Crystal measurements with the nanoVNA #measurement #applications


hwalker
 

A notification was sent to the group that the file "Crystal measurements with the nanoVNA.pdf" was uploaded to the Files area of the nanovna-users@groups.io group by EB4APL.
=============================================================
Coincidentally, Battosai42 has an open-source project for automated measurement of crystal parameters using either a miniVNA or nanoVNA. The measurement method requires a passive test-fixture according to IEC-444 which is a double-pi network terminating the crystal at 12.5Ohm. The IEC-444 fixture appears to be the same one described in the pdf document uploaded by EB4APL.

A beta distribution of the application is located at https://github.com/Battosai42/amcp/tree/master/python3/dist . I don't have the IEC-444 fixture, but I have run the beta application and verified that it talks to the NanoVNA. A screen-shot of the application is attached.

Keep in mind this is an open-source program that is still under development and, therefore; you should not expect any support - although providing the author with encouraging feedback would surely be appreciated.

- Herb


alan victor
 

On Sat, Feb 29, 2020 at 04:54 PM, hwalker wrote:


Battosai42
Hi Herb, very nice Battosai42.

I did something similar a long time ago using the old school hp vector voltmeter.

One question on the displayed result example. The Co value shown of .56 pF. Seems awfully small. Does he remove the package C somehow in the measurement and calculation or is this just the Co of the xtal blank? That particular package has to have a larger Co than 500 fF. Typically that value is picked off by moving well offset from the series resonate frequency of the xtal.

Alan


hwalker
 

On Sat, Feb 29, 2020 at 11:41 PM, alan victor wrote:

" on the displayed result example. The Co value shown of .56 pF. Seems awfully small. Does he remove the package C somehow in the measurement and calculation or is this just the Co of the xtal blank? That particular package has to have a larger Co than 500 fF. Typically that value is picked off by moving well offset from the series resonate frequency of the xtal. "
===================================================================================

Alan,
I'm not sure if Battosai42 is a member of this forum and questions might be best directed to him at his GitHub page. His application is written in Python and examination of the code might answer your question. The project description at https://github.com/Battosai42/amcp does mention that either the Phase-Shift Method or -3dB Bandwidth Method can be used.

Rune's NanoVNA-saver application started life in a similar manner and today is a mainstay in the NanoVNA community. I wish Battosai42 similar success with his application.

- Herb


alan victor
 

Thanks Herb for the reply and info. I will try to message him direct. Since his call letters are given he may have a QRZ
PAGE and hence a direct email.

Alan


Jerry Shirar
 

I have to assume that this is a 26.000MHz crystal. The Fp is higher than that. It should be lower so that when you load the crystal that the frequency increases from there. So these numbers do not look good.

I would say that the calibration is at issue here. There is an app note at this link http://literature.cdn.keysight.com/litweb/pdf/5965-4972E.pdf . It shows the pi network for the 50 ohm measurement system.

Other than that, you need to assure that the calibration is done properly at the crystal point with this pi network in the system. The open needs to be about .1 pF. The short needs to be less than 0.1 ohms and the 50 ohm resistive load needs to have less than 10 nH of inductance.

The traces look good. More data points will render better values as well. Zoom in on the 26MHz area as this VNA may not have enough data points with this sweep to give decent numbers. Now that I think about it,that may be all you need. You don't need all that area showing the Fs.

I have hardly turned mine VNA on. I have not looked at crystals as I need to build a pi network fixture. I don't know how small a frequency span you can get.

Jerry N9XR


Nick Kennedy
 

I measured an 8.000 MHz crystal with the nanoVNA using my PHSNA notes and
Ignacio's suggestion of using a 2 kHz span and 12 segments. I did a
calibration on that basis with nanoVNA-saver and saved it.

I feel like I got a good measurement as it agrees within about 2% for Cm
and Lm of an earlier measurement of the same crystal with different
instruments a few years ago. The Q was within 1.5% and the Rs value was
dead-on.

I created an automatic crystal measurement function in PHSNA as Ignacio
noted. The key for me was in how to measure Rs, which is often done by
substitution or by using a pot. A formula from Wes Hayward showed me how to
calculate it from the crystal's attenuation at resonance and if you can do
that accurately, you can get a good reading of Rs. See below.

IIRC, I could do a crystal in 10 to 20 seconds or so if you don't count the
time I spent fumble-fingering the next crystal into the socket.

The process worked like this:

The user tells the software the resistance seen by the crystal in his
fixture, which is 12.5 ohms in my case. I've used both 4:1 transformers and
resistive pads but am settled on the latter now. They do give me ~ 23 dB of
loss, but there's still plenty of dynamic range.

The user must also do a one-time measurement of the fixture loss with a
short installed in place of the crystal. The difference between this value
and the loss at resonance with the crystal installed in the fixture is the
crystal's attenuation number needed to compute Rs.

The process starts with the user providing a frequency below the resonant
frequency with enough margin to be below the -3 dB point.

The software then begins scanning upward in 1 Hz steps and looking for the
resonant peak at minimum attenuation. The attenuation at that point is
recorded. Then the program scans down in 1 Hz steps until attenuation is 3
dB below the peak. It returns to resonance and scans upward in 1 Hz steps
until the upper -3 dB point is found. From these readings the 3 dB BW can
be calculated.

Next, the values for Cm, Lm, Rs and Q are calculated:

Rs = 2*Rg(10^(a/20) - 1)

Where Rg is the generator resistance or 12.5 ohms in most cases and a is
the attenuation of the crystal at resonance in dB, entered as a positive
number. Rs is the series loss resistance of the crystal.


Cm = BW / (2*PI*Fc^2*(2*Rg + Rs))

Lm = 1 / (39.48 * Fc^2*Cm) where (39.48 is 4*PI^2)

Q = (2*PI*Fc*Lm)/Rs

Where Fc is the series resonant frequency of the crystal.

In the above, C is in farads, L in Henry, and Fc is in Hertz.

In my measurement today, I got the BW and attenuation from nanoVNA and
cranked out the other stuff manually. But you can see how it could be
automated.

As for the parallel C or Co. No doubt it could be done with the nanoVNA or
with a low frequency instrument like the AADE or eBay versions of it.
Also, You can estimate Co as Co = 220 * Cm

73,

Nick, WA5BDU

On Sun, Mar 1, 2020 at 8:04 AM Jerry Shirar <radio.n9xr@gmail.com> wrote:

I have to assume that this is a 26.000MHz crystal. The Fp is higher than
that. It should be lower so that when you load the crystal that the
frequency increases from there. So these numbers do not look good.

I would say that the calibration is at issue here. There is an app note
at this link http://literature.cdn.keysight.com/litweb/pdf/5965-4972E.pdf
. It shows the pi network for the 50 ohm measurement system.

Other than that, you need to assure that the calibration is done properly
at the crystal point with this pi network in the system. The open needs to
be about .1 pF. The short needs to be less than 0.1 ohms and the 50 ohm
resistive load needs to have less than 10 nH of inductance.

The traces look good. More data points will render better values as
well. Zoom in on the 26MHz area as this VNA may not have enough data
points with this sweep to give decent numbers. Now that I think about
it,that may be all you need. You don't need all that area showing the Fs.

I have hardly turned mine VNA on. I have not looked at crystals as I need
to build a pi network fixture. I don't know how small a frequency span you
can get.

Jerry N9XR




Jerry Shirar
 

Excellent post Nick,

Very detailed and correct. Herb shows a span width of about 50kHz. A span width of 2kHz to 5kHz would be pretty good.

The C0 value shown will be garbage when looking near resonance as the series arm of the crystal shows a very low impedance. It shorts out the C0 value. If you find a span where the crystal is not resonant, you can do a ~2kHz span where the sweep is pretty flat, you can usually get the C0 calculated pretty good. Like maybe 30 MHz on a 26 MHz crystal or 10MHz on a 8MHz crystal. Otherwise a C Meter could work that is oscillating on a frequency not resonant with the crystal. But yeah. The user needs to realize you can't get all that information on the software in one sweep. The C0 needs to be made in a non resonant sweep.

Thanks.

Jerry N9XR


Oristo