AnsweredAssumed Answered

Compression on the PNA

Question asked by BryanFast on Apr 28, 2009
Latest reply on Sep 6, 2017 by Dr_joel
Okay, I'm using a PNA-X to measure compression on my device (though the same method should work for the PNA-C).  I'm not using the built in application, because I need more detail than it provides (like multiple compression levels, and DC current values).  So, I'm simply using a high power setup (test-set amplifier, and the appropriate receiver attenuations) and incrementing the input power level with the standard frequency sweep.

So here's my problem:  Everything looks fine with the data in the small signal portion of my measurement.  However, as my device starts to compress, the value at my B receiver is getting a large ripple versus frequency.  This, of course, directly translates into a ripple in my 1dBc calculations.  I measured the device on a different test bench, and knew from there, that the ripple isn't real.

So, the investigation began.  Firstly, I confirmed that the receivers were not compressing at those levels.  Then I started looking for a bad match in my signal path that would cause this ripple.  After some time, I confirmed that the ripple is being caused right at the output of my DUT.  An attenuator at that location helps the ripple, while the same attenuation on the other side of the cable (right at Port 2) does nothing.

My conclusion from that is that the output return loss of my part gets worse as it compresses.  This worsened return loss is combining with the reflection provided by my cable and test set, and causing a ripple.  So, no big deal.  Just add a attenuator at the DUT output and everything is fine, right?  Well, not completely.  The problem is that I can't add an attenuator close enough to my DUT, due to fixturing, to completely remove the ripple.  To complicate things, even when calibrated out, it adds noise to my S22 measurement on another channel.  Interestingly, I'm actually finding that the value of the attenuator is not as important as it's own reflection performance.

Surely this is a common problem.  We take care to give a great match on our other test setups, but it's easy there.  We don't often need much of a cable (if any), and large attenuations are not a problem.

Are there any tricks to getting around this?  Or is it a limitation of trying to measure too many things at the same bench?

Here are some quick numbers to quantify this:
1) Small signal S22 is -18dB or better in this band.
2) Power output of the device is around 20dBm at the 1dBc.
3) I have not had time to measure 'Hot S22' to see what happens under compression.
4) The ripple seen at my B receiver, without an attenuator, is around 1.2dB peak-to-peak.
5) The ripple seen with the attenuator is around 0.5dB peak-to-peak.

Thanks in advance.