Gain versus temperature.

26aug00 ,2000

Setup:

      A temperature sensor is mounted above the IF/LO racks on the rotary floor (in the dome). It is sampled every 10 seconds.   The sband transmitter transmits 1 Megawatt cw. It is about 50% efficient so 1 Megawatt of power is dissipated in the transmitter room (above the rotary floor). Cooling water is circulated through the klystrons and and a separate cooling line is used to cool   the wave guide and transmit horn in the rotary floor room.

    On 26aug00 an sband transmit experiment ended at 3:00 hours. The temperature in the rotary floor room had risen from 74 degF to 82 degF. The air conditioning in the room got the temperature back down to 74 degF at 6:00 hours. The outside air temperature was about 73 degF.
    At 4:00 hours a calibration run at cband was started on J0318+164 doing position switching. 1 minute on,offs were done with a 1 minute move between the  positions. The cal was fired for 10 seconds on/off at the end of each off position. The scale factor to convert from correlator counts to Kelvins was computed every 3 minutes from the cal data (cal[K])/(calOn-calOff[corCnt]). Since the attenuators were not changed during the run, this value should be a constant unless the gain or the size of the cal changed.  The data showed a large drift in this value over the entire run. The system temperature measurements using the cal showed no long term drift which implies that the drift was in the gain and not in the calValue itself (the cal diodes are also temperature stabilized.

Results:

The plot  shows the % change in the cal scale factor versus time (marked with *). The  (temperature in the room - 70F) is also plotted. Over the 12 degF temperature change, the gain changed by 15%. At 6:00 you can start to see the air conditioner start to cycle on/off and the gain is also following this cycle. It is making about a 1% change in gain per degree change in temperature.
    But there are some things to be careful about:
  • We are only measuring the air temperature in the rotary floor room. There is also gain downstairs.
  • The temperatures of the devices that are affecting the gain can change at a different rate than the air temperature (they probably change more slowly).
  • A small part of the change can come from the za change during the cal on/off.
    •     When calibrating spectra, we assume that the cal scale factor is constant for the integration. Assuming the cal is not changing, then the error in the calibration is the change in the cal scale factor over the integration (or maybe 1/2 that). In our case for 1 minute on /offs it ranges from 0 to 2%. Most position switching is  5 minutes on/off. This would last for 4 of the 1 minute measurements. Even during a normal observation (without an sband xmit experiment) you could get a 2% change over a 5 minute on/off.

    Conclusions:

    1. The current air conditioning needs about 3 hours to get the  rotary floor room back to the normal temperature after a nighttime sband xmit experiment (when the air temperature is 73 degF).
    2. The upstairs gain looks like it changes 1% for 1 degF change in temperature.
    3. The air conditioner cycle of 2 degF over 15 minutes may add a 1% error to the calibration.
    processing: x101/cb/aug00/26aug00.pro and then tempstab.pro
    home_~phil