De embedding S parameters

[jayce3390]

de-embedding s-parameters

Hello

I performed S parameters measurement data of a symmetrical Test Fixture in thru configuration .

The TF is used to make transistor measurement.

I want to get S parameters of a half part of Test Fixture from full TF measurement.

Does anyone have idea to do that?

Thanks

 

[Mazz]

deembedding matrix

This is a problem that I have encountered too.
I think that is not mathematically possible to solve it, as there are more variables than equations (but I'm not sure...).
We measure a 2 port network made by:

-Test Fixture - DUT - erutxiF tseT-
and
-Test Fixture - erutxiF tseT-

how to calculate DUT ???

Mazz

 

[jayce3390]

de-embedding

Thank you for your answer, i think you have to cut the S params matrix.
First, You must convert your into [T] matrix
and then [T]= [T1]*[T2]

To find T1 and T2 parameters, you use symmetrical consideration of the TF perhaps sqrt(S21)=S21 (of T1) * S21 of (of T2) ... but I m not sure .
The problem stays for phases.

If you have any idea, let me know

Thank you

 

[Mazz]

deembedding

The only idea I have now is to stay with the well proven TRL calibration. Longer but accurate.
In any case, if there is a solution I will be interested.
Mazz

 

[madengr]

de embedding s parameters

Google "Two Tier Calibration". I use Maury fixture software which does this automatically.

**broken link removed**

You do a full 2-port cal at both the fixture reference planes (SOLT) and the DUT reference planes (TRL). The software then reads back the error parameters for both cal states and computes the s-parameters for each fixture half. I don't know the mathematics behind it but I'm betting it's a simple matrix division of the two in T-parameters. It is more than likely described in the ARFTG literature.

It can also be done with the adapter removal algorithm built into the VNA however I don't remember how to do it. It was passed on verbally to me by a guru at Maury.

 

[jayce3390]

de-embedding s parameters

Ok thank you very much, but I can't perform a TRL calibration, if I could make a TRL calibration deembedding process will become useless.

 

[madengr]

s12 deembedding

You can probably do what you mentioned with T1=T2=sqrt(T). This assumes the return loss is very low. You can also short the DUT and measure S11, then assume T21=sqrt(T11). I did this using a Matlab script years ago. Beware than sqrt(1)=[1,-1] so you results will periodically flip around the the axis on a polar plot. The key to correct this is to start you measurement at a low frequency before teh phase crosses -180 degrees. That way you can establish a true phase reference and your script can correct for the sqrt(1) problem.

You could do a SOLT, it does not have to be TRL. TRL is just more accurate.

The deembedding process deos not become useless if you can do TRL cal. You still need to know the S-parameters of each fixture half if you want to attach tuners. It's much easier to repeat the calibration at the fixture referecne planes than the DUT.

 

[Mazz]

de-embedding vs trl

The key factor is accuracy.
I'm supposing that jayce3390 wants to know the half-fixture Spar to deembed them from the complete meas and obtain DUT Spar. jayce3390 do you confirm that?
So jayce3390 has one "thru" fixture and one DUT fixture. To make a TRL just make another couple of lines (line and reflect) and you will avoid any Spar post processing (de-embedding) with all advanteges of a direct meas.
Generally speaking (and in my experience) in a test fixture is not easy to get a very high return loss (>20 dB), especially at very high freqs to use the approx needed to use the formula T1=T2=sqrt(T).
So, if you are not interested in accuracy (but how to calculate DUT Spar error using this method?) use it, otherwise SOLT or TRL.
What do you think about it?
Mazz

 

[jayce3390]

de-embeding s parameters

Yes Mazz I confirm I want to obtain DUT Spar from the complete meas (TF+DUT+TF) and thru TF meas.
But I can't perform TRL calibration with my symmetrical TF.

Effectly IRL of my TF is around 10dB.
Finally I can try :
If I suppose that both parts of TF are matched at the DUT accesses I can write:
S22 (half part)=0 and S11 (half part) = 0

And if both parts are identical I can tell:
S21 (first part) = S21 (second part)
S12 (first part) = S12 (second part)

S11 (full) = S11 (first part)
S22 (full) = S22 ( second part)
S21 (first part)=S21 (second part)= sqrt(S21 full)
S12 (first part)=S12 (second part)= sqrt(S12 full)


Is it true?

 

[Mazz]

s-parameter de-embbeding

Let me say that I don't like this way.
The equations seem correct but the approximation you are making gives a low accuracy in the result. If this is OK for you, go on.
By the way, why you cannot perform a calibration?
Mazz

 

[rillyxue]

microwave deembedding

you can use pel language in ic-cap to do it.

 

[jayce3390]

trl delay line

I m using IC microwave TF, and it's hazardous to perform TRL calibration with this kind of TF because IC microwave provides TRL cal kit but you have to use very thin and small peace of line to make the thru for example or to connect TRL delay line and finally you get as much results than of measurements due to difficulty to place the peace of line.

jayce3390

 

[ConstantWave]

You might want to look at our product, Spectro VNA. Using Time Domain Substitution, you can move your reference plane to anywhere that you can temporarily create a short.

 

[cnm]

There is an Agilent VNA fixture cal wizard that is handy in removing two end fixture's S-par responses.

It assumes the two end blocks are nearly identical (not so true at high frequency).

It just needs a center thru block.

One measure is END block+thru+END block; another measure is: END Block+DUT+END block. The wizard will then automatically de-embed the end blocks effect from 2nd measurement.