Contact an Expert

Discussion Forums are temporarily in read only mode. All forums are visible but posting is not possible. Please check back shortly.

Thread: Group Delay Points versus Smoothing Points


Permlink Replies: 32 - Pages: 3 [ Previous | 1 2 3 | Next ] - Last Post: Jul 20, 2012 12:11 PM Last Post By: Dr_joel Threads: [ Previous | Next ]
miyamky

Posts: 111
Registered: 08/03/09
Re: Group Delay Points versus Smoothing Points
Posted: Jul 18, 2012 1:41 PM   in response to: Dr_joel in response to: Dr_joel
Click to report abuse...   Click to reply to this thread Reply
Our DUT uses an external UNY option signal generator (8663B) for its LO. All measurement devices are tied to a single 10MHz reference. The LO is at a fixed frequency.

As to the method, there's another thread someone in here that Dara worked w/ us (mostly our test set engineer) for about 2 months to develop the cal. We did discuss the SMC+Phase method but I believe Dara told us to use the VMC method (let me see if I can find that thread). The final IFBW,averaging,smoothing,# of pts settings (again over that 2-3 month) period was agreed upon by the systems engineer, test set HW engineer and all the unit design engineers.

The reason that all of this came up was that I found that our SW was flawed (not waiting for all of the averages to complete before reading the trace) which you recently help us fix and then I re-ran the SW on a new device (that was previously tested w/ a R&S) and the DUT which passed the variation on the R&S is now failing at hot temperature (and now by a significant margin). It barely passes at ambient temperature (the R&S measurement had about 20ns of margin).
miyamky

Posts: 111
Registered: 08/03/09
Re: Group Delay Points versus Smoothing Points
Posted: Jul 18, 2012 1:54 PM   in response to: miyamky in response to: miyamky
Click to report abuse...   Click to reply to this thread Reply
After talking w/ akalei again, we "think" this is what we started w/ in developing our VMC:

http://na.tm.agilent.com/pna/help/latest/Programming/GPIB_Example_Programs/Create_and_Cal_a_VMC_Measurement.htm

After that you'll find a few threads where we had some problems implementing it and Dr Joel, Dara and Mr Danzy all helped us in about 2010.
miyamky

Posts: 111
Registered: 08/03/09
Re: Group Delay Points versus Smoothing Points
Posted: Jul 18, 2012 1:57 PM   in response to: miyamky in response to: miyamky
Click to report abuse...   Click to reply to this thread Reply
One of the DUT design engineers said that he'd agree to using a 5% of BW smoothing setting but he wants to look at the data in detail.
daras


Posts: 918
Registered: 12/01/05
Re: Group Delay Points versus Smoothing Points
Posted: Jul 18, 2012 3:24 PM   in response to: miyamky in response to: miyamky
Click to report abuse...   Click to reply to this thread Reply
It would be nice if you could point out that previous thread (given the number of different customers and measurement setups I work on, my memory is a bit foggy). the reason I ask is that given the success we have been having with SMC + Phase (and specially our new soon to be released Phase Reference Method), there are very few circumstances, where I would recommend VMC over SMC+Phase, if what the customer ultimately wants is Group Delay and not absolute phase response. So having a link to our previous discussion will inform us of the setup details that led to the current measurement approach and might be helpful in troubleshooting the current issues.

Just as a point of reference, here is a screen shot of a set of measurements where I compared all the different Mixer/Converter Group Delay measurement techniques available on the PNA:



Edited by: daras on Jul 18, 2012 3:29 PM
Dr_joel


Posts: 2,733
Registered: 12/01/05
Re: Group Delay Points versus Smoothing Points
Posted: Jul 18, 2012 5:15 PM   in response to: daras in response to: daras
Click to report abuse...   Click to reply to this thread Reply
From Daras measurements you can see we have only about 0.5 nsec delay ripple with VMC. So, I cannot understand how you get 20 nsec of ripple unless the test setup has some very odd flaw. SMC+Phase should work very well for you setup. The main issue we see is trace-to-trace noise on narrow bands, and you state that you don't see any trace-to-trace noise.

Some common problems with VMC can be analyzed from the following questions:

1) what are you using for a reference mixer? Is it filtered at the IF?
2) what are you using to send LO signal to the reference mixer? Is it isolated between reference and test (sometimes, IF limage and eakage signals from the reference mixer floats out the LO port of the ref mixer, across the splitter and into the LO port of the test mixer to appear as IF at the DUT output). --note: SMC+Phase eliminates this issue
3) What are you using for a calibration mixer? Is it filtered at the IF? What does it's delay look like if you remeasure after calibration? How does it compare to the mixer characterization file.

20 or even 2 nsec of delay error are much to large to attribute to instrumentation error. It must be setup and calibration issues; that or the DUT has the response.

You might send me an email with your details and we can take further discussion off-line.
miyamky

Posts: 111
Registered: 08/03/09
Re: Group Delay Points versus Smoothing Points
Posted: Jul 18, 2012 5:45 PM   in response to: Dr_joel in response to: Dr_joel
Click to report abuse...   Click to reply to this thread Reply
Dr_joel wrote:
From Daras measurements you can see we have only about 0.5 nsec delay ripple with VMC. So, I cannot understand how you get 20 nsec of ripple unless the test setup has some very odd flaw. SMC+Phase should work very well for you setup. The main issue we see is trace-to-trace noise on narrow bands, and you state that you don't see any trace-to-trace noise.

Some common problems with VMC can be analyzed from the following questions:

1) what are you using for a reference mixer? Is it filtered at the IF?
2) what are you using to send LO signal to the reference mixer? Is it isolated between reference and test (sometimes, IF limage and eakage signals from the reference mixer floats out the LO port of the ref mixer, across the splitter and into the LO port of the test mixer to appear as IF at the DUT output). --note: SMC+Phase eliminates this issue
3) What are you using for a calibration mixer? Is it filtered at the IF? What does it's delay look like if you remeasure after calibration? How does it compare to the mixer characterization file.

20 or even 2 nsec of delay error are much to large to attribute to instrumentation error. It must be setup and calibration issues; that or the DUT has the response.

You might send me an email with your details and we can take further discussion off-line.


1) The reference mixer (Marki Microwave M2-0026LB) IS filtered at the IF
2) E8257D (w/ UNY option)
3) Same type as mixer as the reference mixer. --- EDIT akalei says the reference mixer is NOT filtered but the other calibration mixer IS filtered.

Trying to keep this here since there are others watching us (be vewy vewy quiet)

Edited by: miyamky on Jul 18, 2012 5:02 PM

Edited by: miyamky on Jul 18, 2012 5:04 PM
daras


Posts: 918
Registered: 12/01/05
Re: Group Delay Points versus Smoothing Points
Posted: Jul 18, 2012 5:59 PM   in response to: miyamky in response to: miyamky
Click to report abuse...   Click to reply to this thread Reply
I did a cursory review of these prior threads and my memory is well jogged :-) However, it does not appear that at any time we really discussed the merits of SMC+Phase versus VMC and that is partly because this whole discussion started in 2009, when SMC+Phase was just getting off the drawing board. So if you are having systematic problems with VMC today, it might not be a bad idea to revisit the SMC+Phase Reference method (which should be available by the end of August). there is also a new calibration wizard called Cal All, which should make calibrating multiple channels at the same time a breeze. It might be time for you guys to call in the local FE/AE and ask for some demos.
miyamky

Posts: 111
Registered: 08/03/09
Re: Group Delay Points versus Smoothing Points
Posted: Jul 18, 2012 6:10 PM   in response to: daras in response to: daras
Click to report abuse...   Click to reply to this thread Reply
It would almost take an act of god for us to wait until Aug for an upgrade. But we can look at it for a future upgrade IF needed (and I'd have to make a business case). We'll start discussions about w/ our field rep.

It looks like the RDE is preliminary upbeat about my initial findings w/ the new smoothing points. Now we have to convince the systems engineer.

So is Dr Joel saying that I shouldn't see these big swings in variation (max - min) in-band w/ the different smoothing points?

They're also confused as to why the IFBW doesn't seem to have much of an effect (I've only really tried 1kHz vs 50Hz recently).
daras


Posts: 918
Registered: 12/01/05
Re: Group Delay Points versus Smoothing Points
Posted: Jul 18, 2012 6:56 PM   in response to: miyamky in response to: miyamky
Click to report abuse...   Click to reply to this thread Reply
I think that is exactly what Dr_Joel is saying. with a well behaved mixer/converter, the group delay variation that is attributable to the measurement instrument is on the order of +/- 0.5 ns, so anything bigger than that is symptomatic of some other underlying issue and that is what you/we should try to figure out. IF Bandwidth reduction or averaging can only improve the data if the source of the variation is noise or at least noise like (i.e the data changes from sweep to sweep). however, if the variation is due to ripple that was either introduced as part of the calibration or it exists in the raw measurement of the DUT, then changing the IF Bandwidth or increasing the averaging will have no effect. the problem with mixer measurements is that there are many sources of ripple, some of which cannot be calibrated out, no matter how hard you try. in some of those cases, doing simple things like adding 3-6 dB pads to the ends of the cables (not much help if you are on wafer), or using the source attenuation on the RF and IF ports can do a lot to improve the overall measurements, this is why having local support can be a lot of help. We do have AEs that have clearance, so that may be something you could discuss with your local Agilent reps.
miyamky

Posts: 111
Registered: 08/03/09
Re: Group Delay Points versus Smoothing Points
Posted: Jul 18, 2012 7:18 PM   in response to: daras in response to: daras
Click to report abuse...   Click to reply to this thread Reply
I'll check w/ Mike Finnegan and see if your people can someone come into our lab (and maybe they can better explain to you what's going on).

Do you want me to try the pads? Both input and output? Do I need to recal?
Dr_joel


Posts: 2,733
Registered: 12/01/05
Re: Group Delay Points versus Smoothing Points
Posted: Jul 19, 2012 12:14 AM   in response to: miyamky in response to: miyamky
Click to report abuse...   Click to reply to this thread Reply
IF BW reduction does not, in general, reduce variation in the group delay measurement if the variation is due to phase noise of the mixer LO, VNA source or VNA LO. This is because noise in group delay is related to noise in phase traces. As the LO phase noise cause the signal to "wiggle" back and forth across the IF filter, the phase varies up and down. The phase change is on the order of 360 degrees change over the BW of the filter. Lowering the IF BW reduces the noise, but also narrows the response so the phase slope (360 deg/ BW) is increased and the phase deviation due to phase noise is the same as at the wider IF BW. That is why,and only for mixer measurements, lowering IF BW does not reduce noise (sweep to sweep variation) of the GD response of mixers.

But, in a previous post you said the variation in measurement results doesn't change with averaging (which will reduce trace noise on the GD trace). This indicates the variation is built into the calibration.

Adding pads reduces varation from higher order products, and so can improve results. Try pads on the test port cables (6 dB is a good start).

Filter is generally not needed on the reference mixer, but higher order products can exit the reference mixer and remix to generate inband signal. All very tricky.

The most important thing for you to look at is the group delay response of your calibration mixer. If it has a lot of variation, it is likely in error; and this variation will be added to the overall measurement. But, if you turn off calibration, then you see only the raw variation in GD, and you already said it is similar to corrected variation, so I am very suspicious of some other issue that is not well defined, such as lack of isolation between ref and test. Thus, I would suggest you try SMC+phase, as a raw measurment, on your DUT mixer to see its response. You can use SMC+phase without the phase reference calibration and its total ripply shoudl be less than 5 nsec.
miyamky

Posts: 111
Registered: 08/03/09
Re: Group Delay Points versus Smoothing Points
Posted: Jul 19, 2012 11:05 AM   in response to: Dr_joel in response to: Dr_joel
Click to report abuse...   Click to reply to this thread Reply
Dr Joel/Dara,

Please take a look at the following thread:

http://www.home.agilent.com/owc_discussions/thread.jspa?threadID=33586&tstart=0

There, we sort of (where we pulled your teeth) discuss a different test of another unit (w/ the same type of test set). In that test, we're measuring phase rate of change/stability. That test does not even cal the same mixers used in the GD test. However, when we measured the phase rate of change of just the mixers themselves (by testing w/ just a UNY sig gen through the mixer, and then comparing phase stability against the same setup WITHOUT the mixers showed that the phase stability of the mixers themselves added about 0.2-0.3 degrees of change per 100ms.

We think this info may also add to your understanding. We're currently talking w/ our local field rep to get an AE in here (we did find one w/ the proper clearances and tickets - we just need to get his program briefing approved). We're also going to have him take a look at this entire thread before he gets here.
miyamky

Posts: 111
Registered: 08/03/09
Re: Group Delay Points versus Smoothing Points
Posted: Jul 19, 2012 11:21 AM   in response to: miyamky in response to: miyamky
Click to report abuse...   Click to reply to this thread Reply
Dr Joel,

My systems engineer is asking how you derived the 5-10% number for smoothing? I asked him to provide his math (and I'll put it here if it's okay w/ security) on how he derived his much lower percentage.
daras


Posts: 918
Registered: 12/01/05
Re: Group Delay Points versus Smoothing Points
Posted: Jul 19, 2012 11:49 AM   in response to: miyamky in response to: miyamky
Click to report abuse...   Click to reply to this thread Reply
miyamky wrote:
Dr Joel/Dara,

Please take a look at the following thread:

http://www.home.agilent.com/owc_discussions/thread.jspa?threadID=33586&tstart=0

There, we sort of (where we pulled your teeth) discuss a different test of another unit (w/ the same type of test set). In that test, we're measuring phase rate of change/stability. That test does not even cal the same mixers used in the GD test. However, when we measured the phase rate of change of just the mixers themselves (by testing w/ just a UNY sig gen through the mixer, and then comparing phase stability against the same setup WITHOUT the mixers showed that the phase stability of the mixers themselves added about 0.2-0.3 degrees of change per 100ms.


One thing that wasn't made clear in that thread was the actual measurement being performed under the CW sweep condition. Are you measuring a ratio like "B/R1" or are you measuring a single receiver like B? You say you are stimulating the mixer with an external UNY option sig gen; is that going to the RF port of the mixer or the LO port? if it is going into the RF, are you sampling the sig gen with a PNA-X reference receiver, so that you can do a ratio measurement? without doing a ratio measurement, you really can't tell where the phase change is coming from, the source or the mixer, the PNA-X's internal LO or a combination of all. Also, unless you are doing this in a well controlled environment and with high quality phase stable cables, most or at least part of the phase change you do see could be attributed to the cables and any additional plumbing you might have in your setup. this is especially true in a CW measurement, where you can't distinguish a constant phase offset caused by change of length in the cables versus change of phase response of the mixer.

Point your RSS reader here for a feed of the latest messages in all forums