[evlatests] Tests July 29

[Barry Clark]

When arrived, no sync det, no fringes.  Turns out samplers were off, 
because the F rack had been turned off in the interim.

With the samplers turned on, IF C is effectively dead.  Seem to be a 
lot of errors on one color.

Checked delays.  A & B pretty reasonable, D out to lunch.  Looked
at D bandpass; it, too, is out to lunch.  It was double humped, and
with 50 MHz bandwidth, alternate points had opposite phases.

A fringes were week, but amplitudes were stable, though phases seemed
to jump around.

Wanted to investigate proper settings of levels.  The RF power detector
in IF A was not working (reading -INF at all times).  And, as noted, A
fringes were weak, and therefore a little suspicious.  However, I looked
at the effect of swapping attenuation between the RF attenuator and the
output attenuator (measurements made with Tsys correction turned off, so
they are close to a real SNR):

A_RF_Att A_RF_DET A_IF_Att A_OUT_DET Fringes

8        XX       15       -32.1     0.78
11       XX       12       -32.1     0.78
14       XX        9       -32.2     0.80
17       XX        6       -32.0     0.80
20       XX        3       -32.0     0.77
23       XX        0       -31.9     0.81

The fringe amplitude is gratifyingly independent of these changes.  There
is no evidence of added noise from the fron-end amplifier over this range.
This was confirmed on IF B.

B_RF_Att B_RF_DET B_IF_Att B_OUT_DET Fringes

8        -30.7    20       -34.9     4.60
23       -36.0     5       -34.6     4.68

No added noise down to -36 dbm on the RF power detector.

Next I looked at swapping attenuation at the output of the T304 with the
GAIN1 of the deformatter.

RF_att = 8
A_IF_Att A_OUT_DET LinkRMS FIR_G1 Fringes
12       -30.2    11.0     45     0.77
15       -31.9     8.3     64     0.72
18       -33.8     5.6     90     0.85
21       -34.8     4.0     128    0.82
 
RF_att = 8
B_IF_Att B_OUT_DET LinkRMS FIR_G1 Fringes
11       -27.7     16      22     4.30
17       -32.9     7.6     45     4.48
23       -36.3     4.0     90     4.42
 
There seems to be a modest preference for having the RMS on the link be about
six or seven, which is to say, about -33DBm on the output detector.  This
is, gratifyingly, about the same on both IFs.

At this point we switched from 3C 84 to 3C 273.  Fringes went away.  Very
weak fringes appeared on IF B, which became resonably strong either on their
own, or after a Modcomp skip (!).  Thereafter, we had only IF B (very weak
fringes on IF A....).  Switched to L band because phases were getting a little
unstable, and I didn't want the complication.  Next tried swapping gains 
on the outputs of the two FIRs in the deformatter.

With RF_att=8, Output_att=17, output TPD=-33 Link RMS =8.5
GAIN1    GAIN2    Fringes
128      256      2.33
64       128      4.29
32       256      3.92
16       512      3.12
 
This one seems to be a little touchy.  Setting GAIN1 too high presumably
results in loss of SNR due to overflow in the resampler.  Setting it too 
low looses SNR in the resampler, starting to make it look like a one-bit
machine.

I've said before that there seems to be some sort of time problem at antenna
16.  All the above measurements were made with Walsh functions turned off.
Looking at the baseline 8-16 this gave an amplitude of 8.91.  Turning
on the Walsh function, and looking at amplitudes with various offsets gave
Walsh offset  Fringes
-0.0003718    4.15
-0.0003711    3.63
-0.0003704    4.87
-0.0003697    4.68
-0.0003690    3.82
-0.0003683    3.18
(Antenna 14 correlated perfectly with the offset of -0.0003697 or thereabouts.)
This does not look like a simple time offset of a heartbeat or two.  I 
suspect problems with the heartbeat pulse to the L302s (at least L302-2),
or to the MIB therein.

Did some playing around with driving VLA antenna 10 with the Executor.  
Getting closer, but still not quite there (problems with Walsh function
there too :-).