AnsweredAssumed Answered

Group Delay basics... very basics

Question asked by mjz on Jan 24, 2007
Latest reply on Jan 26, 2007 by Forumsadmin
Hi, I'm trying to put together a test for group delay and I've read some material, but the actual specifics on how to do this test is alluding me.  Any articles or insight would be appreciated although I really couldn't make heads or tails of the Agilent white paper on group delay for mixers.  I'll describe my situation below.

The device under test
I'm testing what basically amounts to an amplifier on wafer, but it also has adjustable attenuation and phase.  It's pretty wideband we're talking 2-18 GHz or so.

The setup
I'm using a 8362B PNA to measure the group delay, I'm also going through a pre-amp.  The pre-amp is actually the agilent H81 setup.  It is meant to be used for pulsing, but I have just fed it a static TTL signal while I try to figure this out.

I'm also going through a custom switch network.  Those too are from agilent (87222E), and beyond that I just go through some cable to the wafer probes.

The methodology
At first I tried to just take a sweep over the entire 2-20 GHz range with .25 GHz steps.  That was pretty much a disaster.  Negative group delays and other weird things.

Now I've settled on an algorithm where I setup the PNA to the following parameters for each frequency I'm interested in obtaining group delay for.

Span 100 MHz
Bandwidth 1kHz
Num points 101
Smoothing over 11 points

Basically I'm zooming in on the frequency I want and trying to get a reasonable delta frequency.

The questions
I've read alot about aperture and so forth, but my question is What is a reasonable aperture?  1MHz? 10MHz? Should it be in the kilohertz range?  I guess I'm looking for a good rule of thumb maybe.

I've ran across a strange phenomenon, and I can give you a screen capture if anyone is interested, but when I activate my SOLT cal (which was really done with S-parameters in mind and was done over 18 GHz span with .25 GHz steps) I get a large divot at the frequency I'm interested which actually goes into negative time, but the group delay looks reasonable to my novice eye every where else.  By reasonable I mean 30 or so ns, so my question then is what do I need to look for in the cal to make sure I get a good group delay measurement?  Should I do the cal over the same aperture as my measurement?

My third question also pertains to the cal.  I notice when I turn the cal off there is virtually no difference between most of the measurement points with the lone exception of the point where the divot is.  It goes away. I guess this is really a part of the second question, but I'm really just wondering if this makes sense.  There is literally about a 3-5 ns difference in group delay between 100 out of the 101 points taken.

One last question.  I don't expect anyone would no this without knowing the specifics of the device, but as mentioned, my device under test has adjustable phase.  However, when I have a group delay trace displayed on the PNA and I adjust the phase of the device, I was expecting the group delay to be effected.  It was remarkably uneffected.  Intuitively, does this make sense (ie, should changing the phase of the DUT have an effect on group delay?)

I appreciate anything anyone has to offer.  I look forward to becoming involved with the community.  My expertise really lies with programming instruments.  Some over COM, but mostly over GPIB.  

Outcomes