Jumps in interim correlator sbc 2 polB

march 03

Experiment A1726 used lbn and the interim correlator with 1 sbc per board and 2048 channels. The observing frequency was 1300-1350 so the radar blanker was used. The correlator configuration was:

sbc1 - polA
sbc2 - polB
sbc3 - polA same as sbc1
sbc4 - polB same as sbc 2

240 second on/off position switching was done. For two source jumps in the spectra were seen on different nights. The jumps in one source repeated the second night.
The plots show the total power versus sample (1 second) for the 240 on points and 240 off points. There are 10 spectra (out of 250+) that have trouble.

Fig 1-2: The 0 lag for 10 pairs. The vertical scale i linear in the total power. The black line is sbc 2.The red line is sbc 4 which has the same signal as sbc 2. The green line separates the 240 on position samples from the 240 off position samples. The large jumps are about a db.
Fig 3-4: The power counters for the 10 pairs.

The 0 lag (figures 1,2) are derived from the digital filters and then correlator board. The power counters are VtoF converters in front of the digitizer that look at the entire 50 Mhz bandwidth.
The 0 lag data is blanked by the radar blanker while the power counters are not.

The source of the problem can be narrowed down by:

Figure 3 top shows a jump in the power counters at the same place that the 0 lag jumps. This implies that the problem is before the digitizer/digital filter/correlator board.
The problem does not occur in the red trace (sbc 4) which is the same signal as sbc 2. This implies that the problem is at or after the downstairs mixer for sbc 2.
Figure 1 top: the jump is about 20%. This is close to the 1db steps of the programmable attenuator for sbc 2. Figure one bottom has jumps that are less than this.

Possible causes of the problem could be:

The programmable attenuator for sbc is jumping. It could also be the ttl drive that programs it. The stable 1db jumps could be caused by this. The smaller jumps could occur if the attenuator jumped by 1 db for a fraction of a second and then jumped back. If this was true i wouldn't expect the fraction jumps to remain for many seconds.
A component in the chain from mixer 2 to digitizer is failing. The amps there are surely being driven into saturation by the radar (this is why the power counter data has so many negative going spikes). It may not be recovering within the 400 usecs that the radar blanker blanks the correlator. I'd expect the levels to jump around a lot more if this was the problem.
There is a problem of lo feedback back on the 750 to 260 Mhz mixers with this setup. The same lo is applied to all 4 mixers. Some of the lo goes out the mixer input, through the splitter and back into the input of the other mixers. This is the same frequency as the lo so it gets mixed to dc. The IF gain of the mixers is changed when there is a dc output. This doesn't explain why it only happens occasionally (unless of course it is a function of the IF strength). If this was the problem, the changing the frequency of the 4 sbc by a small amount should make the problem go away.

There is still the question of why the problem is limited to the two sources. It is also strange that it repeated with the same source on two separate nights (it can't be source related since sbc 4 has the same sky signal as sbc 2). Maybe this az,za direction is pointing toward a maximum in the radar transmission into the dome...

processing: usr/a1726/badsbc.pro

 home_~phil