Measurement Series / Experiment

[Michael Wollensack METAS]

We want to make series measurement to get average and standard deviation. Is it possible to get all measurement seris data with average and standard deviation in one file? 

[Michael Wollensack METAS]

Hi,

The standard deviation of the mean of a measurement series is just the random part of the overall uncertainty. It can be computed with the following steps:

1. Measurement Journal / Add Begin Experiment (Type: Statistical).
   
2. Measurement Journal / Add Measurement Series (Mode DUT).
   Measure the DUT multiple times (at least 5 times (8 recommended) for a 1-port DUT or 11 times (16 recommended) for a 2-port DUT).
3. Measurement Journal / Add End Experiment.
   
4. Calibration and Error Correction.
   
5. Data Explorer / Select all error corrected measurements of the DUT and compute the mean value (right-click in the Data Explorer).
   
6. Data Explorer / Select the computed mean file of the DUT.
   
7. Data Explorer / Graph or Table / Unc: Standard, Unc Contributions: Not All, Only VNA Experiment Statistical.
   
8. The mean value and the standard deviation are shown for the measurement series of the DUT.
   

Comment:

No noise and connector repeatability uncertainties are added to the measurements between the ‘Begin Experiment’ and ‘End Experiment’ entries in the measurement journal.

Regards

Michael

[handans...@gmail.com]

Thank you Micheal. I have one more question. Which equation you use to calculate "VNA Experiment Statistical" ?

Also, is it possibble to define the measurement series number instead of pressing many times? 

[Michael Wollensack METAS]

Hi,

> Which equation you use to calculate "VNA Experiment Statistical" ?

x: data

n1: number of samples

n2: number of dimensions, e.g.: n2 = 2 for a one port measurement and n2 = 8 for a two port measurement.

dof = n1 - 1

p = 0.95: probability 95%

k1: coverage factor(dof, n2, p)

k2: coverage factor(Inf, n2, p)

factor = (k1*k1)/(k1*k2)/n1

covariance matrix = cov(x)*factor

That's how it's done in METAS UncLib which is used by VNA Tools II. An other way to compute the factor would be:

factor2 = (n1 - 1)/(n1 - N - 2)

factor2 and factor give similar results for p = 0.95.

> Also, is it possible to define the measurement series number instead of pressing many times?

- Measurement Series / DUT: You have to press OK for each measurement. (New connection).

- Measurement Series / Drift: You can specify the number of measurements you want to do. No user action is required. (No new connection of the DUT).

Regards

Michael

[Jan Grajciar CMI]

Hi Michael,

I forgot to add the beginning and the end of the experiment. Is there any way to get the uncertainty without having to create the whole journal again in offline mode?

Thanks

Jan

[Juerg Ruefenacht]

Hi Jan,

you can always add an entry to an existing Measurement Journal.

- Open the existing Measurement Journal

- Add Measurement / Begin Experiment (Statistical)

- Select the just new generated entry in the Measurement Journal (double-click) and edit the time stamp in this entry so the beginning of the experiment is on the right position.

- a similar procedure is needed for the definition of the End Experiment entry.

Maybe it is a good advice to make a copy of your existing Journal before you start to manipulate it or to save the new Journal with a different name first.
But then you must not forgot to also assigne the new Journal name in the Calibration Config tab and the Error Correction tab.

Hope this helps
Juerg

[Luca Oberto]

Hi Michael and Juerg,

I'm trying the "experiment" functionality of VNATools. I actually did not perform any experiment so far, so I am trying to use previous repeated measurements made individually.

Modifying the existing measurement journal adding the Begin/End Experiment with the proper time stamp seems to work, but my measurements where taken only four times (I measured a load) and the procedure stops and tells that at least 9 repetitions are needed. If I copy one of the measurements five other times (changing obviously the file name and updating the journal accordingly), all seems to work.

Of course I will perform actual measurements to test this functionality instead of doing this weird things, but it was just to see if I am able to use VNATools for this kind of analysis.

Now my question is: if, for any reasons, I have to take less than 9 measurements, is there a way to make anyway the computation?

Thanks,

Luca

[Michael Wollensack METAS]

Hi,

It seems that your measurement data contains all four S-Parameters (S11, S21, S12, S22). So you've an 8 dimensional problem and then you need at least 9 measurements to do statistics. See the appendix called Small Sample Statistics in the Math Reference of VNA Tools.

> only four times (I measured a load)

If you would have only measured S11 then you would have a 2 dimensional problem and you would need at least 3 measurements.

In your case when you measured a load and you collected all four S-Parameters then you have four times a 2 dimensional problem and you would need at least 3 measurements. You can tell VNA Tools that case by checking Independent S-Parameters in the Begin Experiment journal entire.

Hope that helps.

Regards

Michael

[Luca Oberto]

Thanks Michael,

it worked.

Best,

Luca