Thu, 21 Feb 2002 R. Millan-Gabet PICNIC Noise Due to Eltrol Electronics ====================================== It seems that the noise problem has largely been fixed. Details below. --- Feb 21, 02 ---------- Mike Brewer atacked the problem of killing the oscillation in the Eltrol electronics, believed to be the source of noise "picked up" by the PICNIC camera, with great success. === see Mike's notes on separate memo === RMG simultaneously worked on some improvements to the wiring inside the dewar: (a) ground sense connection to ADC is now a direct wire between the pre-amp and ADC boards, (b) temporary direct connection to PICNIC output (for debugging) has been removed, (c) PICNIC output wires are now tightly twisted-pair between the detector and the pre-amp. Before warming the dewar, it was clearly established that MB could debug the Eltrol electronics using other noise diagnostics which were previously determined to clearly correlate with what was seen on the PICNIC signal: noise radiated picked up by AM radio, by antenna set up at input of ADC, by PICNIC signal when connected to shorted pre-amp. This noise could be as bad as 100s du in amplitude and dependent on SD location, as established earlier. After the Eltrol cards mods, there is only a smaller oscillation remaining, +-1 volt (instead of 10 V), 500KHz-1MHz depending on SD location. This oscillation however has no discernible effect on noise measured in above tests. Now that the PICNIC dewar is cooled again and back on-line we get the following: I call the "noise" the rms of PICNIC scans in 1 read 1 loop mode. I call "base noise" what we get when the PICNIC is looking at a cold load with no SD activity. This value is ~ 7 du. This is the same we get under same conditions in: (a) archival NICMOS data from the old classical experiment and FLUOR, and (b) PICNIC readout using the Pentium (data from Sept 01) => good. (but maybe could be as low as 5 du). When the SDs are moving we now get (V = 0.2 - 1.0 cm/s): SD2 has no effect on noise SD1 results in rms ~ 11 du when moving away from home SD1 results in rms ~ 8 du when moving toward home SD1+SD2 same as SD1 alone Note: Other measures of noise, after fixes: ADC input shorted: 1 du pre-amp shorted: 3 du in both cases with no effect from moving SDs. Conclusion ---------- * We seem to be OK to observe, though we'll continue testing to make sure nothing intermittent shows up. * SD1 still introduces some excess noise. This is likely due to the fact that it is the one with no shielding in the long cable going from the VME rack to the tank. Experiments using al foil underway now. The fact that SD1 introduces more noise going away from home is unexplained. To Do: ----- * MB has not given up on killing the remaining Eltrol oscillation. * Note added in proof: The good rms ~ 7 du mentioned above can become 11du by messing around with dewar, card cage contacts with optical table etc, to which SD1 still adds a few additional du of noise as described above. Therefore it seems like we still have room for improvement at that level, if we only knew what is going on with these earth/ground connections. * RMG might re-connect PICNIC to Pentium to confirm above noise figure and allow readout with whole VME chassis off, which will test acusations made to noisy VME environment. --- Update ------ * Pentium readout of PICNIC: VME rack OFF: rms ~ 9 du VME rack ON: rms ~ 11 du => small effect suggestion: add side & back shielding of VME cage * MB has killed the Eltrol oscillations COMPLETELY SD1 cable to tank has been wrapped in al foil Back to VxWorks readout of PICNIC base noise: rms ~ 7 du note: goes up to 12 du when lights in tank ON! something is wrong with the supply to those lights! Drive SDs in vacuum V = 1 cm/s SD1 slewing away from home: rms ~ 10 du SD1 & SD2 slewing away from home: rms ~ 10 du SD2 slewing away from home: rms ~ 7 du SD2 slewing toward home: rms ~ 7 du SD2 & SD1 slewing toward home: rms ~ 7 du => latest improvements have not made a difference, although it sure is nice that Eltrol circuit does not oscillate at all now ... suggestion: make real shield for SD1 cable to tank. apparently we can buy a really nice one from Anorad, like the one we have for SD2. * Dependence on SD speed: (test with SD1 slewing away from home, the worst case) V = 0.05 cm/s rms ~ 6 du V = 0.1 cm/s rms ~ 7 du V = 0.2 cm/s rms ~ 7.5 du V = 0.3 cm/s rms ~ 7.5 du V = 0.4 cm/s rms ~ 8 du V = 0.5 cm/s rms ~ 9 du => there is a small but clear effect, but at normal tracking speeds we seem to be OK.