EMRFD Message Archive 8895

Message Date From Subject
8895 2013-08-05 09:55:25 Bob L Software for the Novice Crystal Filter Designer
As part of designing my first crystal filter, I reviewed the software offerings on the internet. As a good learning tool I recommend The Dishal Crystal Ladder Filter program, by DJ6EV. It is available here, as zip files:



There is no installation. Just extract the files and save them in a folder. The help file is available in German and English.

I like this program because it allows efficient “what if” games. The user can change a parameter and immediately see the result. The program provides pull down menus and programs. It supports Cohn, Butterworth, Chebychev and a ripple-minimized-Cohn filter designs.

Pull down programs support either of the crystal measurement methods found on EMRFD, page 3.18.

Here is an interesting exercise: Compare the filter bandpass for an eight crystal Butterworth filter with a six crystal Chebychev filter with 0.1db ripple.

Now a question: The G3UUR measurement method (EMRFD Figure 3.35) does not provide measurement of crystal Q or series resistance. Are measured Lm, Cm or frequency shift useful in eliminating “low Q” crystals?
8897 2013-08-05 19:56:00 Bob L Re: Software for the Novice Crystal Filter Designer
I asked about calculating crystal Q when using the G3UUR characterization method.

On page 6 of the of the DJ6EV help file is the following paragraph, which I missed on the first read:

"Although the xtal loss resistance cannot be measured directly by this [G3UUR] method, the oscillator amplitude level can be used as a good measure of xtal quality factor and resulting activity. This allows the sorting of xtals according to their relative quality factor and to weed out bad xtals. Only a diode detector at the oscillator output stage and a high-resistance voltmeter (DVM) are necessary for these comparative measurements."

Based on this additional voltage measurement the DJ6EV software, used with the G3UUR oscillator, appears to provide a very practical and sufficient mechanism for crystal filter design.

Does that sound right?
8900 2013-08-05 20:31:15 William Carver Re: Software for the Novice Crystal Filter Designer
You are asking for opinions, Bob, and I expect you'll get a variety of
responses. So heres one.....

The current crop of Chinese crystals often have several junk ones
(Q=35,000) in a batch of 20. I've not looked at oscillator amplitude as
a rough indicator of quality, but if DJ7EV says you can spot crap
crystals by looking at his oscillators amplitude I expect you can.

After doing that, if you want to try something simple like a 3 or 4-pole
Cohn filter you can give it a try without measuring the series
resistance of the crystals. Just don't get upset/discouraged when it
works but has some bumps and dips that you didn't expect.

It is NOT magic, when you put known crystals and known capacitors
together and impedance match them to the circuits, the filter WILL work
and look just like the plot the design software put on the screen.
Measuring the xtals is a lot of tedious gruntwork, but once you get that
right the filter just falls together and looks just like the picture on
the screen.

I didn't skip trying to measure the series resistance and Q, but I still
don't think I got it quite right on the early filters. But I was pretty
jazzed and made a better tester and eventually I reached a point where I
could make filters that money can't buy (at least not in quantity of 1).

Bill - W7AAZ
8901 2013-08-06 19:36:09 Bob L Re: Software for the Novice Crystal Filter Designer
Does the G3UUR oscillator process give me the measures I need to identify a junk batch of crystals with a Q of 35,000?

For example, is the percent of frequency shift, when the serial crystal is added, a functi
8903 2013-08-06 19:45:08 William Carver Re: Software for the Novice Crystal Filter Designer
To be honest, Bob, I've not used the G3URR method. Someone who has can
probably answer your question.

On Tue, 2013-08-06 at 21:53 +0000, Bob L wrote:
> Does the G3UUR oscillator process give me the measures I need to
> identify a junk batch of crystals with a Q of 35,000?
> For example, is the percent of frequency shift, when the serial
> crystal is added, a functi
8904 2013-08-06 21:53:19 Kerry Re: Software for the Novice Crystal Filter Designer
The G3UUR method gives Cm; Co is required for the calculation and this must be measured separately with a capacitance meter or the like.

Since frequency is measured during the test it is then possible to use the value of Cm to calculate Lm using the normal resonance equation .

Rm and Q cannot be determined by this method.

It's a most ingenious method and I doffs me lid to G3UUR for discovering it but it's not a complete method for determining crystal parameters.

As Bill said earlier, it's most desirable that all parameters (Q, Rm, Cm, Co & Lm) be measured with some accuracy.

Discounting VNAs, I think the 3dB resonance method is probably the simplest way for most experimenters to obtain all parameters.

It requires a signal generator, a test fixture, perhaps an attenuator or two and a detector. A capacitance meter is required to measure Co.

You have a signal generator; it requires some "clean-up" as I understand it but, provided a reasonably clean signal can be obtained, it should be OK.

Here is part of the original setup that I used for this method;


I used a maximum power to the crystal of about -30dBm; that doesn't seem a lot, 1 microwatt, but it is a lot in crystal terms.

The attenuator reduces the input power and helps set a 50-ohm impedance environment for the crystal.

Note that a 50-ohm environment is not ideal; the lower the impedance the more accurate the measurements will be. This is because we are loking for unloaded Q, Qu, but we can't measure it in an "unloaded" test setup so we measure loaded Q, Ql, and convert that to Qu.

Essentially, the lower the load the more accurate is the conversion; Bill has espoused load environments as low as one ohm which will certainly give good results but require a very sensitive detector as the output from a one-ohm fixture is very small.

The EIA standard is for a 12.5 ohm fixture, possibly chosen so that simple 4:1 impedance transformers can be used. I have not used them; I use impedance-changing attenuators which bring a penalty of about 15 dB loss (and there is one on each side of the DUT).

But, for starters, a simple fixture such as the one in the photo will give quite good results; remember Wes' advice that any measurement is better than none, to which I would add "provided you understand the limitations of the method you are using".

The detector is a bit more difficult; the photo shows the head of my home-made AD8307 meter, based on Bob Kopski's fine design which is on the 'net for download, I think from ARRL.

I built it years ago and it has been the most-used and most instructive tool I've ever had; I've "graduated" to spectrum analysers, VNAs etc since then but Old Faithful still gets a lot of use.

I strongly recommend suspending work on crystals and building one of these meters; digital readout preferably although I really like my digital and analogue readouts, each has its use.

Otherwise, perhaps a simple diode detector would work; there is probably enough output from a 50-ohm fixture fed with -30dBm.

If you use the "3dB attenuator" method of finding the half-power points the detector need not be calibrated or even accurate as you will just be looking for the same reading at three different frequencies.

A receiver, with AGC disabled and an audio voltmeter on the output, can be used as a sensitive detector but this requires modulation of the generator which you don't have. This is a venerable technique; my HP 415E SWR meter works on this principle. :)

I don't think a 'scope would be sensitive enough; it might be if you put a lot of gain in front of it.

(Yes; that works. I just tried it. With a 43dB amplifier, -80dBm @ 10 MHz almost fills the 'scope screen at 2 mV/div).

If you want to know all the parameters you will need more testgear; it depends, I suppose, on your aims & desires.

Take filters; many builders assemble a filter from unknown crystals using a schematic from a book or the 'net and are very happy with the results from their homebrew receiver. Some, like me, want to take their level of knowledge as far as it can go and invest $$$s, time and thought into the project.

Many others (the sensible ones perhaps :) ) are at some point between these two extremes.

Philosophy mode OFF!

The DJ6EV program seems good; I've played with it but not used it in anger (I actually haven't built a filter for a few years).

The only downside for me is that it doesn't do transitional Gaussian types; most of my interest is in test & measurement (I'm not an active amateur although I'd like to build a good receiver some day) and the Gaussian types are well-suited to measurement equipment.

Re your other question; I think it's possible to estimate relative Q from the amplitude of the oscillation but it's a rough method at best and doesn't tell anything about the value of Q.

It's not a technique that I would use.

Wes has said that, initially, he placed a high value on good Q but that, later, he's less-inclined to see high Q as vital.

I certainly like Q as high as I can get it; it affects filter loss and Q is also a term in the equation for filter end resistance.

I know that Wes' Ladpac will reject a proposed termination resistance on the grounds of low Q; the proposed figure has to be raised until the program accepts it.

As Bill said, cheap crystals tend to have low Q and there are "rogues" in many batches; this, perhaps, makes Q-mesurement more important than it once was.

I think that, in the pre-microprocessor days, crystals were initially made as "good" as the maker could make them as a matter of routine. Later it was understood that different grades could be made for different applications, many of which did not need highly accurate & stable crystals or expensive enclosures.

The microprocessor created a need for millions (billions?) of low-grade crystals; this market became the major one and the market for very high grades was & is much smaller proportionally.

I have quite a few batches that I regard as useless for my purposes; the trend becomes clear after about ten are tested.

I have occasionally found good batches; the best batch I can recall showed very consistent F and Q in the range of about 150 000 to 200 000. I could only get those in 6 MHz; that sets my IF choice!

Wes mentions (rather casually) in EMRFD using crystals with Q of 240 000; I dream of those!

I hope this uncontrolled rant is of some use to you.

Kerry VK2TIL.
8905 2013-08-07 15:45:52 Bob L Re: Software for the Novice Crystal Filter Designer
Thank you.

That is a great summary. It answers my questions.

8919 2013-08-08 12:37:07 Wes Re: Software for the Novice Crystal Filter Designer
Hi Bill, et al,

While I've used the G3UUR method extensively, I have not paid much attention to the operating level of the oscillator. One might be able to get some indication of the unloaded Q by looking at the amplitudes, but it would have to be calibrated. There are other direct ways to get at the crystal Qu.

The method of EMRFD Fig 7.70 works well in the determination of crystal Q. I've found good correlation between the Q results obtained with that scheme and measurements with a n2pk type VNA and Paul's program "xtal2.exe".

You mentioned construction of a simple crystal filter as a method of Q measurement. I've done this. Just build a filter with a single crystal. Use 50 Ohm terminations on each side of the crystal and 1000 pF shunt capacitors. I did the measurements at 10 MHz. Measure Lm with a G3UUR set up or a vna and then drop the crystal into the simple filter. The insertion loss is measured. You can then model the filter in GPLA and tune on crystal Q in the program to get the same loss value.

This single element filter method is more sensitive if you transform to lower terminating impedance values. In the extreme, you end up with a crystal test fixture like the one that DeMaw and others did back "in the day."

Have fun.

73, Wes

8921 2013-08-08 15:02:23 Bob L Re: Software for the Novice Crystal Filter Designer
Is this different from what Todd is doing at the bottom of this page?


8922 2013-08-08 15:06:05 Wes Re: Software for the Novice Crystal Filter Designer
No, same ole thing.


8926 2013-08-08 21:18:40 Kerry Re: Software for the Novice Crystal Filter Designer
To add a couple of things;

Following Wes' example of "try it and see" in the Rigol topic I took a batch of crystals that I had measured some time ago, selected the one with the lowest Q, the one with the highest Q and three at random.

I inserted each into my G3UUR tester and recorded the amplitude (RMS) of oscillation with my HP digital 'scope.

The results (expressed as Q / RMS Amplitude in mV);

106,578 / 372; 101,665 / 340; 91,231 / 354; 82,499 / 335; 72,379 / 325.

There is some correlation, although the second crystal does not follow the trend. Perhaps my Q measurement is in error.

I don't know of any way that the crystal Q could be calculated from the oscillator output voltage; all this test would tell us is that one crystal probably has better Q than another, it won't give us a value of Q for either crystal.

This test is, of course, not comprehensive; I should check each of the batch (60 I think) for a meaningful study, but I'm afraid that I don't have the time or the "need-to-know".

Another topic; literature.

There are several good papers on crystals available for download; Wes' paper and Jack Smith's paper are notable. Jack Hardcastle's papers are also worthwhile.

I recently stumbled on a very good paper; it's rather "buried" but go to


and put "simplified methods for measuring quartz crystal parameters" into the Search box.

The first paper of the four results is an excellent one both on measurements and on the construction of basic filters.

In particular I like the test fixture; it combines a couple of different tests, including Hardcastle's phase-shift test, into a simple fixture with a low-impedance environment.

The end result is that all parameters can be measured or calculated.

If I were not quite spoiled by my VNAs I would build a version of that fixture.

Kerry VK2TIL.
8932 2013-08-09 07:01:32 dk7jb Re: Software for the Novice Crystal Filter Designer
*** As a good learning tool I recommend The Dishal Crystal Ladder Filter program, by DJ6EV. It is available here, as zip files:

All files by Horst dj6ev on filters you can find on my website:
But most texts are in German.

I always find the parameters with the VNWA. He has already built in a quartz tester tool.

Should you be interested in my recent crystal filters, you can find them here:
(Band 2: Quarzfiltermodul und Vorverstärker-Mischer-ZF Modul)
My texts are in German. The traces are of course still readable.

By chance I came across your group. Maybe you can start something with my information.

Joern, DK7JB
8933 2013-08-09 07:39:00 Nick Kennedy Re: Software for the Novice Crystal Filter Designer
I did a little presentation / overview on crystal characterization and
filter design tools to a QRP group five years ago. I made a web page
mainly to provide links to some documents.


You can't get too in-depth in a 50 minute talk, but the slide show and PDF
of notes do include a good list of references on the subject.

In my own world of QRP & homebrewing, my introduction to crystal
measurements came via Jim Kortge K8IQY and some articles he did in the
NJQRP and AmQRP (I think) Homebrewer quarterlies. He built on earlier work
by Wes Hayward. Jim provided some nice circuits including a precision VXO
you could build using a couple of crystals from your test set. It provided
the signal source for testing. Also included was a test fixture for making
the measurements.

The detector can be a simple diode detector, but I use my W7ZOI / W7PUA log
power meter. I sure wish I had one of those N2PK VNAs. Better put it back
on my list.


Nick, WA5BDU

[Non-text portions of this message have been removed]
8934 2013-08-09 10:05:57 Bob L Re: Software for the Novice Crystal Filter Designer
Let me expand on my last post.

I start with the G3UUR oscillator.

I remove the series cap and replace it with an inductor and variable cap, turning it into a VXO. I insert a crystal and tune the oscillator to peak output. I remove the crystal and measure the loss. Using GPLA, I estimate the crystal Q. This is the process just described by Wes.

I now remove the inductor and variable cap and replace the original series capacitor. Its back to its G3UUR oscillator configurati
8935 2013-08-09 10:06:40 Bob L Re: Software for the Novice Crystal Filter Designer
I've discovered that I need a clean sine wave for crystal evaluation. Can you swing a VXO made from the filter crystals to perform the insertion loss test or 3db tests we have been discussing?

Could it be an oxymor