Correlator polB sbc 1 gain drifts

march 04

a1705 up to 15apr03
list of all files with drifts 15apr03 to 27may03

Resolution: 01jun03

The 260->baseband mixer chassis was tested on the bench and the tubular lowpass filters (k&E) (100 mhz) after the mixer were causing the problem. We cooled all 8 tubular filters and 1b,4a,4b showed large gain variations. Removing these filters, we saw that the impedance between the center conductor and ground varied from 1 Mohm to 2 ohms. Opening the filters we found that the mylar sheet that was wrapped around the filter core and acted as the dielectric material was causing the problem. It is a flat sheet that wraps around the filter core leaving a thin line where the filter could be tuned. At the ends of the filter, the gap got too large and the filter core (actually the capacitor lugs) were getting close the outer case and shorting. As the temperature changed, the impedence would change and cause the gain variations that we were seeing. It looks like this is a basic design flaw with the filters. We ordered a new set of 8 filters and have now become suspicious of these "inactive" devices. All 8 filters were replaced on 27jun03.

a1705 and if2 noise source (thur 15apr03)

Experiment A1705 was experiencing drifts in polb sbc 1 of their data. It was centered near 1400 Mhz.
The plots show the 0 lag power for the entire night. The large jumps are when the attenuators were changed. The narrow spikes are the cals. The slow variation that occurs in 1 pol and not the other is the problem. The gradual rise in all sbc is continuum.

Fig 1 has the 0 lage for the 4 sbc. Black is PolA, red is polB
Fig 2 plots polA-polB. black-sbc1, red:sbc2, green:sbc3, blue:sbc4.

A1705 data

This is the data from A1705.

Noise source data:

This data was taken with the if2 noise source inserted. It is 1 second dumps just like the a1705 data.

09apr03 shows drift of sbc1B
10apr03 shows drift of sbc1B with the outputs of the IF2 switched sbc1B, sbc2b. The drift remains in sbc1B so the problem is after this switch (in the correlator).
14apr03 The cables leaving the attenuator box in the correlator were switched (sbc1b and sbc2b). The drift remained in sbc1b. This shows that the problem is the cable switched or the mixer box that the cables go to. At sample 4600 i retightened the switched cables. At sample 5500 bill sisk retightened these cables. If the problem was a loose cable, then after tightening the cables, it should have gone back to the maximum gain. Instead it stayed about 30% below the maximum value. The power counters show this same drift. They also come out of the same mixer box. This implies that the problem is a component in the mixer box sbc1b.

Here is some of the evidence that has been gleaned(in time order)

It is a change in gain, not noise. I computed the calon/caloff -1 for the entire night and this value remained constant. If noise had been added then it would have changed (assuming the cals are constant).
It is only polB of sbc1 so the problem is at or after the 750 to 260 downconverter in the downstairs if/lo rack for this sbc.
It also appears in the power counters.This says that it is before the analog signal is split sending 1 to the power counters and 1 to the a/d converter. This rules out a/d converter, digital filters, correlator board, etc...
The slow drift looks like a temperature variation either a component or a cable that is not making contact or has a center conductor pin that does not stick out far enough.
I checked all of the external cable connections and the problem is still there.
The drift occurs with corr attenuator values 6,7 8 (which were the main ones used) so it is not a single value on the cor attenuator.
The drift occurs across changes of the cor attenuator change. So it's probably not the attenuator individual settings ( recant this below...)
Running off of the noise source I saw a drift in sbc 1B when the output of the iflo rack cables had been switched: sb1a and sbc1b. This says that the problem is not in the rcvr room iflo racks (since the problem didn't switch).
I realized later that the attenuator steps we use are 1db , but the attenuators are .5 db step attenuators with the lowest bit grounded. Much of the variation that we see is between two fixed levels. It is stable, it drifts, and then it is stable again. This was also seen with the noise sources. The difference between the two levels is .5 or 1 db. The data from 09apr03 noise source is very close to .5 db. The times i saw it drift during attenuator changes, could have been the .5 db bit changing.
14apr03 jump with the output of the attn box switched showed that it was not the attenuators!!. The jumps did not switch. Must be some component in the mixer chassis. Opened the mixer chassis and tightened all of the cables in the chassis. The cable on the inside of the bulkhead bnc feed through was loose. Also replaced the outside cable from atten box to mixer chassis for sbc1b.
15apr03,16apr03,18apr03,21apr03,22apr03,22apr03,23apr03,25apr03... Ran the correlator during the day for many hours (> 4hours/day). Saw no more large gain changes in sbc1b. Must have been the loose cable or the outside cable that was removed.

processing: usr/a1705/apr03polBDrift.pro

After 15apr03

The problem went away and then reappeared. Here is a list of all of the observations that have had drifts 15apr03 thru 27may03. There are 123 files that were checked. 23 files had gain drifts. Some of the files measured one polarization per board so we would never see a 1b drift in their data (eg. a1708).

file	drift
corfile.09apr03.if2noise.1	1b
corfile.10apr03.if2noise.1	1b (1)
corfile.14apr03.if2noise.1	1b (2)
corfile.09may03.x101.3	4a
corfile.12may03.a1312.1	1b
corfile.13may03.a1457.1	1b
corfile.15may03.a1708.1	4a
corfile.16may03.a1703.2	4a
corfile.16may03.a1708.1	4a
corfile.16may03.reu03.2	1b
corfile.17may03.a1312.2	4a
corfile.18may03.a1703.1	4a (3)
corfile.19may03.a1708.2	4a
corfile.19may03.if2noise.1	4a
corfile.20may03.a1708.1	4a
corfile.23may03.a1703.1	4a
corfile.24may03.a1545.1	1b
corfile.24may03.a1703.1	4a,1b(3)
corfile.24may03.a1789.1	4a
corfile.24may03.x111.1	4a
corfile.25may03.a1703.1	1b,4a(3)
corfile.25may03.a1708.1	4a
corfile.26may03.a1703.1	1b*

if2 rack outputs 1b,2b switched. if2_1b -> corInp_2b, if2_2b->corinp_1b. The switch was done at the output of if2 rack. The problem stayed in cor_1b, so the problem is not in the if2 racks.
The correlator attenenuator chassis outputs attn_1b,attn_2b were switched at the output of the atttenuator chassis. The problem stayed in _1b so the problem is downstream from the attenuator box (cable, mixer chassis, etc). The power counters also drift so it is before the split of the power counters in the mixer chassis..
A large fraction of the datafile had gain drifts...

processing:usr/a1705/sbc1bdrift/driftmay.pro

 home_~phil