 0.9151 0.9121 0.9181 0.8985 0.9055 -0.39 0.39
0.802 0.8297 0.8273 0.8320 0.7993 0.8047 -0.33 0.33
0.701 0.7394 0.7371 0.7417 0.6985 0.7035 -0.36 0.36
0.601 0.6462 0.6437 0.6487 0.5984 0.6036 -0.43 0.43
0.501 0.5491 0.5466 0.5517 0.4984 0.5036 -0.51 0.52
0.401 0.4482 0.4457 0.4507 0.3986 0.4034 -0.60 0.60
0.301 0.3434 0.3411 0.3457 0.2988 0.3032 -0.72 0.72
0.200 0.2336 0.2313 0.2359 0.1979 0.2021 -1.06 1.06
0.100 0.1210 0.1181 0.1240 0.0974 0.1026 -2.58 2.59
0.000 0.0046 0.0003 0.0089 -0.0036 0.0036 --- ---
0.210 0.2446 0.2423 0.2469 0.2079 0.2121 -0.9996 1.0004

The calibration curve's uncertainty is acceptable for concentrations > 0.21 ppm.

5. Stability Test using Schuirmann’s two one-sided tests (TOST)

As discussed in Subsection 2.1.5.2, the stability test requires at least three initial
measurements of the candidate standard plus at least three additional
measurements following a period of 7 days or more. The standard's
concentration must be in the well-characterized region of the analyzer per
Subsection 2.1.4.2.

Concentrations are estimated using the calibration curve from at least three
measurements for the initial assay and at least three measurements for the
second assay. Schuirmann’s two one-sided tests (TOST) will be used to test for
equivalence with a acceptance criterion (i.e.,tolerance) of 1.0%. The calculation
follows:

Initial Assay Second Assay (after holding time)

C1 C4
C2 C5
C3 ` C6

147
 EPA Traceability Protocol for Gaseous Calibration Standards

n1 = number of observations in initial assay

n2 = number of observations in second assay
x1 = (C1 + C2 + C3) / 3
x2 = (C4 + C5 + C6) / 3
s1 = standard deviation of (C1, C2, C3)
s2 = standard deviation of (C4, C5, C6)
tol = acceptable level of difference between the means = 0.01 (i.e., 1.0 percent)
alpha = significance level of the test = 0.05
t(1-alpha,df) = value of Student's t for which the distribution function value is 0.95
and degrees of freedom = n1+ n2 - 2

The pooled variance of the six measurements is calculated from

sp2 = [(n1-1) s12+( n2-1)]s22/[( n1-1)+( n2-1)]

First calculate the confidence interval on the difference of means as

x1- x2 +/- t(1-alpha,df)* sp*sqrt(1/ n1 + 1/ n2)

Note that we are using alpha, and not the usual alpha/2, in the t-value.

If this confidence limit is within the TOST acceptance region, which is calculated
as

x1- x2 +/-(x1+ x2)/2*(1+tol)

then the null hypothesis is rejected and the two sets of results are considered
equivalent.

If the confidence limit is not within the TOST acceptance criterion, the stability
test can be repeated after an additional 7 days or more using the second and
third sets of results in the calculations, as above. If another statistical failure is
found, then the candidate standard is determined to be instable and it is
disqualified for further use.

Example:

Initial Assay Final Assay (after 7-day holding time)

0.995 ppm 0.993 ppm
0.996 ppm 0.997 ppm
0.992 ppm 0.991 ppm

n1 = 3 n2 = 3
s1 = 0.0020817 ppm x1 = 0.9943333 ppm
s2 = 0.0030551 ppm x2 = 0.9936667 ppm
sp2 = 0.002614065 tol = 1.0%

148
 EPA Traceability Protocol for Gaseous Calibration Standards

x1 - x2 = 0.000666667
t(1-alpha,df) = 2.131847
CI lower limit = -0.00388
CI upper limit = 0.00522
TOST lower limit = -0.00873
TOST upper limit = .0.01061

The confidence interval fits in the TOST acceptance region, so the two sets of
measurements are equivalent and the candidate standard can be certified.

6. Recertification

Per Subsection 2.1.11, a standard can be recertified after the certification period
has elapsed, if the standard has been reassayed and the Appendix C
spreadsheet indicates that the TOST acceptance criterion has been attained. To
determine whether the concentration of the standard has changed since the
initial certification, new measurements are made using a pollutant gas analyzer
that has been calibrated according to Subsection 2.1.4.2. Original certification
data are used to provide initial estimates of mean (x1) and standard deviation
(s1). The recertification data are used to estimate a second mean (x2) and
second standard deviation (s2). These are used in Schuirmann’s two one-sided
test (TOST) to determine if the measurements are equivalent. As outlined in the
previous section, for a 95% confidence level a critical value for t is based on a
90% confidence interval on the difference of the means with degrees of freedom
equal to the number of initial and recertification data minus 2. A pooled estimate
of the standard deviation (sp) is derived from s1 and s2. If the confidence interval
is within the TOST acceptance region, then the two sets of measurements are
equivalent and the standard is stable. If not, the standard cannot be recertified.

Example:

Original Data Recertification Data

0.995 ppm 0.989 ppm
0.996 ppm 0.990 ppm
0.992 ppm 0.994 ppm
0.999 ppm
0.999 ppm
0.993 ppm

n1 = 6 n2 = 3
s1 = 0.0029439 ppm x1 = 0.9956667 ppm
s2 = 0.0026458 ppm x2 = 0.991 ppm
sp = 0.002394438 tol = 1%
ppm

149
 EPA Traceability Protocol for Gaseous Calibration Standards

x1 - x2 = 0.00466667
t(1-alpha/2,df) = 1.894579
CI lower limit = 0.00145891
CI upper limit = 0.00787442
TOST lower limit = -0.005266
TOST upper limit = 0.014600

The confidence interval fits within the TOST acceptance region, so the standard
may be recertified. The certified concentration of the standard is the grand mean
of the combined data set.

Certified Concentration = mean (initial data + recertification data) = 0.994 ppm

150
 EPA Traceability Protocol for Gaseous Calibration Standards

Appendix E - Statistical Spreadsheet for

Procedure G3
1. ReadMe Worksheet: This worksheet is the same as the one in Appendix A

2. Zero Test Worksheet

STEP 1 Enter Calibration Data

Enter the calibration data in the shaded spaces below. The first column (I)
simply counts the calibration points that you enter. The second column (X) is for
the certified concentrations of the calibration gas standards. The third column
(Y) is for the instrument responses corresponding to the calibration standards.
The number of points cannot exceed 50.

i Xi Yi Color Code
1 0.0000 0.1900
2 0.0000 0.2000 red = derived value (protected)
3 0.0000 0.2100
blue = entered value
4 1.0000 1.1400 (unprotected)
5 1.0000 1.1500
6 1.0000 1.1600 black = fixed text (protected)
7 2.0000 1.9900
8 2.0000 2.0000
9 2.0000 2.0100
10 3.0000 2.7400
11 3.0000 2.7500
12 3.0000 2.7600
13 4.0000 3.3900
14 4.0000 3.4000
15 4.0000 3.4100
16 5.0000 3.9400
17 5.0000 3.9500
18 5.0000 3.9600

STEP 2 Review the Parameter Estimates

Review the estimates of the coefficients (b0, (b1,...) for the linear and quadratic
models, their confidence, and the residual error variances (s2).

Linear Model Expanded Uncertainty

Parameter Estimate Lower Upper
b0 = 0.3667 0.2492 0.4841
b1 = 0.7500 0.7112 0.7888

151
 EPA Traceability Protocol for Gaseous Calibration Standards

s2 = 0.0176 0.0097 0.0407

s= 0.1326 0.0987 0.2018
df = 16
t= 2.1199
Quadratic Model Expanded Uncertainty
Parameter Estimate Lower Upper
b0 = 0.2000 0.2000 0.2000
b1 = 1.0000 1.0000 1.0000
b2 = -0.0500 -0.0500 -0.0500
s2 = 0.0001 0.0000 0.0002
s= 0.0089 0.0066 0.0138
df = 15
t= 2.1314

The approximate level of confidence associated with the expanded uncertainties

is 95%.
The corresponding intervals take the form estimate +/- t*u, where: estimate is the
quantity of interest, u is its uncertainty, and t is a coverage factor.

For each model, the coverage factor, t, is determined from the t-distribution with
appropriate degrees of freedom (df).

Note that the uncertainties are not displayed, but can easily be derived from the
estimate, expanded uncertainty, and the coverage factor.

Comparing the two models:

F ratio = 219.69
F critical = 2.3849 (5% significance level)

The quadratic model produces a significantly smaller error variance. The

quadratic model appears to be the better choice.

If cubic or quartic models are supported by compelling scientific theory or data,

then review the following estimates for those models. Otherwise, go to Step 3.

Cubic Model Expanded Uncertainty

Parameter Estimate Lower Upper
b0 = 0.2000 0.1891 0.2109
b1 = 1.0000 0.9789 1.0211
b2 = -0.0500 -0.0605 -0.0395
b3 = 0.0000 -0.0014 0.0014
s2 = 0.0001 0.0000 0.0002
s = 0.0093 0.0068 0.0146
df = 14
t= 2.1448

152
 EPA Traceability Protocol for Gaseous Calibration Standards

See above notes on calculation of expanded uncertainties.

Comparing quadratic and cubic models:

F ratio = 0.9286
F critical = 2.5536 (5% significance level)

The error variances are not significantly different at the 5% level. The quadratic
model appears to be a better choice than cubic.

Quartic Model Expanded Uncertainty

Parameter Estimate Lower Upper
b0 = 0.2000 0.1880 0.2120
b1 = 1.0000 0.9559 1.0441
b2 = -0.0500 -0.0921 -0.0079
b3 = 0.0000 -0.0133 0.0133
b4 = 0.0000 -0.0013 0.0013
s2 = 0.0001 0.0000 0.0002
s= 0.0096 0.0069 0.0155
df = 13
t= 2.1604

Comparing quadratic and cubic models:

F ratio = 0.9286
F critical = 2.5536 (5% significance level)

The error variances are not significantly different at the 5% level. The cubic
model appears to be a better choice than quartic.

STEP 3 Review the Charts

View the charts named Curves 1 and Residuals 1. Curves 1 shows bands
illustrating expanded uncertainties for the estimated regression. Compare these
bands with those of the quadratic regression, Curves 2. (Note: You can change
the associated approximate confidence level by changing the "p-value" in cell F4
of the worksheet named Chart Data.) Residuals 1 shows how the calibration
points deviated from the calibration line. Look for a simple pattern (such as a
quadratic curve) in the chart. If such a pattern appears, the quadratic model may
be better. View Residual 2, the deviations from the best- fit quadratic curve. If
Residual 2 effectively removes the simple pattern observed in Residual 1 and if
the magnitude of the deviations has been significantly reduced (as evidenced by
a reduction in the uncertainty u2), then the quadratic model is superior. An F-test
can be run to determine if the two overall uncertainties of fit are significantly
different.

153
 EPA Traceability Protocol for Gaseous Calibration Standards

F =219.6875 Prob. of greater F = 7.52E-15

The quadratic model produces a significantly smaller error variance. The

quadratic model appears to be the better choice.

If cubic or quartic models are supported by compelling scientific theory or data,

then view their Curves and Residuals charts.
STEP 4 Test for Zero Material on Same Day as Initial Calibration

This test can only be used when using the linear or quadratic model.

Enter the impurity specification for the zero air material estimation and prediction

Maximum Concentration = 0.1

Maximum Response that can be measured at 95% confidence to make sure

concentration is less than Maximum Concentration when Predicting Response.

0.2821

STEP 5 Test for Zero Material on Different Day from Initial Calibration

This step applies to zero material tests on a different day than the initial
calibration.
This test can only be used when using the linear or quadratic model. Before the
zero material limit is input, the measurement system is challenged with zero and
span checks. Three or more discrete checks of the zero gas and three or more
checks of each span gas are made. If necessary, second span should be higher
concentration than the first. Enter the results below:

2 different span values need to be tested in the quadratic model

(Only Quadratic Models)

Response to First Span Response Second Response
Zero gas Conc. to Span Span Conc. to Span
0.2 1.000 1.150 2.000 2.000
0.21 1.000 1.160 2.000 2.010
0.22 1.000 1.170 2.000 2.020

Max. Span Calculations

n= 3 n= 3
mean = 0.210 mean = 2.010
s= 0.010 s = 0.010
Cal. Resp. = 0.200 Cal. Resp. = 2.000
s = experimental standard deviation

154
 EPA Traceability Protocol for Gaseous Calibration Standards

Zero Gas Results Max. Span Gas

Results
Std. Uncertainty = 0.0058 0.0058
Rrs/100 = 0.0201 0.0201
Std. Uncertainty is Std. Uncertainty is
OK OK
Relative Difference -0.50% 0.50%
(RD)= RD is OK RD is OK

RD calculations above are based on linear/quadratic model.

Following successful completion of the zero and span checks, zero material limit
can be predicted by the appropriate model

Coefficients are not significantly different. Consider including the new data as
part of original calibration (Step 1).

Maximum Concentration = 0.1

Maximum Response that can be measured at 95% confidence to make sure

concentration is less than Maximum Concentration when Predicting.

0.2916

155
 EPA Traceability Protocol for Gaseous Calibration Standards

Appendix F - Statistical Spreadsheet for

Procedure D1
1. ReadMe Worksheet

INSTRUCTIONS FOR GAS DILUTION WORKBOOK

The general workflow for this workbook is from right to left. The steps involved
are:
To be done annually:
Zero Gas Annual Calibration
Pollutant Annual Calibration
Annual Gas Check
To be done monthly:
Monthly Flow Check
To be done on day of use:
Day of Use

Organization

The workbook consists of several worksheets, which are displayed as tabs at the
bottom of the screen. The functions of these worksheets are described below:

Read Me (this sheet) - describes the workbook, explaining how to use the
worksheets

Day of Use - to be used on day of GDS use to calculate GDS settings

Monthly Flow Check - to verify that GDS calibration has not shifted since
annual calibration

Annual Gas Check - to be performed after annual calibration to verify that the
GDS actually produces the concentrations of diluted mixtures for both
nonreactive and reactive gases that agree with theoretical concentrations

For both the Pollutant Annual Calibration and Zero Gas Annual Calibration:

Calibration - allows for user input of calibration and other analytical data

Curves 1 - displays the calibration data, the best-fit line, and expanded
uncertainties.

Residuals 1 - displays the difference between the observed responses

and those estimated by the best-fit calibration line.

156
 EPA Traceability Protocol for Gaseous Calibration Standards

Curves 2 - displays the calibration data, the best-fit quadratic curve, and
expanded uncertainties.

Residuals 2 - displays the difference between the observed responses

and those estimated by the best-fit quadratic regression line.

Chart Data - includes the data used to create the Curves and Residuals
charts

Pollutant and Zero Gas Annual Calibrations Background

This worksheet supports linear and quadratic regression models:

Linear: y = b0 + b1x+ e
Quadratic: y = b0 + b1x + b2x2 + e

The inputs are:

x the valve setting for the corresponding gas
y the measured flow rate
e a random effect due to variation of influence quantities

It is assumed that the valve setting standards (x) have negligible uncertainties.

The output quantities are estimated parameters (bs) and related uncertainties.
The workbook then helps the user perform the following functions:

- determine which model (linear or quadratic) is best.

Conventions

The Pollutant Gas and Zero Gas Annual Calibration worksheets includes
instructions that guide the user through the steps in its use. Each worksheet is
also color coded to simplify use. Shaded cells that are bordered in blue text are
for input of data. These cells are unprotected, but all other cells of the
Calibration worksheet are protected.

Derived values and statements are colored red. These cells contain formulas
and are protected to prevent alteration.

Spreadsheet background colors indicate the order of the polynomial supported

by the calculations in the area.
Light green is used for the linear model.
Yellow is used for the quadratic model.

Reference: EPA Traceability Protocol Assay and Certificatoin of Gaseous

Calibration Standards, USEPA, 2012

157
 EPA Traceability Protocol for Gaseous Calibration Standards

2. Pollutant Gas and Zero Gas Annual Calibration Worksheets

STEP 1 Enter Calibration Data

Enter the annual pollutant flow rate calibration data in the shaded spaced below.
The first column (i) simply counts the calibration points. The second column (X)
is for the pollutant flow controller settings. The third column (Y) is for the flow
rate reference standard readings corresponding to these settings. The number
of points cannot exceed 50.

i Xi Yi Color Code
1 78.0000 78.4529
2 70.0000 70.1984 red = derived value (protected)
3 60.0000 60.2776 blue = entered value (unprotected)
4 50.0000 50.1985 black = fixed text (protected)
5 40.0000 40.0262

STEP 2 Review the Parameter Estimates

Review the estimates of the parameters (b0, b1, …) for the linear and quadratic
models, their expanded uncertainties and the overall uncertainty of the fit (u2).

Linear Model Expanded Uncertainty

parameter estimate lower upper
b0 = -0.2914 -0.8714 0.2886
b1 = 1.0088 0.9993 1.0182
s2 = 0.0082 0.0026 0.1141
s= 0.0906 0.0513 0.3378
df = 3
t= 3.1824
Quadratic Model Expanded Uncertainty
parameter estimate lower upper
b0 = -0.3069 -5.7275 5.1138
b1 = 1.0093 1.0026 1.0160
b2 = 0.0000 0.0000 0.0000
s2 = 0.0123 0.0033 0.4862
s= 0.1109 0.0574 0.6973
df = 2
t= 4.3027

158
 EPA Traceability Protocol for Gaseous Calibration Standards

The approximate level of confidence associated with the expanded uncertainties

is 95%. The corresponding intervals take the form estimate +/- tu, where:
estimate is the quantity of interest, u is its standard uncertainty, and k is a
coverage factor. For each model, the coverage factor, t, is determined from the
t-distribution with appropriate degrees of freedom (df). Note that the
uncertainties are not displayed, but can easily be derived from the estimate,
expanded uncertainty, and the coverage factor.

Comparing the two models:

F ratio = 0.67
F critical = 19.1643 (5% significance level)
The overall uncertainties of fit are not significantly different at the 5% level.
The linear model appears to be the better choice.

STEP 3 Review the Charts

View the charts named Curves 1 and Residuals 1. Curves 1 shows bands
illustrating expanded uncertainties for the estimated regression. Compare these
bands with those of the quadratic regression, Curves 2. (Note: You can change
the associated approximate confidence level by changing the "p-value" in cell F4
of the worksheet named Chart Data.) Residuals 1 shows how the calibration
points deviated from the calibration line. Look for a simple pattern (such as a
quadratic curve) in the chart. If such a pattern appears, the quadratic model may
be better. View Residual 2, the deviations from the best- fit quadratic curve. If
Residual 2 effectively removes the simple pattern observed in Residual 1 and if
the magnitude of the deviations has been significantly reduced (as evidenced by
a reduction in the uncertainty u2), then the quadratic model is superior. An F-test
can be run to determine if the two overall uncertainties of fit are significantly
different.

F= 0.6667 Prob. of greater F = 0.64642

The overall uncertainties of fit are not significantly different at the 5% level.
The linear model appears to be the better choice.

If cubic or quartic models are supported by compelling scientific theory or data,

then view their Curves and Residuals charts.

3. Annual Gas Check Worksheet

This worksheet should be used for the annual nonreactive gas calibratoin check
(Sec. 4.1.6.3) and the annual reactive gas calibration check (Sec. 4.1.6.4). It
should only be used after filling out the Pollutant Annual Calibration worksheet
and the Zero Gas Calibration Worksheet.

159
 EPA Traceability Protocol for Gaseous Calibration Standards

You can only fill in the cells in:

STEP 1 Concentration and Uncertainty of Reference Standard

Date of annual calibrations for flow controllers 04/29/12

Fill in the certified concentration of the pollutant 50

Enter the expanded uncertainty (U) from the reference standard's certificate of
analysis. This value is the coverage factor (k=2) times the combined standard
uncertainty (uc). For example, absolute values are given in ppm and relative
values are given in percent. Enter the value as a decimal value (e.g., 1 percent
should be entered as 0.01)

Expanded uncertainty of concentration of the reference standard 0.01

Is the uncertainty absolute? (yes or no) no

It is necessary to answer whether or not this value is absolute or relative.

STEP 2 Concentration and Uncertainty of Check Standard (Lower

Concentration)

Enter the concentration of the check standard 0.5

The check standard concentration must be at least a factor of ten less than the
reference standard concentration and must fall in the well-characterized region of
the calibration curve

Enter the expanded uncertainty (U) from the check standard's certificate of
analysis. This value is the coverage factor (k=2) time the combined standard
uncertainty (uc). It is necessary to answer whether or not this is absolute or
relative. For example, absolute values are given in ppm and relative values are
given in percent Enter the value as a decimal value (e.g., 1 percent should be
entered as 0.01)

Expanded uncertainty of concentration of the check standard 0.01

Is the uncertainty absolute? (yes or no) no

STEP 3 Enter Settings for Pollutant and Zero Gas Flow Controllers

Select settings for the gas flow controllers that fall in the same range as were
used in the annual calibrations and that produce a diluted concentration of the
reference standard that is equal to the undiluted concentration of the check
standard.

160
 EPA Traceability Protocol for Gaseous Calibration Standards

Enter the pollutant flow controller setting 60

Enter the zero gas flow controller setting 5940

Predicted Concentration of Diluted Pollutant 0.499185

Concentration of Check Standard 0.5

Match -0.2 percent

Adjust the pollutant and zero gas settings so that these two concentrations match
within 1 percent. Do not proceed until the Match value is less than 1.0 percent.

STEP 4 Measure Diluted Reference Standard and Undiluted Check Standard

Using a previously calibrated analyzer, measure the diluted reference standard

and the undiluted check standard. You must take at least 3 measurements.

Diluted Reference Undiluted Check

Standard Standard
Measurements Measurements
1.005 1.005
1.000 1.000
0.995 0.995

Mean of Dilution Measurements Mean of Standard Measurements

0.5 0.496

Relative Agreement between Mean Responses

0.0 percent

Acceptable Relative Agreement

Items shaded in light green are internal calculations

Predicted pollutant flowrate Predicted zero gas flowrate
60.23422 5973.016

Pollutant valve relative uncertainty Zero gas valve relative uncertainty

0.001504 0.002996

Reference std concentration Check std concentration

50 0.5

Predicted diluted concentration Relative concentration difference

0.499185 -0.16 %

161
 EPA Traceability Protocol for Gaseous Calibration Standards

Relative uncertainty of reference std Relative uncertainty of check std

0.01 0.01

Mean response for reference std Mean response for check std
1.000 1.000

Std deviation of RS responses Std deviation of CS responses

0.005 0.005

Relative combined std uncertainty for std concs, flow rates, and anal responses
0.016163

Margin of error for confidence

0.03

Lower limit for diluted conc Upper limit for diluted conc.
0.39 0.60

1 1 indicates diluted conc in range, 0 indicates it is not

4. Monthly Flow Check Worksheet

This worksheet should only be used after filling out the Annual Gas Calibration
Check worksheet

It is only valid within one year of 04/29/12

You can only fill in the cells in:

Date of Monthly Flow Check 04/30/12

STEP 1 Enter Flow Rate Uncertainties of Flow Rate Reference Standard

Enter expanded uncertainties (U) for both pollutant gas and zero gas flow rate
reference standards

Pollutant Gas Zero Gas

0.01 0.01

STEP 2 Enter Flow Controller Settings

Enter flow controller setting within range of the annual calibration for both
pollutant gas and zero gas

Pollutant Zero Gas

60 5940

162
 EPA Traceability Protocol for Gaseous Calibration Standards

Predicted pollutant flow rate Predicted zero gas flow rate

60.23422 5973.016

STEP 3 Measure Flow Rate

Enter the flow rates measured by the flow rate reference standard

Pollutant Zero Gas

60 5940

Pollutant Passed Zero Gas Passed

Items shaded in light green are internal calculations

Predicted pollutant flow rate Predicted zero gas flow rate
60.23422 5973.016

PG flow rate RS relative uncertainty ZG flow rate RS relative uncertainty

0.005 0.005

PG flow rate relative standard ZG flow rate relative standard

uncertainty from calibration uncertainty from calibration
0.001504 0.002996

PG flow rate relative ZG flow rate relative

combined std uncertainty combined std uncertainty
0.005221 0.005829

Pollutant gas flow rate Zero gas flow rate

acceptance criterion acceptance *
1.231344 129.3639

5. Day-of-Use Worksheet

This worksheet should only be used after filling out the Monthly Flow Check
worksheet
It is only valid within one month of 04/30/12

You can only fill in the cells in:

STEP 1 Enter Pollutant Gas Concentration and Expanded Uncertainty

Enter pollutant gas concentration

Pollutant Gas Concentration

50

163
 EPA Traceability Protocol for Gaseous Calibration Standards

Enter the relative expanded uncertainty (U/conc) from the reference standard's
certificate of analysis. This value is the coverage factor (k=2) times the
combined standard uncertainty (uc). Enter the value as a decimal value (e.g., 1
percent should be entered as 0.01)

Relative Expanded Uncertainty for Pollutant Gas Concentration

0.01

STEP 2 Enter Desired Flow Rates for Pollutant and Zero Gas

Enter the desired flow rates for both the pollutant and zero gas to achieve your
target dilution and obtain the proper flow controller settings to achieve these flow
rates.

Pollutant Gas Flow Rate Zero Gas Flow Rate

60 5940

Pollutant Gas Flow Controller Setting Zero Gas Flow Controller Setting
59.76781 5906.997

STEP 3 Resulting Diluted Concentration of Pollutant and Expanded Relative

Uncertainty

Diluted Pollutant Gas Concentration Relative Expanded Uncertainty

0.5 0.018599

Items shaded in light green are internal calculations

Relative std uncertainty of PG Relative std uncertainty of ZG
flow rate reference std flow rate reference std
0.005 0.005

Pollutant flow rate setting Zero gas flow rate setting

59.76781 5906.997

Pollutant flow rate relative std Zero gas flow rate relative std
uncertainty from calibration uncertainty from calibration
0.001516 0.00303

Pollutant gas flow rate relative Zero gas flow rate relative
combined std uncertainty combined std uncertainty
0.005225 0.005846

Diluted pollutant gas concentration Relative combined std uncertainty for

diluted concentration
0.5 0.009299

164
 PRESORTED STANDARD
POSTAGE & FEES PAID
EPA
PERMIT NO. G-35

Office of Research and Development (8101R)

Washington, DC 20460
Official Business
Penalty for Private Use
$300