Hi!

We have a rather interesting problem and may be why they make 4 port Analyzers. Here are the specifics:

1) We are testing a 4 port hybrid coupler in a temperature/vacuum chamber environment.

2) We are testing with a 2 port PNA.

3) DUT P1 & P2 are the inputs, P3 & P4 the outputs

4) We need to test the following:

a) P1 to P2 for Isolation

b) P1 to P3 for Input & Output Return Loss

c) P1 to P4 for Input & Output Return Loss

d) P2 to P3 for Input & Output Return Loss

e) P2 tp P4 for Isolation

5) Since we don't have direct access to the DUT (it's under vacuum) we need 5 different cal's (P1/P2, P1/P3, P1/P4, P2/P3, P3/P4. So far so good.

6) We connect up all 4 ports of the DUT to a bulkhead on the chamber wall. On the other end of the bulkhead we have access to the 4 RF paths. 2 of the RF paths will connect to the PNA & the other 2 will have 50 ohm loads.

Our problem is is that I'm getting about a 3 to 4 dB change in Isolation when comparing a termination connected directly to the DUT vs connecting to the DUT via the bulkhead (due to, I assume the uncalibrated cables in between the bulkhead and the DUT.

Question: Is it possible to configure 2, 2-port Network Analyzers to test a 4 port device?

We have a rather interesting problem and may be why they make 4 port Analyzers. Here are the specifics:

1) We are testing a 4 port hybrid coupler in a temperature/vacuum chamber environment.

2) We are testing with a 2 port PNA.

3) DUT P1 & P2 are the inputs, P3 & P4 the outputs

4) We need to test the following:

a) P1 to P2 for Isolation

b) P1 to P3 for Input & Output Return Loss

c) P1 to P4 for Input & Output Return Loss

d) P2 to P3 for Input & Output Return Loss

e) P2 tp P4 for Isolation

5) Since we don't have direct access to the DUT (it's under vacuum) we need 5 different cal's (P1/P2, P1/P3, P1/P4, P2/P3, P3/P4. So far so good.

6) We connect up all 4 ports of the DUT to a bulkhead on the chamber wall. On the other end of the bulkhead we have access to the 4 RF paths. 2 of the RF paths will connect to the PNA & the other 2 will have 50 ohm loads.

Our problem is is that I'm getting about a 3 to 4 dB change in Isolation when comparing a termination connected directly to the DUT vs connecting to the DUT via the bulkhead (due to, I assume the uncalibrated cables in between the bulkhead and the DUT.

Question: Is it possible to configure 2, 2-port Network Analyzers to test a 4 port device?

Two 2-port PNA's will probably complicate the matter since it is hard to orchestrate the source and load match conditions between the sweeps.

As you already pointed out, the trouble you see is undoubtably coming from the effective load match at the port interfaces inside the chamber.

I think that you could, with a bit of work, create a "generalised" S-parameter network with your device, and through a round robin set of measurements, obtain sufficient information to get at the fully corrected 4-port, but it would be a lot of work (I'm sure you could justify the cost of an extension test set based on the risk that you don't get the mesaurmenet right, and the time it would take to figure out all the math). I think to figure out all the math, you need to measure your cables terminated in 50 ohm loads, for each cable, Then for each 2 port you create a generalized 4 port network where you know that (using examples of port1 and port 2 measured), that you know the row of b2 = a1 s11 + a2 S12 + a3 S13 + a4 S14, but you only know that a2=0 and b3/a3 is the reflection measured for that termination, and b4/a4 for it's effective termination. You don't know a3 or a4 by itself, but you know you can use b3/a3=gamma3 to remove one of the terms from the equation. Then you repeat this for every b. This gives you 2 conditions known (for a1 and a2 drive). Then you switch around the loads and write the equations again, except this times all the a's and b's are new (superscript them with port pairs such as a1super21, and a2super21, then the next time it would be a1super31...).

If you work this out far enough, you may find that you have enough equations to solve simulatenously for the S-parameters, which are the only things that don't change between the measurment setups. But you'd have to keep careful track of the gammas (reflections) for each loading of the unused ports and make sure you use the same loads for each scenario as what you previously characterized.

I can tell you it took 3 engineers about 7 months to work out the multiport math and validate it, and we still get caught occasionally with a missed use case (just fixed on last week!, hope it's the last).

The extension test sets are the N4419B, N4420B or N4421B depending upon whether you need 20, 40 or 50 GHz (we also have an N4421BH67 for 67 GHz four port). If you are making really precise measurements, you'll want option 100 which adds a buffer amp to improve noise floor and residual source match.

There... a nice simple easy answer!