A1459 3 min on/offs with lots of baseline ripple.

07sep01. The baseline ripples (vs freq) and the temperature variation of the room.
08sep01. Turret room temperature variation.
08sep01. Check cband, lband narrow, lband wide for jumps in total power.
14sep01. Total power jumps coming from new airconditioner compressor.
20sep01. lbn Image while turning air conditioner on/off.
28sep01. lbn image with magnetic contactor installed on new ac compressor.
05oct01. Turn louvre motor, blower,  Ac compressor, 208Volt breaker on/off.
31oct01. Timer module for compressor causing the problems.

  A1459. The baseline ripples. (topa1459)

    polA plots (5mb ps file) and polB plots (5mb ps file) show on/off-1 in kelvins for each day. For each day there are 4 frequencies. Each strip has been offset from the previous by .2 kelvins for display purposes. The baseline ripple is:
 

date	polA	polB
30aug01	bad	bad
31aug01	better	better
01sep01	better	better
02sep01	better	better
03sep01	better	better
04sep01	bad	bad
05sep01	bad	bad
06sep01	bad	bad
07sep01	bad	bad

Bad means that there was ripple in a large fraction of the observations. Some of the better days had large ripple in a few strips. The ripple was on the order of the  strip offset (.2 Kelvins).  The 25 Mhz integration over 180 seconds with hanning smoothing (and no addition of polarizations) should have given 30K*sqrt(2)/sqrt(25e6/1024*2*180)=.015 Kelvins. Some properties of the ripples were:

1. 30aug, and 04sep->07sep were worse than 31aug->03sep.
2. The ripples are in all 4 sbc so it is probably in or before the 1st IF (or in the correlator).
3. It is on both polarizations so it is not a single amplifier/mixer in the 1st IF.
4. The ripples in polB are shifted about 2.5 Mhz down from the same ripple in polA. Look at 07sep01 1396.4 polA (6th strip from the top) and 1394 polB (6th strip from the top).
5. We took data at cband tracking blank sky. We took adjacent 3 minute sections and computed on/off-1. 10 of these on/off-1 are shown in the plot (each offset for plotting). The vertical scale is Tsys rather than Kelvins. There is a 1 Mhz standing wave (since we did not do position switching) but the rms is < .001*( tsys=30) <. 03 Kelvins. This is close to the theoretical value and an order of magnitude less than the bad ripples at lband. This rules out the correlator as the problem. The setup differences were the 1st IF (lb 750Mhz if, cband 1.5 Ghz IF), dewar/postamps, and frequency.
6. The ripples are across all of the 100 Mhz. If it is rfi , it must be outside the IF bandpass and driving something into saturation. It could be a noise diode from one of the other receivers that is firing when it isn't supposed to (we had this problem originally when the ttl cal triggered all of the receivers at the same time way back in 199?).

Turret room temperature variation. (topa1459)

    A new air conditioner was added to the turret room in early aug01. There are two of them set at different thresholds. The plot shows the room temperature 01aug01 (bottom) through 08sep01 (top) (2mb psfile). Each day has been offset by 2 degF for plotting. The new second air conditioner was installed around 8aug01. With the new air conditioner, the temperature cycle is now about 1 degree over 10 minutes.

Total power variation of cband and lband narrow with temperature. (topa1459)

1. The cband data showed no jumps in power.
2. Figure 1 top has the lband narrow total power samples taken  on 08sep01. The lower plot is a blowup after averaging all 4 sub correlators (freq) for each polarization. Near the bottom of each temperature cycle there is a jump of about .3% Tsys down by all sub correlators and both polarizations. About a third of the way up from the bottom of each temperature cycle there is a corresponding jump upwards of about .3% Tsys. With a Tsys of  28K this is about .1 Kelvins. Since there are two air conditioners in the room, it looks like one of the air conditioners is turning off causing the downward jump. The temperature rises, the air conditioner turns back on  causing a positive jump in the total power. The two air conditioners have their turn off thresholds set to 73.4F (23 C) and 71.6f (22 C). Their turn on value is probably a fixed value above this. The temperatures differ from those plotted since the room temperature sensor is located in the center of the room while the air conditioner sensor is measured at the air intake. On top of the discreet jumps, there is the same temperature variation of gain as cband.
3. Figure 2 is a blowup of 800 seconds of figure 2 data 72800 through 73600 seconds from midnight. Plotted is sbc 1, polA  (total power-median(totalpower))*500. The temperature is over plotted in color. An image was made (200kb gif)  of the 800 spectra that correspond to these total power points. The spectra was bandpass corrected using the average bandpass. The vertical axis on the image (seconds from midnight) corresponds to the horizontal access of the line plot.  Looking at the image you can see that the jumps are not constant in frequency.  There are .5 mb wide stripes distributed across the band.
4. Figure 3 has lband wide total power samples. The receiver had the hybrid out and only pol A is plotted (pol B has a large 100 second oscillation).  The 1%  per degF power fluctuation is also present in this receiver.  You can see the 100 sec oscillation in this polarization as well. There may also be a .3% jump when the temperature cycle goes through the minimum but it is difficult to say because of the other junk going on.

14sep01 total power jumps while air conditioner power cycled. (topa1459)

    On 14sep01 blank sky was tracked during the day with the lband narrow receiver. A 25Mhz bandwidth was sampled at a 1 hz rate close to 1418 Mhz. While the data was being taken the two airconditioning units had there power turned on and off (at different times) to see how they affected the total power.

The figure shows the results. It plots the total power versus time. The colored vertical lines are:
 

vertical solid red. new airconditioner turned on.

vertical dashed red. old airconditioner turned on.

vertical solid green. new airconditioner turned off.

vertical dashed green. old airconditioner turned off.

 

 

These are split level units with the consoles in the turret room and the compressors on the service platform. The power to the consoles was turned on/off with a remote unit. Between 53 and 80 seconds after the new Ac was turned on, the rf power jumped by .3%. This did not occur with the older unit. The new unit would then turn it's compressor off before we shut off the power. It turns out that the evaporator unit was icing up so it was not running continuously (even though the temperature in the room was 90 F).
The jumps in the lband narrow receiver are caused by the new Ac compressor running.
processing: usr/a1458/doall.pro,x101/010908/tpjumps.pro,lbnimage.pro,x101/010914/aconoff.pro

20sep01 lbn image of air conditioning on,off (topa1459)

On 20sep01 during the day the lband narrow receiver was used to take data while the new airconditioner unit was turned on and off. 1024 channels were taken over a 25 Mhz bw at 1 second integrations. There were 1060 seconds of data. The image was flattened in frequency by  using the time average of the data. It was flattend in time by using 50 channels away from the interference. The image shows on/off-1 versus frequency and time. There are 3 sets of dashed lines added to the plot:
• short dashes. The remote control was used to turn on the airconditioner console.
• midsized dashes. A person watching the compressor heard it come on.
• long dashes. The remote control was used to turn off the console and compresssor.
Using 1417 Mhz as a reference you can see that:
1. At 36 seconds the console was turned on.
2. At 50 seconds the compressor was heard to come on.
3. At 100 seconds the console/compressor turned off. The image becomes dark implying that the power has gone done.
4. At 150 seconds the console was turned on. No rfi yet.
5. At 190 seconds the rfi appeared but they did not hear the compressor turn on.
6. At 400 seconds the console/compressor was turned off and the rfi went away.
7. At 450 seconds the console was turned on. No rfi yet.
8. At 590 seconds the compressor was heard to come on and the rfi appeared.
9. At 750 seconds the console/compressor was turned off and the  rfi went away.
10. At 840 seconds the console was turned on. No rfi yet.
11. At 940 seconds the rfi appeared but they did not hear the compressor start.
12. At 1060 they heard the compressor come on.

So the rfi is always turning off with the console/compressor. Twice the rfi started when the compressor started. Two times the rfi started without hearing the compressor turn on. The other air conditioner was running at the time. I doubt that the rfi is fromthe old air conditioner since it was not turned off when the new unit  was (and the rfi always disappeared).
processing: x101/010920/lbnac.pro

28sep01 New magnetic contactor on ac installed. (topa1459)

The airconditioner compressor came with a magnetic contactor to start the compressor.  This contactor was defective on the new ac compressor and it was replaced with a solid state (SCR) switch. The SCR switch could have been generating noise so on 28sep01 it  was replaced with another magnetic contactor.  The image shows lbn tracking blank sky. The image processing  and embedded horizontal lines are the same as the 20sep01 image above. For this experiment the old air conditioner was shut off.
• Small dots are console on.
• The next largest set of  dashes is compressor on
• The next largest set  (8 dashes/page) is everything off.
• The area between the largest dashes (2 dashes per page) was used for the bandpass correction.

 

 

There are ripples that start right after the console turns on and go off when the compressor goes off so it is not just the compressor. The area 800-1000 seconds looks the flattest since it  was used for the bandpass correction. Since we were tracking blank sky, as we move away from this location the normal standing waves increase. This image in idl shows the start/stop at the console on/off a lot more clearly.

The figure shows  total power versus time for the above image (and two other frequency bands). Three different  25 Mhz bands are over plotted (1420,1400,1380).  The vertical scale is  totalPower[i]/median(totalpower) - 1.  The vertical colored lines are:

• Red - air conditioner console on (small dots in image).
• Green- air conditioner compressor comes on (medium dashes in image).
• Blue - console/compressor off (8 dashes per page in image).
All three separate bands jumped up 10 seconds after the console came on. They all jumped back down when the console/compressor went off.  The compressor did not come on till much later so the problem is not in the compressor.

The next thing to check is the louvre motor that causes the louvres to swing back and forth and the blower motor. Both of these turn on shortly after the console is turned on.
processing: x101/010928/ac/doac.pro

05oct01 New air conditioner, louvre motor, ac breaker on/off. (topa1459)

Blank sky was tracked with lbn. 25 Mhz bandpasses were dumped once a second  centered at 1420,1400,1380 Mhz. 5 separate 300 second integrations were performed. The old air conditioner was turned off. After the first plot the breaker for the old air conditioner was also shut off. Figure 1 shows the total power versus time (normalized to the median total power) for each of the 5 observations. A 1st order polynomial removed the za dependence. Color vertical lines are drawn:
• Red. When the air conditioner console was turned on.
• Green. When the compressor came on.
• Dark blue. When the console and compressor were turned off.
• Pink. When the 208 volt breaker for the air conditioner was turned off.
The black line is the power. The light blue line is the turret room (temperature- mean(temp))/1000 (so .001 step is 1 degree F).
All of the plots show a step up in the power close to when the compressor came on (red). All of the plots show a step down when the console/compressor goes off (blue).  The last 3 plots show a spike when the console is turned off (before the drop). Turning off the 208 breaker makes no difference.
• Plots 1,2,3 the louvre motors where turned on/off multiple times during the 300 seconds without making a noticeable change in the power.
• Plots 4,5 had the blower motor come on 60,30 seconds before the compressor with no change in the power.
• Plots 4,5 we delayed the compressor coming on by setting the temperature threshold to 30 C.  In plot 4 it took 70 seconds before the compressor came on. In plot 5 it came on within 30 seconds.
results:
1. The louvre motors and blower do not cause the jumps.
2. Turning the 208 breaker off makes no difference.
3. Turning the compressor/console off causes the power to jump back down.
4. On 05oct01 data the power jumps up when the compressor comes on. On the 28sep01 the power jumps up 10 seconds after the console comes on.


  On 28sep01 the data was contiguous in time. After the compressor was turned off, the unit would not allow the compressor to come back on for 300 seconds (see the plot 28sep01 and the distance between the blue line and the next green line).
    On 05oct01 we did 5 minute position switching with the telescope (to try and cancel ripples in the bandpass for the images). After every 300 seconds of air conditioning on/off, there was a delay of 360 seconds where everything was off before the next cycle began.
    This shows that the thing that makes the power jump up must be some electronics that starts when the console wants to turn the compressor on (e.g.. the temperature is above the turn on threshold). The actual turning on of the compressor is delayed because of the safety timer.
processing: x101/011005doit.pro

31oct01 Timer module for compressor causing the problem. (topa1459)

On 31oct01 blank sky was tracked using lbn. The correlator was setup as:

sbc1 1432Mhz 12.5 Mhz bw

sbc2 1400Mhz 25.0 Mhz bw

sbc3 1413Mhz 12.5 Mhz bw

The Ac console sends 24 volts to the compressor to start. In the compressor is an analog timer module that delays this 24 volts (for 5 minutes after the last turn off so the pressures can equalize) before passing the 24 volts on to the contactor that starts the compressor.
    We inserted double pole, double throw switches before and after the timer module. The datataking sequence was:

start taking data everything off.

@20  secs: turn on the power breaker.

@64  secs: switch on the ac console.

@175secs: switch on the input switch to the timer module.

@275secs: switch on the output switch from the timer module.

@384secs: switch off the output switch from the timer module.

@488secs: switch on the output switch from the timer module.

@532secs: switch off the output switch from the timer module.

@700secs: switch off the input switch to the timer module.

@841secs: turn off ac console.

The figure plots the total power versus time. The 3 colors are the 3 frequency bands. The top plot is pol A while the bottom plot is pol B. The vertical dashed lines bracket where the output switch from the timer module was on. The total power jumps by about .005 Tsys (.005*32=.16Kelvin).

The jump occurred when the output switch from the timer was thrown, even though the 24 volts from the timer module was not being passed to the compressor contactor (it hadn't timed out yet).  The timer module must be creating rfi that is being radiated out the wires that go to the compressor. We completely bypassed the timer module and saw no jump in the total power when the 24 volts was sent from the console to the compressor contactor. The module has been removed and we are checking out how to "fix it". The Ac can't be used until the timer is replaced.
processing: x101/011031doit.pro