Topics

Crystal parameters

Jack D. Generaux
 

Nick and others:


I am building Scotty's spectrum analyzer and working on a final filter design.  I used the PHSNA crystal routines to measure a batch of 11+ MHz crystals and get Lm values of around 4 mH using the newest crystal fixture with the ERA installed.  Cm is about 50 fF, Rs about 2, and Q of about 145000.  I measured the holder capacitance at 4 pF.


I have been trying three programs to compare and contrast designs: XLAD, AADE, and Dishal.   In the Dishal Program, its gives me a warning that these values don't appear reasonable. When I look at the reference material for the Dishal program,  I see Lm values a factor of 10 higher (around 70), with corresponding lower Cm.  I know that program mentions the G3UUR frequency routines, but also the Hayward methods.  I am confused over this warning.  Any ideas? 


Also Hayward mentions he was happy with a Gaussian topology for a spectrum analyzer.  For the final filter, I would like minimal ripple in the passband and reasonable rejection.  Is the Gaussian topology the best approach? 


I appreciate the help,

73

Jack (W0FNQ)

Nick Kennedy
 

Hello Jack,

Sounds like a great project. At first glance, I did think the Lm value was kind of low compared with what I've seen. OTOH, when you have millihenries and femtofarads, you're somewhat in the ball park.

I've looked at some crystal measurements I have logged over time, many using pre-PHSNA methods and of course they vary a lot with frequency. The closest I have on the high side was some 13.5 MHz crystals that had Lm in the 4.5 mH area. 

On the lower side of 11 MHz I have a 10 MHz crystal measured with 10 mH and some 9 MHz at 18 mH.

So it's all over the place. Your readings seem a bit off, but not completely out of line.

At times we've tried a few "round robin" measurements where we've mailed some crystals around to different people and compared readings. A good way to see if we're getting consistent results. If you have any extras in the set, I'd be glad to take a look at one or more to see what I get and maybe others would too.  Of course, you'd get them back, but there would be some delay.

Another thing - have you measured other crystals of various frequencies and makers and seen if you get readings similar to what others get?  I've got a lot of readings of 8 MHz, 7 MHz, 5.3 MHz, 4.915 MHz and similar crystals we could compare.

I don't have enough expertise and/or experience in the merits of different filter types to comment on your options. Hayward's word is gold to me. Of course, if you've done the research, you can always make an informed choice based on what you want.

Would like to hear more about the SA. Were you able to buy boards and bag-o-parts or similar? 

73-

Nick, WA5BDU

-- It's hard for me to get used to saying "femtofarads".



On Sun, Mar 13, 2016 at 8:51 AM, 9artus@... [PHSNA] <PHSNA@...> wrote:
 

Nick and others:


I am building Scotty's spectrum analyzer and working on a final filter design.  I used the PHSNA crystal routines to measure a batch of 11+ MHz crystals and get Lm values of around 4 mH using the newest crystal fixture with the ERA installed.  Cm is about 50 fF, Rs about 2, and Q of about 145000.  I measured the holder capacitance at 4 pF.


I have been trying three programs to compare and contrast designs: XLAD, AADE, and Dishal.   In the Dishal Program, its gives me a warning that these values don't appear reasonable. When I look at the reference material for the Dishal program,  I see Lm values a factor of 10 higher (around 70), with corresponding lower Cm.  I know that program mentions the G3UUR frequency routines, but also the Hayward methods.  I am confused over this warning.  Any ideas? 


Also Hayward mentions he was happy with a Gaussian topology for a spectrum analyzer.  For the final filter, I would like minimal ripple in the passband and reasonable rejection.  Is the Gaussian topology the best approach? 


I appreciate the help,

73

Jack (W0FNQ)


Jack D. Generaux
 

Nick,

Thanks for the response.  I do have a couple extra crystals-- I'll send them to you.  Is your address correct on QRZ?  If not, email me at 9artus <<at>> gmail <<dot>> com.  No need to return them they were very inexpensive.  Thanks for the offer.  

The "spectrumanalyzer"  Yahoo group has a lot of into.  Scotty participates in the discussion and lends trouble shooting help.  Another ham and I just sponsored a group buy of the boards and hard to get components (Almost 50 participants and this is the 3rd group buy.)   I think the extra boards are all spoken for, but I can check with Stew.   There is a special MiniCircuits package that sources the major MiniCircuit pieces and that is available directly from MiniCircuits.  I think the project will come in for me at about the $400 mark.  Kind of high, but nothing commercial can touch it that I have seen.  I am building the version with the tracking generator and VNA.   I have my unit up but not calibrated,  It has been a challenging project but great experience.  

I better not get too far off-topic; but glad to continue discussion via email, or on the other group.

73,
Jack (W0FNQ) 

David Collins
 

For what it's worth, the attached spread sheet shows the results of testing and sorting 18 crystals for use in the Measurement Receiver's crystal filter.  Three different PHSNA versions were used.


73,
Dave Collins - AD7JT

dick faust
 

The basic reference authority on filters is "Handbook of Filter Synthesis" by Zverev, Anatolii  In particular Chapter 8 - Crystal filters, you also would be wise to look at the first 3 chapters as well. The book is available for lending at openlibrary .org, you will have to register to borrow the book - probably have a wait as well as it is very popular.  Link below.
Handbook of filter synthesis (Open Library)

 


As to your wish to have minimum ripple, you will have to carefully adjust every capacitor value in the filter to achieve that result.  The basic formulas that you are referencing are adequate for many applications and do provide satisfactory filters for most applications.  If you want the highest performance filter it is best to look at the bridge configuration which is far easier to adjust and can provide steeper slopes, minimum passband ripple and lower insertion loss.  Crystals however, for the Bridge are quite expensive.  In either case the impedance match to the surrounding circuits are critical to insertion loss and ripple and careful attention is required here.

I think Dishal is likely right that the Lm needs to be larger for this filter.  In basic oscillator design the crystal parameters are not as critical or controlled as it is for filters and it is likely that you have crystals manufactured for oscillators.

The standard industry practice for filters use to be to adjust the filters for the required specifications with ceramic trimmer capacitors, using a sweep technique similar to PSHNA hundreds of times to obtain the desired response and then replace with ceramic plate caps [now very difficult to find] that were ground down from a slightly higher value on a Boonton Q meter to the exact same value as the trimmers they they replaced.

Good luck with your project.

Dick K9IVB
 


---In PHSNA@..., <9artus@...> wrote :

Nick and others:


I am building Scotty's spectrum analyzer and working on a final filter design.  I used the PHSNA crystal routines to measure a batch of 11+ MHz crystals and get Lm values of around 4 mH using the newest crystal fixture with the ERA installed.  Cm is about 50 fF, Rs about 2, and Q of about 145000.  I measured the holder capacitance at 4 pF.


I have been trying three programs to compare and contrast designs: XLAD, AADE, and Dishal.   In the Dishal Program, its gives me a warning that these values don't appear reasonable. When I look at the reference material for the Dishal program,  I see Lm values a factor of 10 higher (around 70), with corresponding lower Cm.  I know that program mentions the G3UUR frequency routines, but also the Hayward methods.  I am confused over this warning.  Any ideas? 


Also Hayward mentions he was happy with a Gaussian topology for a spectrum analyzer.  For the final filter, I would like minimal ripple in the passband and reasonable rejection.  Is the Gaussian topology the best approach? 


I appreciate the help,

73

Jack (W0FNQ)

Jack D. Generaux
 

Thank to all for the replies.  I have sent Nick a couple of my crystals to see how his numbers compare. 

Dave, your Lm numbers are quite a bit higher than I am seeing but I am measuring 11 MHz crystals and my Q values are similar.  I notice you used the Midnight Design SNA or terminal -- I am currently using the Midnight Design SNA as well.  I am using the PHSNA Crystal Fixture with the ERA amp.  I put a 10 dB attenuator on the output so that I would not overdrive the crystal.  I am just not sure of the variability in crystals -- I assume it can be very large between different varieties.  

Dick, thanks for the great information -- I will try to read the Zverev chapters.  This is all a great learning challenge -- seems like there is quite a bit of art in this science.  

Jack (W0FNQ)

David Collins
 

HI Jack,

I just used the test fixture shown in Tutorial M5 - Testing & Grading Crystals.  The fixture has no amplifier, just a couple transformers for impedance matching and -3 dB pads on the input and output.  The tutorial also includes the results of testing 30 more crystals for use in the Measurement Receiver plus a plot of the final crystal filter's frequency response.


73,
Dave Collins - AD7JT

Nick Kennedy
 

I received the crystals and measured each twice and took the average although the two readings were close. My results do deviate quite a way from Jack's.

Crystal A Jack Nick
R 2.32 6.5
Cm in fF 49.1 19.959
Lm in mH 4.11 10.36
Q 126519  missed
Fq 11055 11054.7
Crystal B Jack Nick
R 2.54 6.5
Cm in fF 48.7 19.79
Lm in mH 4.26 10.4715
Q 116437 111903
Fq 11055.01 11054.7

Pasted from Excel with alignment a little screwy.

I'm trying to think of things that could throw off accuracy. The parameters entered which are the dBm reading through a shorted fixture and the resistance looking into the fixture (typically 50 or 12.5) are important of course.

I think the dBm reading is supposed to be taken in the normalized mode.

Otherwise I'm not sure what's happening.  Jack said he doesn't need these crystals back. Anyone else who is currently set up to do crystal measurements want to take a shot?

73-

Nick, WA5BDU


On Sun, Mar 13, 2016 at 9:08 PM, 9artus@... [PHSNA] <PHSNA@...> wrote:
 

Thank to all for the replies.  I have sent Nick a couple of my crystals to see how his numbers compare. 


Dave, your Lm numbers are quite a bit higher than I am seeing but I am measuring 11 MHz crystals and my Q values are similar.  I notice you used the Midnight Design SNA or terminal -- I am currently using the Midnight Design SNA as well.  I am using the PHSNA Crystal Fixture with the ERA amp.  I put a 10 dB attenuator on the output so that I would not overdrive the crystal.  I am just not sure of the variability in crystals -- I assume it can be very large between different varieties.  

Dick, thanks for the great information -- I will try to read the Zverev chapters.  This is all a great learning challenge -- seems like there is quite a bit of art in this science.  

Jack (W0FNQ)


David Collins
 

Hi Nick,

I'll take a shot at them.  My address is good on QRZ.com.

73,
Dave Collins - AD7JT

Jack D. Generaux
 

Nick,

Thanks,  

I believe in your numbers more than mine.  Since we last communicated, I put together a 3KC 5 pole based on Dishal's program-- It is not performing well.  I'll run your numbers through the program and see if I can rebuild the filter.  

73
Jack (W0FNQ)

Jack D. Generaux
 

Nick and all,

I found the problem with my new PHSNA fixture -- I had flopped two of the output xformer leads.  When I fixed that,  the numbers from similar crystals look like yours.  I also went back and tested with an old fixture I had -- it agrees with the corrected new fixture.   I have a set of measurements I took with the old fixture and will run them with the Dishal program and see where it all leads.  I am glad I sent you those crystals, made me revisit my setup.  Thanks again.

73,

Jack 

Nick Kennedy
 

That's so easy to do.  Glad you found the problem.  I'm sending the crystals off to Dave and then we'll have three sets of numbers to compare.

Hope the filter works well now.

73-

Nick, WA5BDU

On Thu, Mar 17, 2016 at 7:52 PM, 9artus@... [PHSNA] <PHSNA@...> wrote:
 

Nick and all,


I found the problem with my new PHSNA fixture -- I had flopped two of the output xformer leads.  When I fixed that,  the numbers from similar crystals look like yours.  I also went back and tested with an old fixture I had -- it agrees with the corrected new fixture.   I have a set of measurements I took with the old fixture and will run them with the Dishal program and see where it all leads.  I am glad I sent you those crystals, made me revisit my setup.  Thanks again.

73,

Jack 


Jack D. Generaux
 

Well,

Works much better -- I redid the filter with a 1.65KHz Dishal design.  You can see the results in the attached pic.  Not perfect but what I expected -- BW is a little low 1.4.  Thanks again for the help.

73,
Jack (W0FNQ)