Drift Propagation

[Adam Petrus]

I wish to be able to incorporate uncertainty into some of the material measurements such as Split Post Dielectric Resonators (SPDR) and Free Field Focus Beam measurements.  The calibration options within VNA Tools do not allow me from what I understand to apply these specific calibrations. In fact the SPDR data is un-calibrated and a qfactor and frequency shift is used to determine the materials. In fact what I require is something more simple which is the ability to propagate the uncertainty none error corrected data.

In the pursuit of this I looked at taking the VNA device and Journal and raw uncalibrated raw measured files and populating them with corresponding Jacobi dependencies.

To start this I wanted a simple example to compare. Using the TransNorm and RefNorm I should be able to generate this to force the normalisation to be 1.0.

Attached is some notes with some questions. My main question is comes from my conclusion that VNATools applies a sqrt(time) scaling function for drift. Is this correct? After doing some drift analysis using some of our analysers suggests this may be too aggressive.  

[Michael Wollensack METAS]

Hi,

I'll try to answer some of your question first. This should help you to go one step further.

If you want to apply uncertainties to your raw data you can perform a dummy calibration where you set the calibration standard definitions equal to the measured raw data. This results in the following error terms: Directivity = 0, Match = 0 and Tracking = 1.

> Note I tried doing a one port S21 but vdatb seems to be the only ?datb option. Why is sdatb not an option for single parameters?)

A stab file can just store a complete S-parameter matrix, which is the case for e.g.: S11 or S22 or S11, S21, S12 and S22 as complete set but not for S21 as single parameter. Therefor I've vdatb.

Unc Budget:

> Another point is the sign of the Jacobi does not seem to displayed just the magnitude.

You should update to VNA Tools V2.8.0 or newer then the Jacobi value and the input standard uncertainty is stored and displayed separately. In older versions only the product was stored (with the sign) and showed in the Unc Budget (without sign).

Drift:

In the database of VNA Tools you specify the drift for 24 hours.

The variance of that drift value is then scale with the time: u^2(drift_i) = t_i / (24*60*60) * u^2(drift_spec)

This gives a square root behavior for the standard uncertainty as you observed.

> This means the drift uncertainty can be minimised by making measures minimising the time between measurements. This makes sense.

Ye this was the idea of it.

The AnalyseDrift.m MATLAB script is obsolete. This has been written before the drift was assumed to be correlated over time.

You should analyze the drift as it's written in the EURAMET guide.

Hope this helps.

Regards

Michael

[Adam Petrus]

Hi 

Thank you for your insanely quick reply. I'll look into the EURAMET guide again, but the drift measurements data over time we have measured appears to have a uncorrelated and correlated part.  Although  at this stage need to relook with the context of the EURAMET content. 

I'll look at the new version as well, I have been meaning to do this anyway. Lost my admin rights after not using them for a while, hence not installing it on my laptop. Reapplied so hopefully in a week I'll have it installed or I'll just use one of our lab PCs if I fancy sitting in a windowless room. I have another two queries, one on noise where the noise appeared to use the source port rather than receiver, but check with the new version. The other is the windowing in the time gating dialog. The window for me was not changing the output as expected. However I'll point out at this stage, there have been cases where I thought there was something not right only to find I had made a mistake.

Thanks

Adam