AFE meeting, April 17, 2001 John Anderson, Jerry Blazey, Fred Borcherding, Alan Bross, George Ginther, Paul Grannis (video), Stefan Grunendahl, Gaston Guiterrez, Don Lincoln, Thomas Nunnemann, Paul Rubinov, Harry Weerts, John Womersley Paul Rubinov reported that the RHB has been worked on and thought to be ready for tests with the LHB in master/slave mode. (Hopefully on Wed. /Thurs. Apr. 18/19). Paul showed results from temperature control: the resistor was calibrated against the stereo board in both neighboring slots of the cassette. We observe a 3 minute oscillation under temp. control with a peak to peak amplitude of about 2.5 units out of about 360 baseline. The operating temperature was about 9K, so we have met the 100mK spec. Without the control, variations at the 10% level are seen. No problems are observed with the cryo control operation. Thomas Nunnemann showed very nice analyses of both DNL and threshold variation. He used the mean variance of data and his photopeak fits to prior data to determine a correction to the channel response. For both raw and smoothed ADC distributions, this helps suppress the DNL effects. The chisquare of the fit decreased from about 480 to 200 with the first correction; use of an average DNL correction for an entire MCMC reduced it further to 100. He thought that more elaborate algorithms could do better, and that at least for many channels, reliable photopeaks can be found. Thomas analyzed a series of runs on LHB taken in the past three days to determine the threshold variations. Six MCMs were examined, one with more care than the others. His plots are accessible from the AFE web page. For the best understood MCM, two measures characterize the threshold variation -- the rms of the mean threshold location across a SIFT, and the mean of the rms of the thresholds. They agree qualitatively that the rms of the threshold is in the range of 1 -2.7 counts. For the cassette used, we believe that the conversion factors are 2.5 ADC counts == 0.125 pe == 1.0 fC so 20 cts/pe; 2.5 cts/fC; 8 fC/pe. The raw noise rates show a tail that deviates from Gaussian behavior at the high end of the threshold turn-on curve, so we cannot use gaussian approximation to determine rates. Thomas used the actual average noise vs. threshold on masked channels for a full SIFT to translate into single fiber efficiencies. He showed two plots (the 9th page of each of his series of plots 'thresh_rxx ...' that give the single fiber hit efficiency for =8 as a function of occupancy, and the hit efficiency for vs. the MEAN pe yield of the fiber at the VLPC. These plots (see run246) show all channels at >98% efficiency for 0.5% noise occupancy over the full SIFT. All but 3 of the 16 unmasked channels in the MCM show efficiency well above 99%. Other MCMs do not do as well, as there are some channels with weak response, or threshold curves not fully explored. Some of these are expected to be either bad VLPC/mask or bad fitting, but they are not understood in detail. There is reasonable evidence that the noise is primarily contributed by the VLPC and not AFE board noise. Paul Rubinov noted that the dark current pulses are dependent on bias voltage, and tuning needs to be done. One MCM has one SIFT with rather different threshold from the other 3; this can be a problem since if one SIFT is noisy it affects the other SIFTS. Paul noted that if the pclamp is removed, the threshold tend to return to a common value, so that in regular (non-calibration) mode, we can do better uniformity on thresholds. (No pclamp gives smeared photopeaks.) Both the threshold scan and the ADC distribution for one channel was fit for the average number of photoelectrons. The agreement was good to 10%. However because the ADC distributions were very noisy only the one channel out of 448 could be fit. Alan Bross reviewed MC calculations of efficiency. He noted that his assumption of 1/r^2 for doses is based on a point source of radiation; a line source gives 1/r. He has used the measured gains from the VLPC database, the measured attenuation in WGs, and has ignored truly dead channels. He provisionally put in a flat distribution for the threshold variations between +2 and -2 fC. His MC simulates the distribution of traversal length in the fibers. His results show that for thresholds of 7fC and no radiation, eight layer doublet efficiency is 0.997 (0.996 for |eta|>0.5). For 30 fb^-1, this degrades to 0.996 (0.994) at 7fC threshold, and 0.969 (0.939) for 16 fC threshold. See results on the AFE page. Alan and Thomas agreed that their measures of efficiency were in approximate agreement. Overall, the studies are encouraging; the remaining question is why there are some channels that seem ill-behaved or weak in response. Paul noted the differences of trace lengths and pathways on the MCM layout that could explain some of the effects -- channels 1,2,63,64 (and 7??) seem to be low response more often than others. The MCMs are all delivered, so its not clear what could be gained through AFE mods if this proves a problem. Paul reported a highest temperature of 100F (38C) on the board components when operating L and R together with fan (no filter) and baffles in place. John Anderson said components should be OK up to 70C. John reported operating AFE VSVX at 25 and 40 MHz for 10^6 events in DAB6. There were no errors in this study, when events without SASEQ did not trigger are excluded. Discussion turned to whether we have any reason to further hold the Sanmina stuffing. Indeed the weak channels are a worry. We have not tested for a long time for stability, with high and low gain cassettes. Solutions for the DNL problems are not complete. We have not operated with ambient background light simulating high rate operation. John, Fred and Paul do not see any effects that suggest problems that could be fixed by component changes to the AFE. The one test that we strongly want to do before releasing Sanmina is operation of L and R boards together with temperature/cryo control operating. The updated checklist for specifications is posted on the AFE page. The meeting ended with a decision from Harry Weerts: +----------------------------------------------------------+ | So in the end we decided that we will go ahead with the | | stuffing by Thursday afternoon 5:00pm, unless there is a | | clear and compelling reason not to do so. We will | | probably have a short meeting at 5:00pm on thursday. | +----------------------------------------------------------+