16khz comb in 327 receiver.

27aug05 may05: Initial report
aug05: the 16Khz comb is coming from the alfa motor contoller.
01sep05 Grounding motor shield 29sep05 improved things but then deteriorated.

may05: Initial report

A 16 khz comb was seen in the 327 receiver. It was reported by p1979 and A2057 in may05.
On 03jun05 data was taken with the 327 receiver to see what may have been causing it.

03jun04 data:

Data was taken with the interim correlator and the ri.

a 300 second integration with the 16 khz comb (.ps) (.pdf) . The interim correlator was set to 190 hz/channel (after hanning smoothing) and 1 second dumps. The comb is strong in polA and very weak or non existent in polB.

high frequency resolution data set (.ps) (.pdf) was taken by base band sampling a 200 kHz band for 48 seconds. The top plot shows the spectra smoothed to 1 hz resolution. The bottom plot is a blowup with .02 hz resolution. The hires spectral width shows that the birdie has a width of around 1 hz (so it is not a harmonic that is locked to the station clock).

Comb strength vs azimuth (.ps) (.pdf) : The dome was placed at 18 degrees and then the azimuth was swung from 270 to 660 degs and back. The top plot shows the average of the comb channels (white, red) and the adjacent channels (tsys). The bottom plot is the comb strength (comb channels - offCombChannels). There is no apparent azimuth dependence in the strength of the comb.

We then did some test to see where the comb was coming from. We did the following:

Turn the dome cameras on/off. This did not affect the comb.
Open/close the sband transmitter door, open/close the turret room door. The comb did not get stronger or weaker.
Turn off the compressor monitor boxes above the dome. The comb did not change.
Stop the dewar monitoring, The comb did not change.
Turn off the room air conditioners. no change.
Open the turret controller huffman box in the turret room. The comb did not change.
open the iflo racks in the turret room. No change in the comb.
move the first lo by 3 khz. The comb stayed at the same rf frequency so the comb is entering the system before the first mixer (probably through the horn).
Look for the comb with the gregorian 430 Mhz receiver. The comb was not seen.

P1979 data (may05):

P1979 used folding mode on the wapps with 1.56 Mhz Bw and 1024 lags. Typical integration times were 600 seconds. 4 separate frequency bands were taken and a frequency hop of about .5 Mhz was done between files. For each 600 second data file, all of the spectra were averaged for each frequency separately (4 freq x 2 hops). This data was taken between 21may05 and 31may05.
The first set of plots are dynamic spectra images that were made using the average spectra from each file. The vertical axis is data file. This is increasing in time but it is not evenly spaced since the data were taken on different days:

polA freqHop 1 The redline is the strength of the birdie at 325 Mhz.
polB freqHop 1
polA freqHop 2
polB freqHop 2

One thing that stands out is that the 16 Khz comb is strongest when the 325 Mhz birdie is strongest (the red line on the left of each image shows the strength of the 325Mhz birdie.).

The next set of plots shows the relative strengths of the birdies:

Overplotting the spectra for polA freqHop 1 (.ps) (.pdf) Note that the different boxes are scaled differently. The 325Mhz birdie is 20 times Tsys (in 1.5Khz channels).
A blowup around the strongest 4 birdies (.ps) (.pdf) These plots show the strongest 4 birdies. The vertical dotted lines are the comb (15.987 Khz centered at 324.999 Mhz). None of the other birdies align with the 16Khz comb (the 321 Birdie is close). The combs in the 4 bands are all aligned with the 325 Mhz birdie.

summary

The 16Khz birdie is actually 15987 Hz
It is strongest when the 325Mhz birdie is present but is seems to always be there at a low level.
It is not coming from the cameras. Stopping/starting the dewar monitoring did not affect the comb.
the comb is about 1 hz wide so it is not a birdie locked to the station clock.
The comb is coming in before the first lo since it stays at the same rf frequency when the first lo is changed.
The 32Khz spacing is stronger than the 16Khz spacing.
There is no strong azimuth dependence of the comb.

processing:x101/050603/docor,dori,dowapp.

aug05 the 16Khz comb is coming from the alfa motor contoller: (top)

The 16khz birdie in the 327 Mhz receiver was initialing reported in may05. Data taken in jun05 showed that the comb was about 1 hz wide. On 24aug05 we searched in the dome using a portable spectrum analyzer. The setup was:

tektronix sbt250 spectrum analyzer with an input temperature of 720K and a minimum rbw of 100 hz.
A small loop antenna that could be used as a probe when close to the source (for local consumption we used the larger loop. The round ball probe was more sensitive but gave little directionality).
We looked at the two comb members; 331.5903 and 331.6542 (since they were the strongest). To not get fooled we required both of these to be visible

    Sitting at the equipment rack for the 327 receiver we saw a 2-3 db (above the noise floor) birdie at these two freq. After sniffing around it looked stronger at the cables that went over to alfa (this rack also contained the control cables that went to the alfa filter bank).
    The signal was stronger at alfa. We then opened the alfa motor controller rfi chassis and the signal increased. When we turned off the amplifier/controller the birdie went away. This chassis also contains the cryodynes for the alfa dewars. Turning these off made no difference.
    Down in the lab we setup the spare amp/controller in the screen room and looked at its rfi output. We saw the 16khz comb at 331.5903 (may 5db above the noise floor). Opening up the span we also saw a 40Mhz comb coming from the amp/controller. This comb goes out beyond 1400 Mhz. It was 20 to 30 db above the spectrum analyzer noise floor.
    Switching to a span of 50 Khz around 320Mhz (40*8) we saw the 40 mhz comb plus the 16Khz comb centered on the 40 Mhz. The comb was a lot stronger than at 331.5903 Mhz where we had been looking. You could see the amplitude of the 16Khz comb fall off as you got farther away from 320Mhz.
    The specs of the amplifier/controller show that:

The current and commutation frequency is 16Khz.
There is another 2khz frequency that is uses.
My bet is that the dsp in the amplifier runs at 40 Mhz.

To verify that this all hung together, we went back on 25aug05 and took some data with the interim correlator. The setup was:

327 Mhz receiver centered at 320.01 Mhz. The interim correlator can give a birdie at the enter of its band so we used the 10Khz offset to stay away from this.
The interim correlator was setup for 90 Hz resolution.
We took 65 seconds of data in a normal observing mode. The alfa motor controller was on and the rfi shielded chassis was closed. The telescope was sitting so the sky was drifting by.
The motor controller and the cryodynes inside the rfi shielded chassis were turned off and then the chassis was closed. Data was then taken for about 55 seconds.

The plots show the 40Mhz birdie and the 16 Khz birdie (.ps) (.pdf): The first page has a 195Khz span. The second page is blownup to a 40 Khz span (both of channel widths of 90 hz).

Fig 1 top; 195 Khz around 320.01 Mhz. The black plot is with the controller on. The green line is with the controller off. The birdie at 320Mhz (40*8.) that is 180 times Tsys goes away when the controller is turned off.
Fig 1 center: The vertical scale is blown up (0 to 10*Tsys). The red dotted lines are spaced every 16 KHz centered on 320.Mhz. The odd harmonics are strongest and the strength falls off as you move away from 320Mhz.
Fig 1 bottom: The vertical scale is blown up again (0 to 2*Tsys). You can see a lot more high frequency rfi when the motor controller is on (black).
Fig 2 Top: The horizontal scale is limited to 40Khz around 320.01 Mhz. The red dotted lines are spaced every 16 Khz.
Fig 2 center: The vertical scale is now 0-10*tsys. You can see the 320.01 birdie from the interim correlator digital filters. There are some other birdies that are there with both the controller on and off.
Fig 2 bottom: The vertical scale is 0-2*tsys. The red dots are spaced every 2 Khz. There is a 2 Khz comb that is also centered on 320Mhz. This is the other frequency used by the controller.

The encoder and motor cables that enter the rfi box are feed thrus. They do not have their shields tied to the box as they enter the box (even though the feed thru connector has connections for this). This may be causing the problem.
To see whether any extra shielding works, we should concentrate on the 320Mhz comb since it is 25 db stronger than the 16Khz comb.

I also took a look at all of the 40 Mhz comb members that were inside the lbw receiver. Using a 90 hz resolution and 60 second integration, i did not see the comb present.

processing: x101/050825/doit.pro

01sep05 Grounding motor shield 29sep05 improved things but then deteriorated.

    The alfa motor controller amplifier is causing a 16khz and 2khz comb in the 327 receiver. It is centered at multiples of 40 Mhz (the dsp clock?? of the motor controller). The motor power and encoder cables pass directly into the rfi shielded box that contains the amp. Their shields are not grounded to the box where they enter the box. On 29aug05 the shields for these cables were grounded to the rfi box. This decreased the signal by about 10db. On 01sep05 the 327 receiver was checked again. The grounded shields were no longer as effective as they were on 29aug05 when they were installed.
    The plots show the strength of the birdies around 320Mhz (.ps) (.pdf):
    The units are Tsys. The resolution bandwidth is 190 hz (after hanning smoothing). The red lines are the initial birdies from 25aug05 with no shield grounding and the amps on. The dashed red lines show the locations of the 16Khz comb about 320Mhz.

Top: 25aug05 the birdies before the shields were grounded. The amp is on. The 320 Mhz birdie is about 200 times Tsys.
2nd: 25aug05 before shields were grounded. Red is amp on, white is amp off. You can see the 16Khz birdie disappear into the noise. The birdie at 320Mhz went from 180Tsys to 2Tsys. This shows that there is another source of 320Mhz than the motor amp.
3rd: 29aug05 after the shields were grounded. The amp is on. The 320Mhz birdie is 10 times Tsys. The 16khz side bands are not visible. We did not turn off the amp, so we don't know if the 320Mhz birdie strength if from the amp or the "other source" seen on 25aug05.
4th: 01sep05. The shields are grounded. The 320Mhz birdie is now 200Tsys. This is stronger than it was on 25aug05 before shielding. We did not turn off the amps so we don't know whether this is coming from the amp or the "other source".
bottom: 01sep05. A blowup of the 4th frame. You can see the black 16Khz side bands of 320Mhz again. They are .5Tsys. So the grounded shields are not working as well as they did on 29aug05 (or the signal inside the cabinet has increased a lot..).

processing: x101/050901/doit.pro

home_~phil