Fun with crystal filters

[joeqsmith]

I am no math wiz, for sure...   What is Z0,  is that the reference resistor used during the test?

Edit:  This slide from Agilent seems to show 3 methods of measuring impedance as well - are they the same as the Copper Mountain ones?

Where Z0 is characteristic impedance of a transmission line (commonly 50 Ohm);

Yes indeed, these are the only three I am aware of.  What makes this setup unique is while it's a series thru arrangement, it has these added impedance matching transformers that have to be accounted for.   

The original article from Copper Mountain:
https://coppermountaintech.com/measurement-of-electronic-component-impedance-using-a-vector-network-analyzer/

[SilverSolder]

The Agilent document goes through a couple of variations on those 3 themes, basically varying the input impedance of the ports on the VNA (1Meg as well as 50 ohm options).  Interesting to see how it compares.

I don't understand the purpose of the common mode transformer in the Copper Mountain setup...  looks like a way to prevent ground loops?  Why would that be necessary, since everything is connected to the same device?

Attached Agilent document for reference.

[joeqsmith]

The Agilent document goes through a couple of variations on those 3 themes, basically varying the input impedance of the ports on the VNA (1Meg as well as 50 ohm options).  Interesting to see how it compares.

I don't understand the purpose of the common mode transformer in the Copper Mountain setup...  looks like a way to prevent ground loops?  Why would that be necessary, since everything is connected to the same device?

Attached Agilent document for reference.

Yes, they are trying to show you why you may want to use one method over another.  It will depend what you are trying to measure.   This is what the Copper Mountain document is also trying to convey.   

I am not sure what Copper Mountain setup you are referring to that is using a common mode transformer.   

Assuming you are actually asking why I am using the two impedance matching transformers with my setup compared with the resistive pads I started out with, then the answer is simple.  The pads have too much loss to get reliable results with the Nano, depending on the crystal being measured.   The reason that the resistive pads work fine with my older HP systems is that they have a much lower noise floor than the Nano.   If you look at the graphs I've provided, the first fixture has a loss of 30dBish compared with 6 with the new fixture.   When you are playing with a $50 instrument, you need to give it a little help.   IMO, this isn't a big deal.  It's $50.  The fact I am able to pull out sub PPM data with it is impressive.   

When I made the videos for the Nano, I had played with a couple of crystals.  One was measured using S11 (shunt) where I provided a simple example of making a model for  SPICE.   The other, I had one in series showing the resonance.    The first thing I did was try to use the crystal directly connected and the standard deviation was very poor.   The papers I attached early on are what led me to use the pad to try and improve the match.     

I think if you go back and read these papers, you will find I am not doing anything unusual or novel.   What I have shown is we have a $50 VNA that can directly making ALL of the required measurements using a single fixture.   


Still no work about the reference crystal. 

[SilverSolder]

The Agilent document goes through a couple of variations on those 3 themes, basically varying the input impedance of the ports on the VNA (1Meg as well as 50 ohm options).  Interesting to see how it compares.

I don't understand the purpose of the common mode transformer in the Copper Mountain setup...  looks like a way to prevent ground loops?  Why would that be necessary, since everything is connected to the same device?

Attached Agilent document for reference.

Yes, they are trying to show you why you may want to use one method over another.  It will depend what you are trying to measure.   This is what the Copper Mountain document is also trying to convey.   

I am not sure what Copper Mountain setup you are referring to that is using a common mode transformer.   

Assuming you are actually asking why I am using the two impedance matching transformers with my setup compared with the resistive pads I started out with, then the answer is simple.  The pads have too much loss to get reliable results with the Nano, depending on the crystal being measured.   The reason that the resistive pads work fine with my older HP systems is that they have a much lower noise floor than the Nano.   If you look at the graphs I've provided, the first fixture has a loss of 30dBish compared with 6 with the new fixture.   When you are playing with a $50 instrument, you need to give it a little help.   IMO, this isn't a big deal.  It's $50.  The fact I am able to pull out sub PPM data with it is impressive.   

When I made the videos for the Nano, I had played with a couple of crystals.  One was measured using S11 (shunt) where I provided a simple example of making a model for  SPICE.   The other, I had one in series showing the resonance.    The first thing I did was try to use the crystal directly connected and the standard deviation was very poor.   The papers I attached early on are what led me to use the pad to try and improve the match.     

I think if you go back and read these papers, you will find I am not doing anything unusual or novel.   What I have shown is we have a $50 VNA that can directly making ALL of the required measurements using a single fixture.   


Still no work about the reference crystal.

Ah, I actually found a different Copper Mountain article from the one you referenced...   attached.    Here, they use a Common Mode transformer with their VNA to get better performance.  Not sure I understand how...

[joeqsmith]

Ah, I actually found a different Copper Mountain article from the one you referenced...   attached.    Here, they use a Common Mode transformer with their VNA to get better performance.  Not sure I understand how...

I'm not understanding why it would matter.  Crystals filters don't normally have sub ohm resistances.
**
Just to be clear, I am not sure why there would be a need to read such low values when working with filters.   I don't doubt the need to control the paths when measuring this low.   

[joeqsmith]

If we were working in the 900MHz (or even 300MHz) with the Nano, and trying to look at filters with having a high roll off, it may not work out so well.   
 
Attached are a few screen shots of a nine section 1GHz interdigial filter with a 3dB BW of around 130MHz.    A long way from the 12MHz that the OP asked for. 

****
Added narrow sweep for Nano

[paul002]

I also use the nanovna to measure the crystals and DISHAL to design and adjust the crystal parameters. It takes some practice to really understand all the ins and outs. Also tried to use quandl to simulate the filter but somehow the parameters does not match. I use G3UUR methode to measure the crystals using 12 ohm impedance transformers and the nanovna bandpass filter option to calculate -3db points. There is also some crystal measurement software on github. But did not work for me. But for now I am happy with the result using the nanovna.

[joeqsmith]

Their software looks like it is coming along nicely.   

Why do you need these 12 ohm impedance transformers for the G3UUR method?  Everything I saw on it, they just used a basic oscillator.  Maybe scribble a picture of your circuit. 

[paul002]

Ah I see, I don't use the G3UUR, I use the method described by WA5BDU in crystal_slide_show.pdf (slide 10 and 11). And in DISHAL use the xtal tuning function to calculate correct values. And use Nanovna-saver software to determine the -3db points

[joeqsmith]

Makes more sense.  I am using something similar for the fixture. 

How did you validate your results?   Did you use the G3UUR at one point to correllate with the Nano?   Did you measure your Nano's frequency accuracy? 

I'm curious how your setup compares with what I have shown so far.     I never did hear back about the standard reference crystal.   Outside of getting an email stating that they received my request, there have been no other communications.   

[paul002]

I also used the G3UUR method but not with a lot of success, measurements were not precise enough, but also it takes to much time. Using the Nanovna is much faster. I remeasured some crystals with my signal generator which is build in in the scope, but it is not precise enough. But I assumed that if the nanovna has a frequency error, it will make the error on all crystals. Unfortunately I don't have a GPS reference to tune the NanoVna. So yes there is a frequency error. When building the crystal filter I used a bunch of variable air capacitor to be able to adjust very easy the capacitors. After a while you get used to how the capacitors influence the shape of the filter. It is not difficult the get the passband shape, but you have to tune quite well to get a acceptable swr on the crystal filter. I think if you are able to tune the nanovna to an exact frequency this problem can also be solved. It is possible to have external clock on the si5351 maybe that is a solution. So it is science for a part but still a large part experimenting.

[joeqsmith]

I have no way of knowing which method is more accurate when I tried them.  They both could be off a fair amount and I have no way of knowing.  Seeing 2% between the two gives me some confidence in the measurement.

Speed wise, for me it's a wash.  The majority of the time is waiting for the parts to settle.   If I were to measure several parts, I think I would make some sort of holder for them so I could pick them up easily with some plastic tweezers and make some sort of ZIF socket for the fixture.   I thought about adding a small thermal controlled block that would contact the case to at least remove the effects of the room temperature.

With both setups, I have to visually watch them settle.  A better approach would be to automate that part as well.   

If temperature wasn't a problem, the Nano may have a leg up speed wise.   If I were to automate the C0 measurement, it may be a bit faster to use the frequency.   That Nano just crawls along.    So far, using the fitter has worked well but again, no idea on accuracy.    I'll take slow and tight stddev over fast inaccurate results any day. 

[joeqsmith]

After all these months, I have never heard back from Saunder's about the standard they offer.   However, at this point I am much more confident in the data I am collecting. 

Shown is the last filter I constructed for the fun of it. 

With the new low cost VNA having much better update rates (possibly 400 - 500 sample points per second),  I would like to pick one up and try it with some of these filters.  With it being this fast, it would make adjusting some of the mechanical trimmers so much nicer. 

[Grandchuck]

Nice Joe!

Are you going to share the frequency response curve?

[joeqsmith]

I am still thinking to make a video about the whole process which is why I built this particular filter.   

It could end up being more of a VNA shootout of my older higher end equipment vs the ultra low cost modern systems flooding the market.  I would like to show some of the new ones that OWO is involved with.   The speeds they are sweeping at exceed what I can do with my 1970's HP and some of the other claims seem really impressive.   

After playing around a bit above 1.5GHz,  I could maybe do a decent job comparing them at higher frequencies.