I'm trying to measure the return loss on a POTS splitter over POTS band (1k to 4kHz). I normally do this using an Agilent 4395A network analyzer with a transmission/reflection adapter using 50:600 baluns to match the splitter impedance and then let the analyzer do all the work. However, the sample I am trying to test now is for the UK and has a complex nominal impedance and I'm lost on how to accurately measure it. I started going down the balanced bridge method so that I could match the complex impedance on the bridge, but am not sure how to calibrate the analyzer using this approach. Any ideas will be appreciated. Thanks, David

I don't have a good answer, but the approach that I would take is to measure the device as a single-ended 50 ohm impedance measurement (really need a 3 port to do this, so you'll have to measure it 2 ports at a time and compute the differential impedances that you need). If you don't know how to compute the differential impedance, send me an email and I'll show you.

Next, you will do an impedance transformation on the Sdd terms to convert them to the desired real impedance. We do this by using S-parameters of the ideal transformer (Z1=12*Z2, so N=sqrt(12) ).

Then the tricky part is dealing with the complex bit. We do this by adding in the conjugate of the the complex bit to the result, sort-of like port matching. Thus, if you have a result that has R+jX, it will be S11=0 when the load is R-jX. So if you have a result that is Z, and you are trying to see S11 with Z0=R-jX, you take your 50 ohm result, and subtract jX from it, that is, you add a series element of -jX. This is the tricky part. The analyzer wants to see a real impedance. By adding the -jX, you get S11 when your dut presents 50+jX, which is the conjugate match condition.

Finally, there is some controvery about whether S11=0 should occur for Z1=Z0 or Z1=Z0* We chose the conjugate match condition for the simple reason that it matches exactly what ADS does. NIST guys argue for power waves (Z1=Z0), with good reasons, but we haven't implemented that yet.

So, this means a bit of off-line processing, but it is clearly a very proper way to do this.