Richard J.

Harris, G3OTK
10 South Street, South Petherton, Somerset TA13 5AD, United Kingdom; [email protected]

An Automated Method for

Measuring Quartz Crystals
G3OTK proposes a method for automatically measuring all of the components
of the equivalent circuit of quartz crystals using a Colpitts oscillator.

Ladder crystal filters have become popu- without. Many years ago G3UUR proposed for, say, 100 crystals is time consuming.
lar in home-brew transceivers and QRP kits a simple method using a Colpitts oscillator, This article describes a further develop-
because of the availability of very low cost and many references can be found to it on the ment of the oscillator technique that not only
crystals. It seems to me that many projects Internet.1 As originally proposed, this method gives an accurate figure for the motional
use filters that have not been designed, but only gives a ball-park figure for the motional capacitance, motional inductance, holder
have been constructed on the “if it sounds capacitance because it does not take into ac- capacitance and series resonant frequency
right then it is right” principle. Although count the two capacitors of the Colpitts oscil- but also gives a good estimate for the mo-
software for the design of ladder filters is lator, nor the holder capacitance of the crystal. tional resistance. A microprocessor controls
available, a filter cannot be designed if the Also, it does not give any information about the measurements, makes the calculations,
crystals have not been characterized and the the motional resistance of the crystal or its shows the results on an organic light emit-
equivalent circuit components, or motional series resonant frequency. ting diode (OLED) display and also sends
parameters, determined with some accuracy. All of the methods for evaluating the them to a computer running a spreadsheet
Many methods for measuring the equiva- motional parameters of quartz crystals that program. Each crystal can be characterized
lent circuit of quartz crystals have been pro- I have read about in Amateur Radio maga- in a few seconds and the results sorted into
posed over the years. These can be divided zines require the measurements to be en- groups those with similar properties. The
into two classes: those requiring a stable sig- tered into formulae by hand to give the final tabulated data will also give an insight into
nal generator as an excitation source and those parameter values. Making the measurements the spread of the motional parameters as-
where the crystal is used in an oscillator. This and undertaking the subsequent calculations sociated with inexpensive crystals. These
latter class has the advantage that the principle are intended to be used in oscillators, and the
item of test equipment is a frequency coun- information required for filter design is not
ter, which few serious experimenters will be Notes appear on page 8.
1
part of the specification.

QEX – November/December 2013 3

 QX1309-Harris01 C3 C4
+V Cosc = [Eq 3]
Crystal
C3 + C4
R1 C5
Y1
X The small signal equivalent circuit is
Q1 shown in Figure 2. Cm, Lm and Rm are the
motional parameters of the crystal. With the
S1
C3
Output switch SW1 at position 0, the frequency of
2
R2
oscillation f0 is given by Equation 4.
0 1
1
C4 Q2 ALC f0 = [Eq 4]
C1 C2
L C ( C + Cosc )
X' 2π m m h
Gnd Cm + Ch + Cosc

Figure 1 — This schematic is a basic crystal test circuit.

With the switch in position n, where n is 1
or 2, the frequency fn is given by Equation 5.
1
fn =
 Cn Cosc 
resistance, Rosc, is the real part of Zin, then at Lm Cm  Ch + 
Outline of the Measurement Method  ( Cn + Cosc ) 
The outline is shown in Figure 1. The
a temperature of 21°C (70°F), it is given by 2π
Equation 2. Cn Cosc
crystal under test, X1, is connected in series Cm + Ch +
with a switch, SW1, which selects 0 V, C1 [Eq 2] ( Cn + Cosc )
 BI 
or C2. Q1, C3 and C4 form a Colpitts oscil- Rosc = −  2 e  [Eq 5]
lator with the amplitude maintained at 1 mV  f C3 C4 
by an Automatic Level Control (ALC) loop. We will ignore the effect of the holder ca-
At this low level, Q1 operates in a linear and pacitance Ch because it is much smaller than We can derive two equations for Cm by
predictable manner. For the amplitude to the series combination of C3 and C4, and will combining the equations for switch positions
be constant, the motional resistance of the be taken into account when the constant B 1 and 0 (Cm10) and switch positions 2 and 0
crystal must be balanced precisely by the is determined. The motional resistance, Rm, (Cm20), eliminating Lm in the process.
negative resistance generated by the Colpitts is equal to the magnitude of the oscillator
Oscillator. This is proportional to the emitter resistance, Rosc, and so is proportional to the   f 2 
current and gives a means of determining the emitter current.   1  − 1
The imaginary part of Zin is the series   f0  
motional resistance.  
combination of C3 and C4. We will call Cm10 =
The loading effect of the bias resistors R1  
and R2 is very small and can be ignored. At this combination Cosc, which is given by  2 
Equation 3.  1  f  1 
a frequency, f , the input impedance of the − 1 
Colpitts oscillator at points X-X’, as seen by  C1 Cosc   f 0  Ch + Cosc 
the crystal, are given by Equation 1.   Ch + C1 + C  
 osc  
[Eq 6]
 −q I e 
Z in = B  
 ( 2π f ) C3 C4 K T 
2
  Crystal
[Eq 1] X
j  1 1 
−  +    f 2 
2π f  C3 C4    2  − 1
where:   f0  
Lm
Cm 20 =  
K is Boltzmann’s constant (1.3807  
× 10–23 m2 kg s–2 K–1) Cm
Ch C osc
 2 
T is the temperature in kelvins  1  f  1 
− 2 
q is the charge on an elec- Rm  C2 Cosc   f 0  Ch + Cosc 
tron (1.6022 × 10–19 C)   Ch +  
 C2 + Cosc  
Ie is the emitter current. R osc
S1 [Eq 7]
I have added the constant B to make a first 2
The holder capacitance of the crystal will
order correction for the effects of the reduced 0 1 be the value of Ch that makes Cm10 equal to
current gain of Q1 at frequencies typical of Cm20, which will then be the motional capaci-
the crystals used in CW and SSB filters. I will C1 C2
X' tance Cm. I don’t have an analytical solution
describe a method for determining it later. for determining Ch, so I solve it numerically.
The 1 or 2 W emitter bulk resistance of the QX1309-Harris02
For AT-cut crystals in standard or low profile
2N3904 oscillator transistor can be ignored HC49 packages, Ch will be within the range
if C3 and C4 are chosen so that the emitter 1.5 to 6 pF. The microprocessor controller
current is less than 0.5 mA. If the oscillator Figure.2 — Here is a diagram of an oscillator
small signal equivalent circuit. measures the three frequencies and then steps

4 QEX – November/December 2013

 Figure 3 — The schematic/block diagram hybrid of the author’s crystal measurement unit.

through the range of possible values for Ch C2 (47 pF), both NP0 ceramic capacitors, tor for Cref, which I measured to be 220.0 pF
in increments of 0.1 pF, calculating Cm10 and could be measured before being fitted into using an Almost All Digital Electronics
Cm20 at each step, until equality is found. the circuit, but relays K1, K2 and K3 add (AADE) LC meter. The inductor and the
Once Cm and Ch have been determined, additional stray capacitance and the calcu- Colpitts capacitors are chosen so that f0 is
the series resonant frequency, fs, of the crys- lated motional and holder capacitances will about 7 MHz, midway between the frequen-
tal can be calculated from any of the three be in error. My solution is to measure the cies of the crystals to be measured. The
measurements of frequency. The simplest capacitance in situ and this is the purpose microprocessor that I use cannot measure
formula is when SW1 is in position 0. of the Colpitts LC oscillator. Capacitance is frequencies above a couple of hundred ki-
measured relative to the reference capacitor lohertz, and so the frequency is divided by
 Cm  Cref. With the crystal oscillator turned off, 100 before being counted. I found that the
=f s f 0  1 −  [Eq 8]
 2 ( Ch + Cosc ) 
and with K2 or K3 selected, three frequency stray capacitance associated with the relays
measurements are made. was about 7 pF, a significant addition to C1
The motional inductance Lm and Q can With relays K4 and K5 open, let the fre- and C2.
now be calculated by Equations 0 and 10. quency be f0. With K5 open and K4 closed let When crystals are being assessed, the
the frequency be fref. With K5 closed and K4 LC Colpitts oscillator is turned off and K5
1 open let the frequency be fx. K2 and K3 are
Lm = [Eq 9] opened. To bring the frequency within the
( 2π f s ) Cm
2
open or closed as appropriate to the capacitor counting range of the processor, the ampli-
being measured. The unknown capacitance fied output of the crystal oscillator is mixed
Cx can be calculated in terms of the ratio of with an external oscillator and converted
2π f s Lm the frequencies and the reference capacitor, down to an intermediate frequency of 1 to
Q= [Eq 10]
Cref.
Rm 3 kHz. A CMOS exclusive-OR gate is used
as a digital mixer and an RC low pass filter
We have now found all four of the compo-  2  cleans up the output for counting. The crystal
nent values and the Q of the equivalent circuit   f0  − 1  oscillator frequency is calculated from that
for the fundamental mode of operation.  
  fx   of the external oscillator and the intermediate
C x = Cref  2  [Eq 11] frequency.
  f0   The processor is a PICAXE 40X2, a 40
A Practical Measuring Instrument   f  − 1  pin PIC with an integral “PICAXE basic” in-
The block diagram of the measuring   ref  
equipment is shown in Figure 3. Reed relays terpreter, and which is very easy to program.
are used to switch capacitors in series with It controls the measurement sequence and
the crystal being measured. C1 (220 pF) and This method is similar to that described also communicates by means of I2C with a
by Carver.2 I selected an NP0 ceramic capaci- mM-FPU coprocessor that uses 32 bit float-

QEX – November/December 2013 5

6
QEX – November/December 2013
Figure 4 — This schematic shows the amplitude controlled Colpitts oscillator circuit used in the author’s test system.
ing-point arithmetic. All of the calculations
are made within the unit and the results are 30 35

Number Of Crystals
30

Number Of Crystals
displayed on a 4 line by 20 character OLED 25
display and are also sent to a computer run- 20 25
ning Microsoft Excel. Crystals can be mea- 20
15
sured at a rate of two or three a minute and 15
10
the tabulated results sorted to group those 10
5
with the similar values, such as motional 5
0 0
inductance, series resonant frequency and
0.012 0.014 0.016 0.018 0.020 4,998.6 4,998.8 4,999.0 4,999.2
motional resistance.
The level controlled crystal oscillator is at Motional Capacitance (pF) Series Resonant Frequency
the heart of this unit and the schematic dia- QX1311-Harris05 (kHz)
QX1311-Harris06
gram is shown in Figure 4. The crystal and
capacitor switching relays are connected to
Figure 5 — This graph shows the spread of Figure. 6 — This graph shows the spread of
X-X’. Q1 with C3 and C4 form the Colpitts motional capacitance, Cm, for a batch of 100 series resonant frequencies, fs, for the same
oscillator with the emitter current controlled crystals. 100 crystals.
by Q5. A unity gain buffer consisting of Q2
and Q3 drives a high pass filter, C7, L1 and
C8, cascaded with a low pass filter,C9, L2
40
and C10, giving a pass band between 1 MHz 0.020

Motional Capacitance (pF)

35
and 20 MHz. U1 is a wideband current
Number Of Crystals

30 0.018
mode amplifier with a gain of approximately
25
20, which drives the ALC detector U4, an
20 0.016
AD8307 logarithmic amplifier.
15
The long tailed pair Q6 and Q7compares
10 0.014
the detected signal with a reference voltage
5
set by R18 and the ALC loop is completed 0 0.012
by means of Q5. A small voltage is devel- 4,998.6 4,998.8 4,999.0
10 20 30 40 50 60 70
oped across R15, amplified by U5, and then Series Resonant Frequency
digitized and scaled by the processor to give Motional Resistance (kHz)
the oscillator emitter current. Finally, U2 (Ω) QX1311-Harris08
QX1311-Harris07
provides further amplification to drive the
digital mixer. Figure. 7 — This graph shows the spread of Figure. 8 — Here is a plot of the motional
The formula for the negative resistance motional resistance, Rm, for the batch of 100 capacitance, Cm, versus series resonant fre-
of the oscillator has a constant, B, to correct crystals. quency, fs, for the batch of crystals.
for the deviation from the theoretical value
of unity because the reduced current gain
of Q1 at high frequencies lowers the input
resistance of the transistor. We can make a
reasonable estimate of B by first using the
equipment to measure the motional resis-
tance of a number of crystals and choosing
the one with the lowest value. The link in
series with the crystal is removed and several
fixed resistors, for example 10, 15, 22 W, fit-
ted in turn and measurements made again.
The motional resistance now includes a
known fixed resistor. The constant, B, is the
best fit value that makes the increase in the
measured resistances to be the same as the
resistors used. For the 2N3904, I found that
B was 1.20 for 5 MHz crystals and 1.35 for
10 MHz crystals.
The oscillator unit uses “Manhattan” style
construction and is shown in Photo A. The
crystal oscillator, capacitor switching relays
and ALC circuit are on the left hand side. The
lead photo on page 3 shows the assembled
measurement unit and controller. The ex-
ternal oscillator, a homemade DDS signal
generator, is not shown in the photograph.
The calculated values of the motional ca-
pacitance and inductance are displayed to six
significant digits, despite being only accurate
Photo A

QEX – November/December 2013 7

 to about 1%, because otherwise the center so can fs. If motional capacitances are plotted using the oscillator method. This discrepancy
frequency of a filter design using values to, against series resonant frequencies, however, is probably due to the additional inductance
say, 3 significant figures could be in error by then the result is a straight line with a slope of the wiring through the relays, which I esti-
tens of kilohertz. given by Equation 13. mate to be 60 nH.
I also calculated the motional resistances
d fs − fp of the sample from the source voltage, the
Some Results = [Eq 13]
I have measured several hundreds of crys- d Cm 2 ( Ch + C p ) voltage across the crystals and series resis-
tals. I will describe the results of one batch tor value. The motional resistances agreed
consisting of 100 crystals purchased from a where fp is the nominal frequency with the quite well with the oscillator method, being
major UK component distributor. The nomi- load capacitance Cp. The slope is –73.5 Hz / on average 6% higher. I have found that the
nal frequency was 5 MHz, with a load ca- f F for 5 MHz crystals with a specified load motional resistance of some crystals, how-
pacitance of 30 pF and a tolerance of 30 ppm. capacitance of 30 pF and holder capacitance ever, is not constant and varies with crystal
These were housed in HC49/U packages and of 4.0 pF. The motional capacitances and current so that such comparisons are not en-
cost less than 30 cents each. the series resonant frequencies of the crys- tirely reliable.5
Before we look at the spread of these tals are shown in Figure 8, along with the The holder capacitance was within 0.1 to
motional parameters, it is worth consider- least squares trend line, which has a slope of 0.4 pF of that measured using an LC meter
ing what the selection priorities should be –76.1 Hz / f F, close to the calculated figure. for 5 and 10 MHz crystals in both standard
for crystals that are to be used in a filter. My The motional inductance has a similar rela- and low profile packages.
personal view has been that the motional tionship, although with a positive slope.
inductance is the first consideration, and the Conclusion
crystals should be selected to have as small There is a fortunate consequence of this
A Colpitts oscillator has been described
a spread as possible. From this sub-set those relationship. Crystals selected for a small
that can be used as the basis of an instrument
with similar series resonant frequencies can spread of series resonance will also exhibit a
for measuring all of the motional parameters
be selected. Most crystals in a filter will be small spread of both motional inductance and
of quartz crystals. When controlled by a
in series with capacitors to give the correct capacitance. Selecting crystals based solely
microprocessor, the results can be sent to a
mesh frequencies and if necessary these on similar series resonant frequencies is a vi-
spreadsheet and displayed in tabular form to
capacitors can be adjusted to take into ac- able method. Furthermore, crystals from dif-
enable crystals with similar properties to be
count small differences in the series resonant ferent manufacturers with the same nominal
selected. With the present equipment, up to
frequencies of the crystals. Finally, from this frequency, load and holder capacitances can
150 crystals an hour can be measured.
much smaller sub-set, those with similar, and be mixed and selected on this basis.
small, motional resistances can be selected. I constructed a 5 MHz pre-distorted lin-
ear phase filter designed from the tables of Richard Harris was licensed as G3OTK
I numbered the crystals individually and in 1961. He received Bachelor and Master
measured the motional parameters of each k and q values in Zverev using the average
motional parameters of six crystals selected Degrees in Electrical Engineering from
one. The time taken to measure 100 crys- the University of Bath in the UK. He is a
tals was about 40 minutes. The results were from this batch.3 I added L-match circuits to
Member of the Institution of Engineering
displayed in tabular form in an Excel spread the input and output of the filter to match to
and Technology and a Chartered Engineer.
sheet. The important motional parameters the 50 W terminations of the signal genera- Although he has spent much of his profes-
were sorted into various “bins,” 0.5 f F wide tor and detector. The plotted amplitude and sional life undertaking electronic design, for
for the motional capacitance, 50 Hz wide for phase responses overlaid those predicted by the last ten years he has been responsible for
the series resonant frequency and 5 W wide SPICE over a 60 dB amplitude range (the Quality Assurance, Health and Safety and
for the motional resistance. The distributions limit of the test equipment) and a phase range Environmental Management. He is a member
of these motional parameters are shown in of 900°. of the Itchen Valley Amateur Radio Club.
Figures 5, 6 and 7. More than 25% of the
crystals fell within one “bin” on each of the Accuracy Notes
graphs but, unfortunately, there is no guar- I have found that the measured motional 1
Wes Hayward, W7ZOI, Rick Campbell, KK7B
antee that the same crystals were in each of capacitance and inductance of other crystals and Bob Larkin, W7PUA, Experimental
these bins. agreed with another method that I have de- Methods in RF Design, published by the
Inexpensive crystals are manufactured for ARRL, 2009, ISBN: 978-087259-923-9;
scribed to within better than 1%.4 Ultimately, ARRL Publication Order No. 9239, $49.95.
use in oscillators. Provided that the frequency all measurements are related to the accuracy ARRL publications are available from
with the specified parallel load capacitance is of my LC meter, which is specified to be 1%. your local ARRL dealer or from the ARRL
within the stated tolerance, then the motional As a check, I measured the series resonant Bookstore. Telephone toll free in the US:
capacitance and inductance can in principle 888-277-5289, or call 860-594-0355, fax
frequencies of a sample of 30 crystals by con- 860-594-0303; www.arrl.org/shop; pub-
take any value. The series resonant frequency necting the output of a DDS signal generator [email protected].
is given by Equation 12. to each crystal in turn through a series resis- 2
B. Carver, “The LC Tester,” Communications
tor and adjusting the frequency for minimum Quarterly, Winter 1993, pp 19 – 27.
 Cm  3
A. I. Zverev, Handbook of Filter Synthesis,
f s f p 1 −
=  [Eq 12] voltage across the crystal. I then re-measured Wiley-Interscience.
 2 ( Ch + C p )  the same crystals using the Colpitts oscillator 4
Richard J. Harris, G3OTK, “Crystal Bridge
 
method so that I could make a comparison at — A Balanced Bridge for Measuring Quartz
the same room temperature. The series reso- Crystal Parameters”, Rad Com, September
where fp is the specified resonant frequency 2011, pp 50 – 52.
with the specified parallel load capacitance, nant frequency using the direct measurement 5
Richard J. Harris, G3OTK, “The Drive Level
Cp. So if Cm can take a range of values, then was on average 13 Hz higher than that found Sensitivity of Quartz Crystals,” QEX Jan/Feb
2013 pp 14 – 21.

8 QEX – November/December 2013