Need help in measurement S parameters for differential LNA

Hi all,

My LNA design has differential input and output : GSSG. But the problem is our system can't measure NF and S-parameters for differential input and output. So I connect the off-chip baluns at the output and input of the LNA, then measure the total performance of the balun+lna+balun.

Next I connect the 2 balun together, and measure the performance of the 2 balun.

My problem is that after getting the S-parameters of the 2 baluns together, I don't know how to de-embed the performance of 1 balun alone.
I know that we can transform S parameters to T-parameters and use this formula: T(2balun)=T(1balun)*T(1balun) to find the T-parameters of 1 balun.

But square root of a complex matrix will give 4 results and I don't know which one to choose.

IS there any other ways to de-embed the S-parameters?

Million of thank.

If you dont have a direct measure of S3P of balun it will be really difficult to solve this issue.

You don't have any info about amplitude and phase of differential signal generated by the input balun.

Roughly, if you are well matched, you just can divide by two the back-to-back insertion loss and deembed this value from the LNA meas.

I hope it can help.

Mazz

Sorry, but I don't understand why you said that the square root of a complex matrix will give 4 results.

I think the procedure to de-embed the baluns is correct even if it could be not very precise, since you assume the two baluns are perfectly identical one each other.
The multiplication of the two T-matrix has to be element-wise.

Regards

A complex matrix may have up to 4 square roots. For example, The matrix

X =
7 10
15 22

has four square roots. Two of them are

Y1 =
1.5667 1.7408
2.6112 4.1779

and

Y2 =
1 2
3 4

The other two are -Y1 and -Y2.

Regards[/img]

Sorry for my uncorrect answer. Of course you are rigth.
Furthermore I wrote "The multiplication of the two T-matrix has to be element-wise" that is not true: I apologize for that.

However, since the multiplication between T-matrix is not the Hadamard one you cannot apply the square root entry-by-entry, I think you should solve it using f.i. the Denman-Beavers algorithm.

albbg : no worry, I really appriciate your reply.

I can mangae to find the square roots of T-matrix. But as I mentions, there are totally 4 square roots, so I don't know which one is really the T-parameters of the balun.

Is there other way to do the measurement for differential i/p and o/p LNA?

thank

hhnn

of course there are.

What you need is an S4P file. By definition, you can measure it considering the 4 ports of the LNA indipendently and measuring S2P files of 1-2, 1-3, 1-4, 2-2 and so on.

But it is a long way, as you need to terminate unused ports with 50 Ohm.

Then you can convert your S4P in S11dd, S21dd, S21cc (considering virtual differential and common mode ports).

there is a lot of literature about it, for example check agilent web pages, using a software (ADS) and/or instruments (multiport vector network analyzer).

I hope it can help.

Mazz