[evlatests] X-Band Birdies

[Rick Perley]

    New, better shielded T304s were put into antennas 13C and 18C.  We 
believe that 13A and 18A have the earlier shielding/absorbing T304 
modules. 
    I ran a test today to check on the strength of the 8192 and 
3*4096-f(T302) birdies.  Due to a blunder on my part, the latter 
frequency was tuned only for IF A. 

    First a note on what we're looking for.  The requirements are for 
any artificial signal (`RFI') to be at a level 10 dB below the rms noise 
power, when observed with a spectral resolution equal to a velocity 
resolution of 1 km/sec, and after 9 hours' integration.  This 
requirement applies equally to a single antenna and an interferometer 
array. 

    At X-band, where these tests were done, the frequency resolution 
corresponding to 1 km/sec velocity resolution is 33 kHz.  The tests were 
done with a resolution of 1.5 kHz, and with 30 seconds integration.  
With this combination, the noise in an autocorrelation spectrum (which 
is proportional to sqrt(BW/time)) will be a factor of about 7 higher 
than with the time and resolution appropriate for the requirements -- 
hence, to meet the requirements, the RFI should be a factor of 7 * 10 = 
70 below the observed noise level in the autocorrelation spectrum.   
Clearly, if the tests show the RFI spikes at all, then the shielding is 
insufficient. 

    Unhappily, the 8192 RFI spike is easily seen in IFs A and C for 
antennas 13 and 18.  (For other IFs of those antennas, and for antennas 
14 and 16 in all IFs, the 8192 spike is overwhelmingly strong). 

    In antenna 13A, the 8192 MHz spike is seen at a level about 100 
times the noise in the spectrum, while for 18A, it is about 40 times the 
noise.  We hence would need to have a further reduction of about 38 dB 
to meet the requirements. 
    For 13C, the 8192 RFI spike is also about 100 times the noise, while 
in 18C, it is decidedly weaker -- about 10 times the rms noise.  So for 
these, 38dB and 28 dB would be needed to meet the requirements. 

    As noted earlier, I obtained proper spectra for the other RFI spike 
only for IF A.  For this, the news is good -- it is not visible -- so 
we're at least within ~18 dB of the requirement. 

    The requirement noted earlier is for a 'single dish'.  In 
interferometry, we get some help from fringe rotation -- the amount 
ranges from a few dB to many tens of dB, depending on configuration, 
frequency, and source declination.  We should not rely on this for 
resolution of our problem -- we'll be using the array at low frequencies 
and in D configuration with short spacings while observing a far 
northern source -- the fringe winding attenuation is likely no more than 
10 dB. 
    We also get help if the RFI source is incoherent, with respect to 
the target source's phase -- that is, if the RFI phase seen on each 
baseline, at each moment, is randomly distributed.  If so, another ~13 
dB of attenuation is obtained.  This condition might be met for external 
RFI -- which enters through the antenna sidelobes, and has its phase 
messed up by atmospheric propagation -- but it's unclear to me if it 
applies to internally coupled RFI, such as is our case here. 

    Conclusion:  The 3*4096 - L302 coupled RFI is probably low enough to 
be close to the requirement.  The 8192 MHz RFI spike is at least 20 dB 
above the desired level.