[evlatests] switched power issues

[Rob Selina]

Hi all,

Picking up on Paul's insight below, and recent data from the moon compression test, I think I've developed a plausible hypothesis for a significant source of Pdiff compression. I've asked Rick for some time at tomorrow's test meeting to review and discuss. Slides attached for those interested. 

Short version - I think Paul is right that there is gain modulation that leads to antenna power being added or subtracted from the Pcal in the Pdiff measurement (creating Pdiff expansion or compression respectively). However, I think the modulation gets progressively worse with increased system temperature (rather than something unique to the out-of-band cal signals), and is likely due to a perturbance of the LNA or post-amp bias. 

Cheers,
Rob


-----Original Message-----
From: evlatests [mailto:evlatests-bounces at listmgr.nrao.edu] On Behalf Of Paul Demorest
Sent: Friday, May 20, 2016 3:49 PM
To: evlatests at nrao.edu
Subject: [evlatests] switched power issues

hi all,

I've recently been exploring the effect of the VLA switched power signals on WIDAR data.  This has turned up some notable features, some of which may be relevant to everyone's favorite topic, "Pdif compression".  If any of this sounds interesting to you, read on below... the short story is that I think the practice of leaving all receiver cals firing at all times is maybe not so good; this may be causing a small gain modulation at the cal switching frequency; we should change this default behavior, and test what effect this has on the Pdif compression issue.

-Paul

------------

For background:  The default state that is set up when a VLA receiver is selected for use is for its noise diode to be turned on, switching at 10 Hz.  This setting for all other (not-in-use) receivers is left at whatever it happened to be at previously.  The net result is that the typical state of the VLA is for _all_ noise cals (all receivers) to be switching at 10 Hz at all times, regardless of which receiver is actually being used.

By using WIDAR's "pulsar binning mode" it is possible to fold the correlations at 10 Hz and see the effect of the noise cal signals on the data directly, for example by observing a calibrator and solving for antenna gain as a function of the cal pulse phase.  Alternatively, "FRB mode" can be used to dump correlations at 5 ms and do the folding offline; both setups seem to provide consistent results.

What I've seen is that there is a notable difference in how these results look depending on whether the cals for the not-in-use receivers are left switching or are turned off.  The first attached plot shows gain amplitude averaged over many 10 Hz cycles for ea22 at S-band.  Top panel is with all cals on (the usual situation) and bottom is with all cals off except for S-band.  The second plot shows the corresponding autocorrelation data, you can see how this signal lines up with the cal high/low state.  Immediately following a cal low->high transition there is a gain dip, then some ringing.  The same thing with opposite sign happens at the high->low transition.  The magnitude of this effect seems to vary with antenna, I've shown one of the stronger examples of it here.  Turning off the unused cals makes this ringing signal vanish, and leaves only a small, ~constant gain difference between the states (this in turn goes away if the S-band cals are turned off).  I think this latter feature may be a quantization effect due to the 4-bit correlation.

However, I have no explanation for how the "out-of-band" cal signals (which in principle should have no effect) are causing the "ringing".  
If this ringing is a true amplifier gain variation, since it has opposite sign at each transition it will cause an apparent Pdif compression.. basically, the average gain being different in the cal high and low states will couple a small fraction of Tsys into the Pdif measurement (Pdif ~= G*Tcal + DeltaG*Tsys).  The sign and approximate magnitude of this effect seem suspiciously comparable to the Pdif compression results Rick has shown various times.. for example a 0.1% gain difference would result in ~10% change in the measured PDif value when Tsys is increased by a factor of 5.

I think it would be interesting to try some of the strong source observations that show Pdif compression with extraneous cal signals on vs off, and compare the results.  I don't know if this will explain the full effect since there could still be other non-linearity present.  But based on everything I've seen so far, I do expect this change to have at least some effect.  Also I think it would be a good idea to make "cals off" the default for receivers that are not in use.. I can't think of any reason why they need to be left on (except possibly P-band for commensal VLITE observing).  Finally, it would be nice to know what is going on in the electronics to cause this effect, I don't really have any idea what might be doing this.

Comments, alternate explanations, or questions about any of this are welcome!
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 00-EVLA_TESTS_MOON20160407_v05.pdf
Type: application/pdf
Size: 1871888 bytes
Desc: 00-EVLA_TESTS_MOON20160407_v05.pdf
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20160608/e70a2a47/attachment-0001.pdf>