[evlatests] S&R failure statistics

[Vivek Dhawan]

Ken and I looked at this a bit. We had Rick's example from Apr 28,
which changes bands a lot, plus a couple of scripts specially made
for X band with all 8 samplers overlapped at 8-10 GHz (X3lo) and
another for 10-12 (X3hi).

The issue is clearly visible in Psum values (visibilities not needed)
during the couple of minutes that the attenuators, equalizers, and
station-board servos are settling, just before the requantizers kick
in. Also at the 3-bit signal levels in the T304s. other things could
be examined as discussed in today's meeting, but not done yet.

Conclusions so far:

Xhi has lower power levels than Xlo and most other setups. This makes the
optimum attenuator values different by 6-10dB. When switching from another
band, the initial att values quite far off on some antennas/basebands.

On most occasions, over 99%, (script/antenna/FE/baseband) the S&R manages
to get close enough that the loss of SNR is not noticable. In some cases,
esp. Xhi, especially changing from Ku band, especially on ea12 L, the S&R
does funky things and a loss of SNR is apparent.

I'm not ready to recommend any changes yet. One global change might be to
boost the X band front end power by 6-10dB (it is the FE with the lowest
gain). ea12 might be the first one to try.

But: there is more going on. The S&R loop clearly has quirks, which can be
exposed by input conditions.

To be continued.

On 05/09/2018 02:07 PM, Rick Perley wrote:

> Â Â Â  1) Distribution over bands/IFs.
> 
> Â Â Â Â Â Â Â Â Â Â Â  Although these failures (defined here as the power to the 
> sampler low by a factor of two or more) occur at all bands and IFs, they are 
> uncommon *except* in the 10 -- 12 GHz IF band. (Uncommon is defined as five 
> failures or less out of the 27 antennas).  In the 10-12 GHz band, more than 
> half the antennas fail.
> 
> 
> Â Â Â  2) Distribution over time.
> 
> Â Â Â Â Â Â Â Â Â Â Â  The statistics failures are the same over time for the 
> thee runs.  However, it is not true that the same antennas fail each time. 
> Indeed, there is little to no repeatibility in which antennas fail to have their 
> powers set.
> 
> Â Â Â  One other point -- Xband Hi (10 -- 12 GHz) not only fails far more 
> frequently than any other, but the severity of the failure is also very much 
> higher.  I have an example of an antenna whose power was low by nearly a factor 
> of 10,000!