[evlatests] T304 RFI and absorber

[Rick Perley]

    I reported yesterday on a crude sniff for the 8196 and '8797' RFI 
spikes after
absorber and a metal can had been placed within the T304 modules for 13A 
and 18A. 
To the limit of sensitivity in 1 MHz, no RFI was seen in, or between, 
these modules.

However, a 1 MHz channelwidth is too coarse for a sensitive sniff for 
RFI, so today
I ran an experiment with 1.5 kHz resolution for these two birdies. 

The correlator mode was 1A, so only RCP is seen from IF 1.  Antenna 18 was
unfortunately out of the array -- apparently to have its C-band feed 
removed for
repair.  I used 512 channels with 780 kHz BW, giving 1.5 kHz 
resolution.  A 7.5
minute integration was utilized for two different frequencies:

    A)  8192.0 MHz -- to pick up the 2nd harmonic of the 4096 LO

    B)  8799.80 MHz -- which by our calculation will put the 5*4096 - 
f(L302) harmonic
near the center of the band.  (Our calculation was right, by the way -- 
the L302 frequency
was 11680.19 MHz, putting the line at 8799.81 MHz). 

    Results:

    A)  The 8192 birdie is overwhelmingly strong in antennas 16A and 
24A, completely
saturating the correlator's spectral dynamic range.  In 14A, the birdie 
is strong, with a
response in 1.5 MHz about equal to that of the total system noise -- 
perhaps 300 Jy. 
For antenna 13A (with the absorber and RFI can), the line is barely 
visible.  This is good
news!   These results were taken from the autocorrelation spectra, as 
the cross-correlations had to be taken with 13 seconds integration -- 
resulting in many turns of phase in general for these B-configuration, 
X-band observations.  Nonetheless, I looked at the X-correlation spectra 
-- the lines are much attenuated on baselines with 16 and 24 compared to 
the signals seen in the autocorrelation, but the overall results are 
consistent with the conclusions above.  For the baseline 13 x 14, the 
line is barely visible above the noise, for example. 

    B)  This RFI spike is not as strong as the 8192, but still saturates 
the correlator's dynamic range for antennas 14 and 16.  In 24, it is 
strong, but non-saturating.  In antenna 13, it is completely invisible!  
In the cross-correlations,  we get comparable results:  13 x 24 shows 
nothing at all, 13 x 14 is barely visible,  for 13 x 16 it is a few 
times the noise, while for the other baselines, it is a few times 
stronger.  The phase rate is quite high for this response (as it has 
both the natural fringe rate, and that backwards moving phase of the 
L302 on it), so these cross-correlation spectra are greatly attenuated 
by the 13 second integration. 

    Conclusion:  The presence of the absorber and the metal shields has 
reduced these RFI spikes, for antenna-IF 13 at least (and probably for 
18, based on the low resolution data) to near tolerable levels.   A 
small improvement in shielding/absorption might well do the trick, so we 
can avoid a costly and lengthy re-design. 

    Further action:  We need 18A back, so we can do equally sensitive 
tests with it, and it would be interesting to see if absorber/shielding 
is effective on antenna 24A, which has a HUGE 8192 RFI spike. 

    PS  There is clearly a very large delay error on antenna 24:  50 
degrees of phase slope across 700 kHz bandwidth!  About equal to the 
delay change found by Ken yesterday.