I was measuring a filter and I wanted to zoom in to a smaller span. That requires complete re-calibration. Is there a way to avoid that in some way? Interpolation for example? This is not in the current software as far as I know but would be very useful. So would a "save measurement" capability for when you are using the device without a PC.

Mike

On Fri, Jun 12, 2020 at 05:17 AM, Mike Willis wrote:

I was measuring a filter and I wanted to zoom in to a smaller span. That
> requires complete re-calibration. Is there a way to avoid that in some way?
> Interpolation for example? This is not in the current software as far as I
> know but would be very useful. So would a "save measurement" capability for
> when you are using the device without a PC.
==============================
Mike,
The VNA logic in the firmware was written from scratch, and calibration interpolation wasn't implemented. This means if you change frequencies no interpolation is performed and you return to an uncalibrated state.

Per Gabriel, the V2 designer, "Personally I've not had good experience with using interpolation on other VNAs that support it; I've found it to give wrong results every time unless you have an insanely high number of sweep points (1000). However if there is popular demand we can go and implement interpolation (as done on the NanoVNA -me)."

Beside interpolation, Gabriel also decided against implementing other standard NanoVNA features such as powering up with Save 0 as the default configuration, and interactively viewing and controlling the V2 when connected to a PC.

- Herb

I suppose too, that there could be an interpolation model found for approximating two port calibration matrix after SOL one port calibration find or maybe for the SOLT too. The need for recalibration on every sweep change is bad. Another way is to make a time-consuming but one and only full band calibration on the beginning of the setup and then use only frequencies, where the calibration was done.

>From my point of view, the sweep setting with "segment" I would like to see changed to number of steps or frequency step.

What about using nanovna-saver software, calibrate using big start and stop
frequencies, but with extra segmentation (101 points X N segments) to give
it extra detail?

Then, after loading calibration, one can sweep only in the frequencies of
interest, and even reduce the segments to have faster scanning.

The calibration needs to be done once, and for every cable/adaptor
combination, but you would only do it once... Then it's a matter of
recalling the correct calibration and measure.

Suggestion:

50Hz-53MHz (30 segments)
3Mhz-30Mhz (30 segments)
49MHz-71MHz (10 segments)
135-150MHz (20 segments)
420-450MHz (20 segments)
...

This will work ok and you can skip calibration #2 if you want.

Then load calibration, and just sweep for part or all of the calibration
start and stop frequencies...
For antenna analysis I am pretty sure that one would only need one
calibration, ie, without minding the adapters, or cables...





*73 de Luís, CT2FZI*



<http://goog_717545661>


On Sat, 13 Jun 2020 at 10:52, ok1vaw <ok1vaw@volny.cz> wrote:

> I suppose too, that there could be an interpolation model found for
> approximating two port calibration matrix after SOL one port calibration
> find or maybe for the SOLT too. The need for recalibration on every sweep
> change is bad. Another way is to make a time-consuming but one and only
> full band calibration on the beginning of the setup and then use only
> frequencies, where the calibration was done.
> From my point of view, the sweep setting with "segment" I would like to
> see changed to number of steps or frequency step.
>
>
>
>

I'm just kind of guessing but if you have a typical test set with a coax this causes narrow ripples at the coax length-frequencies and that's incredibly hard to empirically interpolate-calibrate. If I were trying to come up with a reasonable way to interpolate I think that modeling the measurement (with a lossy transmission line and passives?) and then using that model with calibration information would be much more likely to work reasonably than a more naked interpolation. My 2 cents.

-Mark

I suppose that the calibration is done by using a scattering matrix and I am convinced, that interpolation between the steps from single elements of the matrix should be possible. Even I think, that simple "lazy man calibration" to adjust a virtual twoport parameters by use of adjusting of gain/attenuation and center such as was done on analog HP8410 VNA could be used.
OK1VAW, Vojtech