Baseline investigations on the 40-52 GHz (Qband) receiver. On January 22, Frank Ghigo took some measurements, with Rick Fisher and Roger Norrod present. The details are in directory CQBAND_012204. The results showed that the baseines in 50MHz mode are not too bad in L1 and R1 channels; poorer in L2 and R2. In 800MHz mode, none of them are satisfactory, but again L2 and R2 are much worse. The baselines show complicated structure with 20-40MHz periods. TBASERDN040126: On January 26, we continued investigation Ruled out the converter rack temperature sensitivity. Since L2 and R2 are much worse than L1 and R1, can do some isolation of the source. Crossed the IF Rack transfer switches, and confirmed the problem is on the FE side of that point. TBASERDN040129: On January 29, we continued. Setup this time so that each FE channel goes through parallel CM->SF->ACS channels. Found that the BL ripples were completely correlated in the parallel channels. Crossed beam switches and found the problem is upstream of that. Did various tests by swapping cables, bypassing beam switches and so forth. Problem definitely seems to be in the dewar. Pulled the receiver off the GBT on January 30, and put it in the Equipment Room. TBASERDN040130: Receiver cold in Equipment Room. Not much time available, but took a few scans. With a LN2 load over window, can see similar results to that seen on the antenna. The fact the ripple amplitude in Ta/Tsys does not scale with Tsys seems to imply that the instability is in the receiver noise rather than gain. Warming up the system. On February 2, Jonah installed isolators between the cold amplifiers and the polarizer in the L2 and R2 channels. Cooling by 5pm. TBASERDN040203: Receiver cold in Equipment Room. The isolators in L2 and R2 made no obvious difference in the performance. Warming up the system. On February 4, Jonah interchanged the mixers in channels L1 and L2. After the change the mixer serial number - channel map is: Chan SN L1 0A26 R1 5F19 L2 5E20 R2 9J02 L3 Not marked R3 0A25 L4 Not marked R4 0A28 These mixers are Spacek Labs model MQQ-11B, and are operated at 15K. TBASERND040205: (Note spelling error in usual ID form.) An hour pirated during weather shutdown for testing with spectrometer. Results were confusing and not enough time to resolve. On February 9, using TP detector we found that the LO multiplier/amplifier for beam 2 is microphonic. Jonah replaced it with a spare which behaves as expected. Hope to test with ACS next maintenance day. TBASERDN040210: Had three hours with the ACS. The baselines in L2 and R2 were essentially of same quality as L1, R1. However, I calculated the standard deviations over "typical" 200MHz portions of the baselines, and these ranged from 0.05% to 0.07% (for two minute scans). The radiometer equation (for BW = 800MHz/2048) reaches this range for 10 second integrations. Hence, the spectral sensitivity still seems to be limited by baseline structure. And, the baseline structure still changes significantly over periods of 10 seconds or so. On February 11 and 12, further troubleshooting found significant (30mV p-p) ripple voltage on the Mixer Bias pins going into the dewar. Capacitors added to the LM317 regulator used to obtain the 3V mixer bias voltage reduced the ripple, but did not eliminate it. Further investigation found that two of the LO Multipliers, DBS Microwave Model 3350X41 SN3 and SN5, connected on band 3 and band 4, were associated with the ripple. The 12V LO Multiplier bias is derived from the system 15V using a 7812 regulator near each multiplier. At the 7812 input terminal, a scope showed ripple on the 15V input wire of ~6V peak, at a frequency of about 2MHz. Removing power to the SN3 and SN5 units eliminated the ripple everywhere checked, including on the mixer bias lines. TBASERDN040213: With ACS again. Not much different than on Feb 10. Began to adjust LNA bias to see if that affected the baselines. Results inconclusive, but there are hints that there could be a relationship. All the LNA's have Cryo3 chips in the first stage. Marian tells me that he has seen unexplained DC bias "jumps" in some transistors from this wafer, although the noise temperature is the best ever seen. February 17: Decided to reinstall the receiver. Left the LNA first stage bias at values that seemed to give a compromise between stability and noise temperature. Checked gain match before beam switches and made necessary adjustments. Installed on the GBT in the afternoon. Pads before Beam Sws (after changes): L1 - 0dB, L2 - 0dB, R1 - 0dB, R2 - 2dB. Pads at IF outputs (not changed): L1 - 2+2dB, L2 - 10+2dB, R1 - 2+3dB, R2 - 2+3dB. TBASERDN040219: Test time with GBT available. Tried unsuccessfully to reproduce the possible link between LNA bias and the baseline ripple. Took measurements with and without load paddle, with GBT at access, as Jonah set several bias settings on R2 first stage. (This was the channel with poorest baselines on Feb 13.) Then did an xon/xoff measurement on sky. Then some DCR scans tracking in TPwCal and then in SwPwrWcal.