AnsweredAssumed Answered

Periodic variations in NF data

Question asked by mattski on Jul 26, 2010
Latest reply on Aug 12, 2010 by mattski
I'm doing some noise figure measurements of some on-wafer FETs.  Currently we're doing 50ohm measurements, but we eventually will be doing full noise characterization in a Maury tuner probe station.

I have just done the 50ohm measurements so far, and I'm seeing a periodic variation in the data, which is visible but pretty small when measuring the entire "through" calibration path, somewhat noticeable when measuring an on-chip attenuator, and very noticeable in the NF data of the biased FET.  My test setup consists of HP 346C noise source, 8970B NFM, external bias tees, and Cascade Microtech probes.  The noise source and NFM were calibrated by Agilent about 5 years ago.  I would also like to test it with 8971C test set in the loop, up to about 4.5GHz.  The NFM is calibrated with a Cascade calibration substrate's Through standard.  We currently do no additional correction for the loss between noise source and DUT.

Here is a picture of the FET data I'm measuring: Image

Now the device itself is currently unmatched, so a VSWR problem anywhere else in the setup could cause the data to appear this way, but I've measured everything individually at -20dB or better return loss (VSWR < 1.22), including the noise source and NFM.  Could even a small reflection like this cause such big variations?

Can anyone think other reasons for this kind of data, or is a VSWR problem the most likely cause?