Brief summary of results from modified Ka-band receiver tests 15 August 2007 A. Harris, S. Zonak, G. Watts This summary covers testing of the Ka-band receiver with the Zpectrometer in the Green Bank Laboratory from 6-9 August 2007. The principal purpose of the tests was to characterize the receiver performance after its reconfiguration as a symmetrical single channel receiver, but we also made tests to optimize and characterize the system for observing. Reconfiguration included removing the OMTs, which eliminated the second channel from the original configuration, but made a symmetrical front-end circuit possible. Beams on the sky are now linearly polarized, with a pair of 45 degree twists rotating the polarizations so the horns are orthogonally polarized. In addition to changes in the front-end circuit, absorber had been installed on the 50 K radiation shield to damp box mode resonances, and we carefully shielded the ambient load we use for tests from breezes with a thick blanket. In short, we verified that the new configuration should yield improved spectral baselines and allow long integration times. ** Stability The main result is that the modified single-channel receiver is considerably more stable than the previous dual-channel receiver. Measured with the Allan variance technique, the transition from white noise to drifts (the "corner time") is now typically beyond 50 seconds, compared with a few seconds for the dual-channel configuration. All lags share these stability times, with no distinction between high (fine-scale spectral structure) and low lags, as had been the case before receiver modifications. System stability was not always equally good, however, with morning measurements (8 or 9 AM) showing more structure in the time series and Allan corner times of about 20 seconds. Measurements on previous trips to Green Bank showed a similar effect, suggesting that the system is somewhat sensitive to its environment and that the laboratory environment is not as stable in the morning compared with afternoons and nights. In general, stability seems to be limited by transient events, often but not always thousands of seconds apart, instead of more gradual drifts. Since we changed three things at once for the receiver tests (front-end circuit, cryostat absorber, and shielded ambient loads), we do not know how to apportion the improvement in stability precisely. The reduction in nonideal structure, discussed next, is likely to be a major contributor. ** Nonideal structure and common-mode rejection Nonideal structure in the cross-correlation functions are a factor of about five smaller with the symmetric input circuit than with the original dual-channel plumbing. This improvement would account for a similar factor in the improvement in stability times, leaving a factor of a few to be divided between box mode suppression, improved test load temperature stability, and possibly a more favorable source impedance for the first amplifiers. In preliminary tests in June with the receiver warm, we saw even smaller structure; the increase probably stems from reflections added by the 45 degree twists or the relatively small-radius waveguide bends needed for mechanical compatibility with the dual-channel configuration. The decrease in nonideal structure indicates an improvement in common-mode signal rejection that should provide better rejection of sky brightness temperature fluctuations, possibly allowing observations in worse weather than before. ** Frequency scale calibration tone We optimized power levels for the calibration test tone needed to establish the frequency scale in the Zpectrometer. This signal is generated by a doubler in the Ka-band receiver. Measurements show that there is sufficient power for good calibration across the entire Zpectrometer band except from 38.5 to 40 GHz, where the IF power from the receiver is low. We found that an SVD parameter of 0.02, dropping the 2% of lowest-amplitude terms in an SVD decomposition of the calibration signal matrix, is appropriate for all sub-bands. We established that 0.25 second integrations per frequency point is adequate. Combined with reductions in the calibration scheme's internal overheads, the frequency calibration cycle has dropped from about 90 to about 30 minutes. We expect that frequency calibrations will be valid for weeks or months during routine operation. ** Amplitude scale calibration with the noise diode Spectra of the noise diode have amplitudes well above zero power across the Zpectrometer bands (with the usual exception of 38.5-40 GHz). This shows that the noise diode can be a good transfer calibrator for the Zpectrometer spectra across the whole band. ** Band flatness measurements We found a new 5 dB peak-peak power ripple across much of the receiver's IF band, measured with the Rhode & Schwartz 40 GHz spectrum analyzer at the output of the Zpectrometer's downconverter outputs. The ripple's 300 MHz period corresponds to a 0.5 m pathlength (0.7 m in coaxial cable), which could be in the receiver or the cable between the spectrum analyzer and downconverter. The latter is unlikely: the standing wave has high amplitude, is present at some level across all downconverter bands, and seems to be present in uncalibrated Zpectrometer spectra. Apart from the ripple, the noise power flatness appears to be sufficient for the Zpectrometer except near the top of the receiver's band, where the noise power drops by 15 dB (linear drop on a log scale) from 38 to 40 GHz. The noise power rises and falls slowly with frequency across the receiver IF, with no narrow structure or discontinuities. A quick look at the match between Zpectrometer sub-band gains and the noise power slopes indicates reasonable net band flatness except in the highest-frequency sub-band from about 36 to 40 GHz. ** Reduction in Manager software overhead Ray Creager modified the Manager code to reduce the overhead between scans from 13 to 7 seconds. A 7 second dead time implies 90% observing efficiency for 63 second integrations, or 80% for 28 seconds. If the Allan variance corner times are 60 seconds, a 30 second integration time is possible, perhaps with the addition of a small amount of nonideal noise. ** Synchronization with a nodding secondary mode We have an initial approach for synchronizing Zpectrometer data acquisition with the secondary motion during nodding. The best approach seems to be an open-loop scheme based on careful timing: the time between nods is known to high accuracy, and the transition time can be estimated. The Zpectrometer will integrate into a data buffer for some time while the secondary is in one beam, then wait a fixed time for transition, integrate into second buffer, wait, integrate in the first buffer, and so on. From the point of view of the data acquisition microcontroller code this pattern is nearly identical to the WASP chop mode, although open loop, and the micro must return data from the two undifferenced data buffers. One open topic is a careful look at the initial synchronization to ensure that the observing pattern starts in the right beam at the right time. In the future, it would be desirable to incorporate the antenna's knowledge of the secondary's position to provide near-real-time software or hardware interfaces containing beam and blanking binary signals. This would allow closed-loop observing with proper error checking and perhaps higher efficiency. ** Related documents GBT Memo #248: Symmetry in the Ka-band Correlation Receiver's Input Circuit and Spectral Baseline Structure; Harris, Zonak, and Watts (2007) GBT Memo #249: Basic Stability of the Ka-band Correlation Receiver; Harris, Zonak, and Watts (2007) NRAO EDIR #318: Cryostat cavity noise and the impact on spectral baselines; Norrod (2007)