Report

As a partial requirement
For
The course on

Engineering Experimentation
ME 3901, D07

Laboratory Experiment 1:
EXPERIMENTS WITH A DIGITAL OHMMETER

Submitted by:

_________________________
Student A [email protected]

________________________
Student B [email protected]

Submitted to:

Prof. Cosme Furlong

DEPARTMENT OF MECHANICAL ENGINEERING

WORCESTER POLYTECHNIC INSTITUTE
WORCESTER, MA 01609-2280

3/26/07

Lab score:
ORGANIZATION AND PRESENTATION: ________
LABORATORY DESCRIPTION AND OBJECTIVES: ________
METHODOLOGY: ________
RESULTS: ________
INTERPRETATION OF RESULTS (DISCUSSION) ________
CONCLUSIONS AND RECOMMENDATIONS: ________

TOTAL: ________
Table of Contents
Table of Contents ................................................................................................................ 2
Table of Figures .................................................................................................................. 2
List of Tables....................................................................................................................... 2
Abstract ............................................................................................................................... 3
Introduction ......................................................................................................................... 4
Methodology ....................................................................................................................... 6
Results and Discussion........................................................................................................ 8
Conclusions and Recommendations.................................................................................. 13
Appendix A: Instrumentation Used ................................................................................. 15
Appendix B: References.................................................................................................... 16
Appendix C: Visual Instrument Used ............................................................................... 17

Table of Figures

Fig. 1 Calibrated batch readings......................................................................................... 8

Fig. 2 Resolution in measurements of a batch 1 kΩ resistor as a function of top limit ..... 9
Fig. 3 Variations of resistor readings with temperature ................................................... 12
Fig. 4 Screenshot of the VI used ...................................................................................... 17

List of Tables

Table 1 Summarizing statistics of the measurements of mix of resistors ........................ 10

Table 2 Instrumentation used ........................................................................................... 15

2
Abstract
This series of experiments used a digital ohmmeter to collect descriptive statistics on a

batch of 10 carbon 1 kΩ-rated resistors. The device’s ability to work with 2.2Ω , 120Ω

and 47kΩ was also checked. Each measurement in the experiment was repeated 10

times to minimize imprecision due to noise and due to the low resolution of the

ohmmeter used. The batch mean was computed to be 971 ± 2Ω , measured with the

resolution of 4.9 Ω. Then 10 random pairs from the batch were measured in series and in

parallel. The mean of in-series measurements was 1965 ± 0.31%Ω . The mean of in-

parallel measurements was 491.2 ± 0.28%Ω . This agreement is approximately in-line

with the theoretical error propagation for this type of measurement. Resolution of the

system was measured, and found to be exactly linear with the higher limit of the expected

range of measurements. Also, the resistance of a 2.2 Ω resistor was found to slightly

increase when heated with a soldering iron, thus confirming the theoretical relation.

3
Introduction
This series of short experiments with the digital ohmmeter served to inspect the

technicalities of analog to digital conversion, as well as analyze what the limiting factors

to the accuracy and precision of the measured value are when making this kind of

measurements. Population statistics were collected and computed for a batch of 10 1 kΩ-

rated resistors. Also, the uncertainty propagation and the effect of temperature on the

resistance of a carbon resistor were analyzed.

The limitations of the device were determined by measuring a mix of 2.2 Ω, 120 Ω and

47 kΩ. The resolution of the device was measured by a series of 100 measurements of a

1 kΩ resistor with expected higher value of the range varying between 1 kΩ, 2 kΩ, 5 kΩ

and 10 kΩ.

Ref. [1] contains the general procedure followed during this series of experiments, for

more details see Methodology. The digital ohmmeter used was controlled using

LabVIEW 8 visual instrument (see Appendix C for screenshot); 4-wire type of

measurements with fixed current and varying voltage were used. This ohmmeter worked

by running a fixed value of current ( 1.0 ± 0.5mA ) through the resistor, and measuring the

resulting voltage across the resistor.

The following equations were used for predicting the uncertainties:

R - resistance being measured

δ R - uncertainty in resistance being measured
I - current put through the resistor (fixed at 1.0 ± 0.5mA)
V - voltage created by the current sent through the resistor
2 2
V § δV · § δ I ·
R= , thus δ R = R ⋅ ¨ ¸ + ¨ − ¸ (Ref. [4]).
I © V ¹ © I ¹

4
The A/D converter used in this experiment was 12-bit, and therefore the resolution of the

Rmax − Rmin
digital ohmmeter can be theoretically predicted by: res = . This equation
212

holds true until Rmax − Rmin does not become too small, and the A/D converter’s

maximum resolution (4.9 mV) does not come into play. As such, we can predict seeing

the resolution max out at a given interval for each resistor.

5
Methodology
A LabVIEW 8 Visual Instrument was created to allow 4-wire resistance measurements

(see Appendix C for screenshot, Ref [1] for instructions on creating the VI). Signal

conditioning chassis and screw terminal from National Instruments were used (see

Appendix A for complete list of instrumentation used). The VI was configured to record

the data automatically in a Microsoft Excel spreadsheet.

The program read the current and voltage running through a resistor and displayed values

of resistance for each resistor. It was set to take ten readings for each resistor, display the

readings on a digital display as well as a dial, and save the data to a Microsoft Excel

spreadsheet.

0. Several test readings were taken to make sure the digital ohmmeter was

functioning (although not necessarily calibrated); that data was then discarded.

Any measurement in this experiment from this point on was repeated 10 times, to account

for fluctuations in current, temperature, and other sources of noise.

1. A batch of 10 1 kΩ-rated resistors was measured, and both the batch and

individual resistor statistics calculated and recorded. For these measurements DAQ

limits (set in the VI) of [1Ω, 10 kΩ] were used. These measurements were made to

analyze the variance in a batch of similar resistors. One of these resistors was also

measured with a multimeter; that data was used for offset in calibration of the ohmmeter.

2. 100 consecutive readings were run on a batch resistor with a varying top limit.

This was done to identify the dependence of resolution on top limit.

6
3. Then the ohmmeter was tested on a mix of three resistors with ratings of 2.2 Ω,

120 Ω and 47 kΩ; limits were varied on these measurements to analyze how high the

ohmmeter’s resolution can go on each scale.

4. 10 random pairs from the batch (picked as a simple random sample) were then

measured in series and in parallel, in order to analyze how the population variance will

propagate into these measurements. Limits were also adjusted in this measurement, to

remove resolution as a source of uncertainty.

5. Finally, the 2.2 Ω-rated resistor was measured with high resolution, and then

heated with a soldering iron (see Appendix A for details on the instrumentation) and

remeasured. This was for qualitatively verifying the variation of resistance as a function

of temperature.

7
Results and Discussion
1. A batch of 10 1 kΩ-rated resistors was measured and both the batch and

individual resistor statistics calculated and recorded. Limits of [1Ω, 10 kΩ] were used;

these limits provided resolution of 2.441Ω , both theoretically (see Introduction) and

experimentally (see Fig. 2 for resolution measurements). Fig. 1 contains the data

collected during this step.

990

985
Accepted
Measured resistance (Ω

Data
980
Rejected
Data
975

Accepted
970 Data Mean

965
1 2 3 4 5 6 7 8 9 10
Resistor number

Fig. 1 Calibrated batch readings

Resistor 10 in the batch was measured (with the digital ohmmeter) to be 979.9 ± 1.7Ω .

When measured with a calibrated multimeter (see Appendix A for instrumentation used),

it was measured to be 969 Ω. Therefore, all batch data was offset by -10.9 Ω, Fig. 1

contains calibrated measurements. No slope calibrations were conducted, so only batch

data is calibrated in this experiment. Also, Chauvenet’s criterion was used to reject

Resistor #4 from the batch. The final batch statistics are 971 ± 2Ω . Thus the probability

8
of a batch resistor falling outside the specified limits (being outside the [950, 1050] Ω

limit) is 0 to at least 15 decimal places. From Figure 1 we can clearly see that the

variance in the batch is primarily due to the differences between the individual resistors,

and not the variances of the individual resistors themselves. The variances of the

individual resistors themselves are mostly due to the relatively low resolution during

these measurements.

3
y = 0.0002x + 0.0004
R2 = 1
2.5
Resolution (Ω

1.5

0.5

0
0 2000 4000 6000 8000 10000 12000
Top Limit (Ω)

Fig. 2 Resolution in measurements of a batch 1 kΩ resistor as a function of top limit

2. 100 consecutive readings were run on a batch resistor with a varying top limit.

This was done to identify the dependence of resolution on top limit. See Figure 2 for the

outcome.

We can see that the resolution exhibited a perfectly linear variation as a function of top

limit. However, the lower limit remained 1 during this measurement, and thus the

number of bits in the signal conditioner never became the issue. It does become the

9
limiting factor when the entire interval is adjusted, as is seen in measuring non-batch

resistors.

3. The ohmmeter was tested on a mix of three resistors with ratings of 2.2 Ω, 120 Ω

and 47 kΩ; limits were varied on these measurements to analyze how high the

ohmmeter’s resolution can go on each scale. See Table 1 for the summarizing statistics

of this data.

Table 1 Summarizing statistics of the measurements of mix of resistors

Resistor (Ω) Bottom Limit (Ω) Top Limit (Ω) x (Ω) σ (Ω) Resolution (Ω)

1 10000 119.0 1.2 2.441

120 110 130 118.11 0.03 0.049

115 125 118.09 0.03 0.049

1 10000 3.1 1.9 2.441

2.2 0.1 5 2.2348 0.0016 0.002

2 2.5 2.236 0.002 0.003

47000 43000 50000 8162.0 1.5 2.441

We can see that with the 2.2 Ω and 120 Ω resistors, the resolution increases up to a point,

and then the number of bits in the signal conditioner becomes the limiting factor. We can

also see that the variance of the samples depends largely on the resolution.

Also, we can see that the 47 kΩ resistor could not be measured properly with this

apparatus. That is because the ohmmeter used in this laboratory is fixed to 1 mA of

current, and that produces voltage of 47 V across the resistor. The apparatus is limited to

[-10, 10] V, and there is no attenuation available in this setup. So the highest resistance

10
that can be measured with this apparatus is one in which no signal attenuation is

10V
necessary, which theoretically is: Rmax = = 10k Ω , assuming absolutely no noise or
1mA

variations in current. However, we can see that the actual measurement peaked at

8162.0 ± 1.5Ω - likely due to variations in current.

4. 10 random pairs from the batch (picked as a simple random sample) were then

measured in series and in parallel, in order to analyze how the population variance will

propagate into these measurements. Limits of [1900, 2100] were used on the in-series

measurements, and [400, 600] on the parallel measurements. That was to maximize the

variance in the batch as the leading cause of uncertainty and minimize uncertainty due to

the resolution of the ohmmeter. Again, the batch statistic was measured in part 1 of the

experiment to be 971 ± 2Ω . Notice, that this number is calibrated by an offset, and the in-

parallel and in-series measurements are not calibrated by an offset. That, however, is not

a problem for this experiment, because the variance is not affected by a bias shift. The

final in-parallel descriptive statistic was 491.2 ± 1.4Ω , and in-series was 1965 ± 6Ω .

Theoretically, if a random resistor from the batch has mean resistance R and

uncertainty δ R , then the error in the mean of a simple random sample of in-parallel pair

δR
measurements will be , and for in-series pairs it will be 2δ R (Ref. [4]). That is the
2

approximately the behavior we are observing here: the error approximately doubled from

the batch to the in-series measurements (from 2 Ω to 6 Ω) and halved for the in-parallel

measurements (from 2 Ω to 1.4 Ω). If more simple random pairs were measured, the

errors in the sample means would further approach their theoretical values (which in this

case are 4 Ω for the in-series measurements and 1.0 Ω for the in-parallel measurements).

11
5. Finally, the 2.2 Ω-rated resistor was measured with high resolution, and then heated

with a soldering iron (see Appendix A for details on the instrumentation) and remeasured.

This was for qualitatively verifying the variation of resistance as a function of

temperature. The 2.2 Ω resistor was used because it should have the least material in it,

so it is easier to heat up; also the variation in the measurements of this resistor are the

least, and the variation due to temperature is thus easier to identify. See Figure 3 for

observed variation.

2.29

2.28

2.27
Measured resistance (Ω

2.26

2.25

2.24

2.23

2.22

2.21
Unheated resistor Heated resistor

Fig. 3 Variations of resistor readings with temperature

We can see that there is a small observed increase in resistance with temperature.
However, there is a large observed increase in variance as well – possibly due to
temperature fluctuations due to air flow.

12
Conclusions and Recommendations
The batch was calculated to have descriptive statistics of 971 ± 2Ω . It is less than the

rated 1 kΩ, but still very well within the specified tolerances. Thus while a linear offset

of 29 Ω may be recommended to the manufacturer, it is not necessary. The probability of

a resistor being outside the specified tolerance is 0 to at least 15 decimal places.

Resolution (minimum difference between two nearby values) increases linearly with

increase in the higher limit, and decreases linearly with decrease in the higher limit. This

behavior is true as long as the number of bits in the A/D converter of the signal

conditioner does not become the limiting factor.

Linear offset was used for calibration. However, it is recommended to also calculate a

slope offset to fully calibrate this ohmmeter. To do so, several additional measurements

of resistors other than 1 kΩ must be taken, and a linear least squares regression conducted

to calculate the β.

The ohmmeter cannot accurately measure resistances above about 8 kΩ, because the

produced voltage across the resistor becomes too high, and the signal conditioner cannot

attenuate the signal.

Error was observed to propagate from batch to in-parallel and in-series readings exactly

as expected from the theory of error propagation. If the number of simple random pairs

measured in-series and in-parallel was increased, the variance in the in-series and in-

parallel means is expected to even further approach the expected values.

Small increase in resistance was observed when a resistor was heated with soldering iron.

It is recommended that more controlled conditions are used to further verify this behavior

– i.e., a non-conducting bath or thermal chamber used to heat the resistor. Also, the

variation observed was very small, so it is likely that the resistors are intentionally

13
manufactured for their resistance to vary as little as possible with the temperature, since

they are not thermistors.

14
Appendix A: Instrumentation Used

Table 2 Instrumentation used

Device Model Number Manufacturer Serial Number Calibration

due date

Multimeter 22-809 RadioShack 06A01 Unknown/Not

Calibrated

Signal SCXI-1000 National N/A Unknown/Not

Conditioner Instruments Calibrated

Chassis

Signal SCXI-1122 National N/A Unknown/Not

Conditioner Instruments Calibrated

Screw SCXI-1322 National ASSY182385- Unknown/Not

Terminal Strip Rev. A Instruments 01 Calibrated

Soldering Iron 25W, 120V, Weller 336B N/A

60Hz

LabVIEW 8 Student Edition National Uknown N/A

Software Instruments

Package

Microsoft N/A Microsoft N/A N/A

Office 2003

15
Appendix B: References
[1]. Documents posted on http://users.wpi.edu/~cfurlong/me3901.html:

Lab1: Description, Procedure: Part 1, Procedure: Part 2, Resistors: Color codes.

[2]. T.G. Beckwith, R.D. Marangoni, and J.H. Lienhard, Mechanical Measurements, 6th

ed., Prentice Hall, 2007

[3]. R.H. Bishop, LabVIEW 8, Prentice Hall, 2007

[4]. John R. Taylor, An Introduction to Error Analysis, 2nd ed., University Science Books,

1997

16
Appendix C: Visual Instrument Used

Fig. 4 Screenshot of the VI used

17