NaN appearing in error corrected Data for reflection measurements

71 views
Skip to first unread message

wickli...@gmail.com

unread,
Nov 30, 2020, 9:56:37 AM11/30/20
to VNA Tools
Dear all

I am encountering a problem, where I get NaN as a Value for the measurement and the uncertainty after error correcting the data. This only happens at 9 kHz which, coincidentally, is also the lowest frequency point.

I read on here that common errors are that the characterisation simply was not done at this point, however, I double checked and all the input data starts at 9 kHz and also the linearity is specified way lower than what we encounter here.

As the Data is way to big to post here I will only post two screenshots for now.
First the raw data:


Now the errorcorrected data:


Once again: this only happens at 9 kHz. Does anyone have an explanation?

Thanks in advance
Patric

wickli...@gmail.com

unread,
Nov 30, 2020, 9:57:54 AM11/30/20
to VNA Tools
Quick update: I just realised, the problem occurs also with transmission measurements

Michael Wollensack METAS

unread,
Nov 30, 2020, 10:24:02 AM11/30/20
to VNA Tools
Hi,

can you make sure that 9 kHz in your raw data is greater equal to the 9 kHz in your cal kit definitions?

And not 8.999999 kHz and 9.00001 kHz

Can you share the sdatb files for your calibration kit definitions and the sdatb file of your raw data?

Regards
Michael

wickli...@gmail.com

unread,
Nov 30, 2020, 10:45:27 AM11/30/20
to VNA Tools
Hi Michael

Sadly the Cal Kit files are too big to be uploaded here, however, the calibration was done by METAS (Cert. No: 217-03130), maybe you have the files somewhere on your server? if not I will gladly upload them to a file server.

As for the raw measurement: you will find one of the files attached to this message.

Thanks in advance

Short_01.sdatb

wickli...@gmail.com

unread,
Nov 30, 2020, 10:58:51 AM11/30/20
to VNA Tools
Two more things:
I did check and they all start at 9000.000 Hz. 
And in case this is important, I did a uThru Calibration, so I used one definition from VNA Tools (unknownThru)

Michael Wollensack METAS

unread,
Dec 1, 2020, 3:45:34 AM12/1/20
to VNA Tools
Hi,

I've checked you're data and I saw that your raw data is starting at 8999.9999999999982 Hz and your definitions for the cal standards are starting at 9000.000000000000 Hz.

You can check that yourself by setting the numeric format to e16 in Table of the Data Explorer.

In the interpolation routine of VNA Tools (UncLib) I check for extrapolation using the following line of code

if (xx < x[0] || x[n1 - 1] < xx) return nan;

This returns nan for your problem at 9 kHz.

I've improved this morning this check by taking into account the epsilon of a double number. So the code looks now like that:

if (!(DoubleHelper.IsLessOrApproximatelyEqual(x[0], xx) && DoubleHelper.IsLessOrApproximatelyEqual(xx, x[n1 -1]))) return nan;

This will fix your issue, but it's not getting released with the next version 2.4.2 of VNA Tools. It will be part of VNA Tools V2.4.3 (around April 2021).

Workarounds for the moment.

Change the start frequency on the VNA to 9000.0001 Hz

Or try to use a segment sweep instead of linear sweep.

Or if the above two workarounds fail I can provide you a script which replaced the first frequency point of your raw data by 9.000 kHz.

Hope this helps.

Regards
Michael

Boris Dardel

unread,
Dec 4, 2020, 4:08:07 AM12/4/20
to VNA Tools
Hello,

I have some similar issue using a "simple line" calibration standard.
My measurements are from 0.3 to 1000 MHz, however I have to put frequencies between 0.3 and 1000Hz in the standard definition, otherwise I get NaN terms in the  Switched error terms.
By doing so the switch error term frenquency range is also from  0.3 and 1000Hz  but the correction in the data between  0.3 to 1000 MHz seems to be done properly.
I guess there must be some confusion somewhere between Hz (requested in the simple line calibration standard) and MHz (used for the measurement data). Where can we define the frenquency range considered in the calibration config ?


By the way, can one explain what the button "varaible gamma" in the  "simple line" calibration standard is used for ?

Best regards.
Boris
 

Michael Wollensack METAS

unread,
Dec 4, 2020, 7:44:06 AM12/4/20
to VNA Tools
Hi,

can you please share all raw data, the measurement journal, the calibration config and all involved database entries.

Thanks and Regards
Michael

Boris Dardel

unread,
Dec 7, 2020, 10:07:46 AM12/7/20
to VNA Tools
Hi,
I checked my data.
In fact, for the sake of comparaison with older data, I used imported measurments that were not in the same frequency range than the original data (and hence of the journal), so I guess the problem came from here.
I don't have this issue on the data measured directly with VNA tools.
Sorry for the inconvenience.
Best regards.
Boris

wickli...@gmail.com

unread,
Dec 21, 2020, 3:28:01 AM12/21/20
to VNA Tools
Dear Micheal

Thanks for your reply. I couldn't find time to fully try the workaround yet but for now it looks like this will resolve the issue.
Thanks very much.

Regards
Patric

Reply all
Reply to author
Forward
0 new messages