![]() |
![]() |
![]() |
|
Components and Circuits For discussions about component types, alternatives and availability, circuit configurations and modifications etc. Discussions here should be of a general nature and not about specific sets. |
![]() |
|
Thread Tools |
![]() |
#21 |
Moderator
Join Date: Mar 2012
Location: Fife, Scotland, UK.
Posts: 22,901
|
![]()
The industry standard test fixture was made by Cathodeon and looked like a set of three piano keys. Two opened clamp bars for the crystal pins, and the third acted as a shorting switch for a 30pf series capacitor.
Electrically it was a pair of minimum loss pads converting 50Ohms of the outside world to about 12 Ohms applied to the crystal/capacitor. The rig was used in a 2-port transfer function measurement with a synthesiser and a vector voltmeter, or with a VNA. Someone in the network analyser division published a paper on using simpler 1-port measurements with s11 on a VNA. Their aim was to flog VNAs to people using old vector voltmeters. As a result of this many Cathodeon fixtures were stripped out and rewired to become posh crystal sockets. Another paper by Neil Wotherspoon in the standards labs analysed sources of errors in the 1-port versus two port approaches. It remained internal, never published outside. It compared results with national standards labs and those of crystal manufacturers. The conclusion was that you were better sticking with the 2-port approach. We bought a new Cathodeon fixture to replace the one which had been gutted. Quartz crystals are extremely high Q and can resolve and PM-to-AM the phase noise of the source in your VNA or synthesiser. The best unit we found was the 3577A 200MHz VNA or the 8663 family of low noise synths. This was all pushing things to state-of-the-art limits, but you can learn surprising things while doing so, and learn a few things which can be avoided in day-to-day work without adding to the work. David
__________________
Can't afford the volcanic island yet, but the plans for my monorail and the goons' uniforms are done |
![]() |
![]() |
#22 |
Dekatron
Join Date: Sep 2010
Location: Cheltenham, Gloucestershire, UK.
Posts: 3,077
|
![]()
I think I can replicate some of the above here at home because I have an old and working HP8405A VVM here and my little Marconi 2024 sig gen should have similar close in phase noise (at CFs of around 10MHz) as the HP8663.
I can't easily check/verify phase noise below offsets of about 10Hz here on my old Tek RTSA but I'm only really doing casual measurements here anyway. I think if I can predict the resonant frequency within 50-100Hz on the VNA and measure the spreads across several crystals it will be good enough for this thread I think. I earlier tried both 1 port and 2 port VNA tests on the 10MHz crystal and had agreement within a few Hz. The close in to far out phase noise response on the E5071 VNA is best described as 'strange' and when I first got mine a couple of years ago I thought it was faulty. But I checked several E5071s at work and they were the same. I don't think it has caused any issues for me yet and I'm not trying for ultimate accuracy in these tests. I'd have to read up on how to measure a crystal with the HP8405A but my first thought would be that it would have to be a labour of love to do it across a span of maybe 50kHz to get the motional LCR of a typical crystal. I'm not sure I'd want to go there but maybe as a means to just verify where resonance occurs with a 30pF test cap it would be worth a try. But I feel I'm heading into nerdy territory here where the added value won't be worth the effort involved. I'm not sure it matters much how accurate these old 4.032MHz xtals are and they are only a small sample across countless other crystal types and specs. So the tests wouldn't prove anything conclusive (about crystal ageing) no matter what I do to vary the test methods ![]()
__________________
Regards, Jeremy G0HZU |
![]() |
![]() |
#23 |
Dekatron
Join Date: Sep 2010
Location: Cheltenham, Gloucestershire, UK.
Posts: 3,077
|
![]()
I just did a quick google and found this:
http://www.k0to.us/HAM/Articles/Appl...te%20171-1.pdf I think I can replicate this setup here at home so it might be fun to try it and compare to the 2 port method on a VNA. When I tested with the VNA I used both a mechanical set of standards and also a $$$ Ecal module I have here and this lets me export a 2 port model of the crystal to an RF simulator on a PC. This method will have various contributors to overall uncertainty but it is very powerful. I got excellent agreement with the mechanical calibration and with the snazzy 4 port Ecal module that came with my VNA. But the VNA method isn't going to be a perfect test method by any means. Hope this isn't taking things too far off topic and I hope it is of interest. I need to read the HP article in some detail so I'm off to get some biscuits and a can of coke. ![]()
__________________
Regards, Jeremy G0HZU |
![]() |
![]() |
#24 |
Dekatron
Join Date: Sep 2010
Location: Cheltenham, Gloucestershire, UK.
Posts: 3,077
|
![]()
If anyone is interested here's a quick screenshot showing an earlier attempt to provide an equivalent LCR model of the 4.032MHz crystal. This has guesstimates for the motional inductance and capacitance etc. The plots below show how closely my LCR model agrees with the VNA measurement of a real crystal. There are actually four traces on the graph but they overlay very closely indeed and I think the LCR model is quite a close match. The graph is for amplitude and phase response through the crystal and also the equivalent LCR based model.
In the second plot below I've deliberately offset the LCR model slightly so you can see that there are really four traces in the first plot. Very good agreement indeed I think!
__________________
Regards, Jeremy G0HZU |
![]() |
![]() |
#25 |
Dekatron
Join Date: Jun 2016
Location: London, UK.
Posts: 3,687
|
![]()
Well that adds another one of the already 800 reasons on my list to "buy an HP 8640". Now I've got an excuse to get a vector voltmeter to go with it.
|
![]() |
![]() |
#26 |
Dekatron
Join Date: Sep 2010
Location: Cheltenham, Gloucestershire, UK.
Posts: 3,077
|
![]()
I've read the HP article now and it looks like a really neat way to measure for series resonance. I think I can do this for the raw crystal and also with an accurate 30pF load capacitor. I need to make the load cap and I'll use a couple of ATC800B 15pF caps in parallel and check the capacitance on the VNA. I've checked/tested that my Mi2024 sig gen can do external FM with DC coupling and the very close in phase noise of the Marconi Mi2024 sig gen should be much better than the HP8640B at offsets of a few Hz at these test frequencies. So I think I'm good to go...
My old HP8405A VVM is now on the bench warming up and I'll leave it an hour or so to fully settle. I also need to make the RC loop filter. I'm tempted to make this in an inline BNC enclosure and I've got one somewhere... This old HP app note deserves a Youtube video to celebrate the elegance of the method but sadly I don't have the presentation skills or the video making skills to do it justice. But I'll post up my findings on here over the next day or so ![]()
__________________
Regards, Jeremy G0HZU |
![]() |
![]() |
#27 |
Moderator
Join Date: Mar 2012
Location: Fife, Scotland, UK.
Posts: 22,901
|
![]()
Yeah, you just drop a resistor into your fixture in place of the crystal to give you a phase=0 reference. Fiddle with the resistor value to get the same voltage ratio on the two VVM channels and your resistor value equals the crystal ESR, and you have the best zero phase reference.
The Cathodeon fixture with its low-Z environment between its minimum loss pads is rather good in this area. Find the SER this way, and find the series res frequencies with and without the 30pf load C in circuit and you have enough to extract motional parameters. There are some superb apps notes by Cathodeon, but I've never seen them on the web. David
__________________
Can't afford the volcanic island yet, but the plans for my monorail and the goons' uniforms are done |
![]() |
![]() |
#28 |
Dekatron
Join Date: Sep 2010
Location: Cheltenham, Gloucestershire, UK.
Posts: 3,077
|
![]()
I had my first go with this and things are looking good and the loop locked up first time! I'm still optimising the setup but the first crude attempt showed the resonance of the crystal on its own (without the 30pF loading)was 4.031720MHz.
The same crystal was tested yesterday (similar room temperature?) on the VNA and the exported model shows resonance at 4.031723MHz as per the image below. This does not have the 30pF loading cap in series and this explains why it is not at the expected 4.032MHz as per the crystal frequency stamped on the crystal. But it agrees very closely with the HP app note method. Just a few Hz difference! What a result! On the simulator I can add a perfect 30pF capacitor and measure the resonance with the recommended loading. I've yet to make the 'real' 30pF test cap for the HP8405A and I realise now that this cap needs to be decent in terms of accuracy. It looks like I'm chasing down to tiny frequency differences between the methods now... I'll have more of a play and then try testing all of the 4.032MHz crystals on the HP8405A and I'll post up the results on here. This method is the fastest/easiest because I just look at the readout on my old Anritsu frequency counter to see the resonant frequency. This is a seriously cool test method!
__________________
Regards, Jeremy G0HZU Last edited by G0HZU_JMR; 15th Feb 2018 at 5:33 pm. |
![]() |
![]() |
#29 | |
Dekatron
Join Date: Sep 2010
Location: Cheltenham, Gloucestershire, UK.
Posts: 3,077
|
![]() Quote:
__________________
Regards, Jeremy G0HZU |
|
![]() |
![]() |
#30 |
Dekatron
Join Date: Sep 2010
Location: Cheltenham, Gloucestershire, UK.
Posts: 3,077
|
![]()
I quickly tested each 4.032MHz crystal in a makeshift socket in the test fixture and on the HP8405A four of them were within 25Hz of the first crystal.
But two were at about -80Hz and one was +80Hz. This is just outside of the manufacturer's original/initial 15ppm calibration spec for this crystal when it was made 20 years ago. Note that I'm just measuring relative differences at the self resonant frequency. Next I need to try them with my first attempt at an accurate 30pF load capacitor to see how close they are to 4.032MHz.
__________________
Regards, Jeremy G0HZU |
![]() |
![]() |
#31 | |
Dekatron
Join Date: Apr 2012
Location: Wiltshire, UK.
Posts: 14,007
|
![]() Quote:
From the job-lot I was able to build half a dozen rather good SSB-bandwidth 8-crystal ladder filters' and point a CW-oriented friend in the right direction to use up the remaining rocks to build a filter or five for his particular predilection. The biggest problem with building 'ladder' filters is that they need to work with consistent source- and sink-impedances, where 'consistent' extends from a few hundred KHz to around 5x the frequency of the crystals themselves. But Wahey! done properly [and with good screening/shielding to prevent IF 'hop-over'] by hand-picking the crystals you could get a great 6:80dB shape-factor - much better than the at-the-time amateur 'gold-standard' KVG XF9- series. |
|
![]() |
![]() |
#32 |
Dekatron
Join Date: Sep 2010
Location: Cheltenham, Gloucestershire, UK.
Posts: 3,077
|
![]()
Yes, it's interesting to see the spread in the frequency across a batch of crystals.
I managed to find a sample bag of 30pF ATC800B caps and the most accurate one on my VNA measured 29.8pF across HF and into VHF. I must admit I was more than a bit excited when I connected it inline with my lashup test fixture. By changing the FM range on the Mi2024 sig gen it's possible to get a very precise calibration for zero phase on the +/- 6deg range on the HP8405A so it's worth playing with the FM setting on the sig gen to get best results. I also ended up reducing the loop filter cap to 1000uF to speed it up a bit. I tried the original crystal on the HP8405A test setup and it measured 4.032015MHz with this 30pF loading. On the VNA yesterday the model predicted 4.032025MHz for the same crystal with a perfect 30pF capacitor so the VNA/simulator was within 10Hz and this wasn't bad!
__________________
Regards, Jeremy G0HZU |
![]() |
![]() |
#33 |
Dekatron
Join Date: Sep 2010
Location: Cheltenham, Gloucestershire, UK.
Posts: 3,077
|
![]()
I tested them all on the HP8405A VVM with the 30pF loading cap and the spread in frequency now seems a bit less.
The highest measured 4.032060MHz (+60Hz) The lowest two measured about 4.031960MHz (-40Hz) The others were within 15Hz of 4.032000MHz. So even the worst one was just about in spec for IQD's 'initial' 15ppm calibration spec (with 30pF loading) and this is after 20 years sitting in a sample bag in a cupboard here ![]() Note that I had to steal the rubber gloves from the kitchen when swapping between the crystals because the heat from my bare fingertips would cause several Hz drift and this would take a while to settle. I tried tweezers and mini grips but this was too fiddly. So I had to resort to donning the Marigolds ![]()
__________________
Regards, Jeremy G0HZU Last edited by G0HZU_JMR; 15th Feb 2018 at 8:30 pm. |
![]() |
![]() |
#34 |
Dekatron
Join Date: Sep 2010
Location: Cheltenham, Gloucestershire, UK.
Posts: 3,077
|
![]()
For completeness, I changed the frequency on the Mi2024 to 10MHz and tried the IQD 10MHz crystal that I tested on the VNA yesterday.
On the HP8405A setup (after a fresh zero calibration) the counter read 9.999975MHz with the 30pF load cap. On the simulator with a perfect 29.8pF cap the VNA/simulator predicts resonance at 9.999960MHz so it was within 15Hz and the room temp feels about the same as yesterday. See also below for a comparison of my LCR model of the crystal compared to the real 10MHz crystal on the VNA. You can see that the traces match almost perfectly so the lumped model is very good here too. I generate this model using the simulator and with an iterative approach the simulator can optimise the model values for motional L, C and R to fit the VNA data for the real crystal.
__________________
Regards, Jeremy G0HZU |
![]() |
![]() |
#35 |
Dekatron
Join Date: Sep 2010
Location: Cheltenham, Gloucestershire, UK.
Posts: 3,077
|
![]()
To do a better comparison I spent a bit of time last night comparing the VNA against the HP8405A method and when tested side by side using the same test fixture with no body heating effects and the same ATC800B 30pF loading cap the two methods usually agreed within about 1Hz. This was measured with several crystals several times going back and forth. I did have to zoom in to a narrow span on the VNA to get this resolution and I used lots of points on the display and a narrow RBW.
Just touching the fixture near the crystal with my fingers for a few seconds would cause a slow drift in the crystal so I had to use gloves and insulated tools to swap between the two setups to do a decent comparison. As the manufacturer's initial calibration spec for a typical crystal is 10-20ppm then these differences are negligible I think.
__________________
Regards, Jeremy G0HZU |
![]() |
![]() |
#36 | |
Rest in Peace
Join Date: Sep 2008
Location: Solihull, West Midlands, UK.
Posts: 4,872
|
![]() Quote:
|
|
![]() |
![]() |
#37 |
Dekatron
Join Date: Sep 2010
Location: Cheltenham, Gloucestershire, UK.
Posts: 3,077
|
![]()
Thanks. What is really neat about the old HP method with the HP8405A VVM is that you can sit there with a box of random crystals and tap in the design frequency of the crystal to the sig gen and then test it very accurately for resonance with/without a 30pF loading cap.
I'm using a shorting bar to speed it up even more and it literally takes a few seconds to do the whole test. eg type in 48.000MHz to the sig gen for a 48MHz series resonant overtone crystal. I think that the 30pF loading cap is not needed in this test if the crystal is marked as series resonant. Then press the shorting bar and calibrate for zero phase (this also means the counter shows 48.000000MHz because the sig gen has 0V DC at the FM input) Then release the shorting bar and with the crystal in place simply read the series resonant frequency on the counter. I've got a bag of old 48MHz OT xtals from Golledge here and they are dated Nov 2007. There are still 7 in the bag but the first one out of the bag read 47.999793MHz in the above test. So just over 200Hz low. Not bad at 48MHz after 10 years! That's within about 4ppm.
__________________
Regards, Jeremy G0HZU |
![]() |
![]() |
#38 |
Dekatron
Join Date: Sep 2010
Location: Cheltenham, Gloucestershire, UK.
Posts: 3,077
|
![]()
I measured two more of them and got 47.999612MHz and 47.999745MHz so they are quite consistent. One thing you have to watch is that the FM setting on the sig gen needs to be carefully set to keep the HP8405A within the linear region of its phase detector on the selected phase display range, i.e. to keep close to zero. I had to change the FM setting on the sig gen quite a bit up at 48MHz.
One way to check the validity of the test is to turn off the FM modulation and close the shorting bar and adjust the VVM for zero phase at 48MHz. Then release the shorting bar and manually adjust the sig gen in 1Hz steps until zero phase is achieved again on the HP8405A. The frequency should then be the series resonant frequency. Then repeat the shorting bar zero cal at this frequency and double check that the phase is still at zero with or without the shorting bar. This proves the phase is zero for both the crystal and the shorting bar at this frequency and so it must be the series resonant frequency of the crystal. I didn't try using a resistor to emulate the crystal ESR here, I just used the shorting bar. But I got good agreement using the automatic FM/loop method and the clunky manual method. So I think it is all good ![]()
__________________
Regards, Jeremy G0HZU Last edited by G0HZU_JMR; 16th Feb 2018 at 2:17 pm. |
![]() |
![]() |
#39 |
Dekatron
Join Date: Sep 2010
Location: Cheltenham, Gloucestershire, UK.
Posts: 3,077
|
![]()
If anyone does read this and tries to repeat the procedure in the HP app note then it's worth emphasising that in section 4 of the 'series resonance' procedure you do have to manually adjust the sig gen mod/freq carefully to keep the phase meter near zero when on the fine (+/- 6deg) range on the HP8405A. Otherwise errors can creep in. I spent some time exploring these errors and I now think it's worth me pointing this out.
Back in those days I suspect that HP were used to testing crystals for use in their classic $$$ OCXOs and they were probably interested in measuring very subtle differences between crystals. Maybe they used a setup like this or maybe they had an even better one but having tried the first procedure in the HP app note I suspect that most users would be better/safer just doing the manual approach without the FM loop connection. It's slower and so you would get less throughput if testing loads of crystals but most of us would only be measuring a handful of crystals at a time. You wouldn't need a sig gen with a decent external FM/DC input either. However, if you had a boxful of crystals like G6Tanuki had then the FM/loop method would be the best, but care would be needed and the phase meter would need to be monitored (and the sig gen adjusted as per section 4) to keep the phase meter close to zero if you wanted to measure things without introducing some subtle errors. You can then test the crystals very quickly once the setup is proven with a manual cross check.
__________________
Regards, Jeremy G0HZU Last edited by G0HZU_JMR; 17th Feb 2018 at 3:10 pm. |
![]() |
![]() |
#40 |
Dekatron
Join Date: Sep 2010
Location: Cheltenham, Gloucestershire, UK.
Posts: 3,077
|
![]()
I also managed to find a regular 16MHz crystal and did a few comparisons with the Golledge 48MHz OT crystal on the VNA. See the word doc below for a few swept plots at 16MHz, 48MHz and 80MHz.
There are definite differences that indicate that the Golledge 48MHz OT crystal is more suited for overtone use. The ESR and response is better at 48MHz for one thing. Note that the Gollege 48MHz OT crystal was the 'best of the bunch' in that the series resonance on this one was very close to 48MHz so it looks almost spot on in the 48MHz plot. I also took accurate 2 port models with the VNA for each crystal at each of these test frequencies. This allows the ESR and motional parameters to be derived from this data at each frequency. Also, these s2p VNA models can be imported to a simulator and married up to an oscillator circuit and the oscillation frequency (and potential VCXO tuning range where oscillation is possible) can be predicted with good results.
__________________
Regards, Jeremy G0HZU Last edited by G0HZU_JMR; 17th Feb 2018 at 3:39 pm. |
![]() |