[evlatests] 128-MHz spectral artifacts

[Paul Demorest]

Hi all,

Thanks to some careful investigation of L-band RFI by Lilia and Pedro, we have found a new (at least to us) type of spectral artifact in EVLA data.  We see low-level, frequency-shifted copies of strong RFI lines at frequencies +/- 128 MHz (and likely +/- 256 MHz as well) away from the real signal.  These signals are about -45 dB in power relative to the real signal, though this ratio is still somewhat uncertain due to the large signal strength and potential non-linearity.  We can see these most easily around 1425 MHz in the clean/protected part of L-band when pointing in specific directions, due to strong satellite RFI around 1550 MHz.

We currently do not think this is a correlator artifact because:
  - The copies appear at the same sky frequency and amplitude independent of WIDAR subband BW (for 64 and 128 MHz subbands).
  - They also appear the same when the LO tuning is shifted slightly, thus moving the signal position with the subbands.
  - There is no change in their amplitude when the WIDAR "F-shift" is disabled.  (Other different artifacts do appear, as expected, when F-shift is turned off)

This behavior appears on all four IFs, and on all baselines.  We might want to quantify the antenna/baseline dependence a bit better, but it does not appear to be limited to one or a few specific antennas.  We don't know if this is a new thing or has always been there.  It does require a very strong input signal to become noticeable.

Some plots are attached to show how this looks.  These show "scalar-averaged" visibility data versus frequency from some of our recent experiments, for a variety of correlator setups and tunings.

One explanation that we think fits the data is if an LO has low-level spurs at N*128 MHz away from the main tone.  This seems vaguely plausible to me since the L302's use a 128-MHz comb as part of their signal generation.  Of course others on this list know much more about the internals than I do.  It would be interesting to know if this sort of thing has been measured for the L302 output.  If not maybe someone could run some spectrum analyzer tests?  I'll put in a Jira ticket shortly for those who would rather track this there.

Other suggestions are welcome of course!

Cheers,
Paul
-------------- next part --------------
A non-text attachment was scrubbed...
Name: rfiraster_5x5_exp_A_002.60548.7729225463.spec.png
Type: image/png
Size: 90511 bytes
Desc: rfiraster_5x5_exp_A_002.60548.7729225463.spec.png
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20240830/c5dbf986/attachment-0004.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: rfiraster_5x5_exp_B_002.60548.787796122684.spec.png
Type: image/png
Size: 102327 bytes
Desc: rfiraster_5x5_exp_B_002.60548.787796122684.spec.png
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20240830/c5dbf986/attachment-0005.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: rfiraster_5x5_exp_EE_000.60548.87374113426.spec.png
Type: image/png
Size: 80153 bytes
Desc: rfiraster_5x5_exp_EE_000.60548.87374113426.spec.png
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20240830/c5dbf986/attachment-0006.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: rfiraster_5x5_exp_A_002.60548.7729225463.avg.png
Type: image/png
Size: 121300 bytes
Desc: rfiraster_5x5_exp_A_002.60548.7729225463.avg.png
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20240830/c5dbf986/attachment-0007.png>