Graeme's Report - May 2001

Results of a recent observation session by Bob Sault indicate the WVRs seem to perform better now. Bob has described a hundredfold improvement. Obvious changes that led to this were the elimination of an oscillation caused by a power supply regulator, and the correction of common mode overload problems on the Digital Acquisition (DAQ) cards. Specific to Unit #2 was an RF level overload on the third amplifier in the chain that almost certainly led to non-linearity. Receiver temperatures measured during the last installation show similarity between both units however they are high relative to the design goals (ie T_sys ~ 200K requiring T_rec ~150K). The table summarizes the measurements.

An initial spec had the required sensitivity as 8 mK. With WVR #2 in the lab, it's first amplifier terminated and a one second time constant the sensitivity equation predicts the smallest observable change to be 17 mK (for the lowest T_rec measured at the telescope). If the receiver was looking at the sky the value drops to 9 mK. Clearly lower T_rec is needed.

The radiometer has an RC time constant of 0.5 secs and with the first amplifier terminated in a room temperature load the sensitivity equation predicts 25 mK. I believe I am measuring 40-50 mK so there is a need to resolve the extra noise. The noise on the signals from the unit is not white and will almost certainly contribute to the poorer resolution. 1/f noise is suspected.

The diagram below compares the fluctuation seen on Channel 2 of the radiometer compared with a battery voltage.

Contrary to previous estimates I think the LSB of the ADC is currently 17 mK due to the input range being set up as -10 V to 10 V. Adjustment will bring better resolution, though I am not sure if we gain a factor of two.

Thermally (in the lab) Unit #2 seems adequate with variation of the order of one part in 16000 at worst. Different channels show different variation and this may be a reflection of the DC amplifiers more than the RF section. On the telescope, with the feed looking at a hot load there is evidence of thermal cycling but whether this can be attributed to the load temperature changing or the unit/input circuitry needs to be determined.

Bob's Report - May 2001

Tests were made of the WVR on the evening of 28 April. The following image shows the results. The red trace shows the measured phase, whereas the blue shows the residual phase after using the WVR data. The WVR correction clearly removes the large excursions that have timescales of approximately 10 minutes. There is drift in the radiometers, which means that the phase does start to wander off outside the radiometers "training interval".

These results are a factor of 50 to 100 better than the results in December last year. They clearly show that the WVR has managed to remove the large excursions.

Conclusions

• We need to understand the non-white noise better, and hopefully eliminate it.
• The system temperatures are higher than had been anticipated. Part of the high system temperature is a result of losses in the waveguide. At the 16 GHz frequency, the waveguide is starting to cut off. Losses in the waveguide will be a function of the vertex room temperature, and so we might see the airconditioning cycling in the WVR output. Possible ways around this are to put the RF amplifiers up near the horn (ugly, and also the worry about temperature effects), place the RF amplifiers in the dewar closer to the horn (a significant change in the dewar design and a worry, again, about the stability of the temperature.), and finally to use some mirrors to elimiate the waveguide and to move the horn close to the WVR units (possible?).
• There has been insufficient thought on how to use (and calibrate) the radiometer in practise. We need to test some calibration schemes on the April data. Also tests of transferring the phase of a calibrator to a program sources have not yet been performed.