##### LTAFILE : node51:/data2/gpuuser/gabtest_1300lo_15aug2013_usb.lta.3 ##### LTAFILE : node51:/data2/gpuuser/gabtest_1440lo_15aug2013_lsb.lta ##### Settings: Observe date 15 AUG 2013 OBJECT 1330+251 LO1 1350 MHz RF 1280 MHz Band = flipped (GUI), meaning LO1 > RF (or LSB) S04 is taken for maintenance No. of Channels 2048 Channel-width 97.66 kHz ##### Note from Nilesh: 1. At 10:00 IST we tried to start GPU. but antennas were not fringing. with GAB LO at 1350MHz, Also there was various instances, 3-4 times when GPU correlator was not responding to online commands ( viz strtndas, stpndas, hltndas etc.). So tried reboot to node51-54 machines and restarted GPU data acquisition. 2. till then GAB LO set to 1440 MHz and found the good correlation between the antennas. files 1-4 shows primary results from tax. ltafile : gabtest_1440lo_15aug2013_lsb.lta cross-amp-with-chan-1440lo-baseC10.jpeg cross-amp-with-time-1440lo-baseC10.jpeg cross-phase-with-chan-1440lo-baseC10.jpeg cross-phase-with-time-1440lo-baseC10.jpeg 3. after 12:30 switched GAB LO to 1300 MHz. and restarted GPU data acquisition. But, Antennas on same ROACH Boards were showing fringe correlations. so once again tried reboot to node51-54 machines which did not help. ltafile : gabtest_1300lo_15aug2013_usb.lta and gabtest_1300lo_15aug2013_usb.lta.1 cross-amp-with-chan-1440lo-baseC10.jpeg cross-amp-with-time-1300lo-baseC10.jpeg cross-amp-with-time-1300lo-baseE06.jpeg cross-amp-with-time-1440lo-baseC10.jpeg 4. After that at 14:40 IST gave power reset to Both the ROACH boards. and restarted GPU acquisition. Still the problem above persists. ltafile : gabtest_1300lo_15aug2013_usb.lta.2 cross-amp-with-time-1300lo-baseC10_2.jpeg cross-phase-with-time-1300lo-baseC10_2.jpeg 5. Stopped GPU chain at 03:47 IST. and restarted it again. details about will be send later on. ##### Note from Yash: Thanks for the quick update on the tests done this morning. As per my understanding, the RF was the 1390 band of the L-band system and we agreed that both options of LO below and above RF should be tried, for which we agreed for 1440 and 1300 MHz settings. I realise in hindsight that 1440 would have been at the very edge of the stopband of the filter (in principle the Lband subbands are +/- 60 MHz around the nominal centre frequencies of the subband) -- this is the reason that the cross-band shapes for the 1440 LO settings start from almost the DC freq. It will be good to see the cross (and self) bandshapes for the case of 1300 MHz LO setting also. Look forward to the more detailed description of the later tests, as you mention in the email. It will also be useful to get some corresponding GSB plots. Some comments and points that require follow-up that come to my mind are as follows : 1. The cross-ampl look reasonably decent : can you tell what was the calibrator that was being used in the GTAC observation? 2. This illustrates that it is, in principle, possible to run the GPU corr in parallel with the GSB (even during regular GTAC observations) and we should work towards exploiting this facility more regularly. If nothing else, calibrator scans from the GTAC observations (from the GPU corr data) can be pulled out for analysis from which we can increase the testing time of the GPU corr and get some idea of the long-term behaviour of the GPU corr. 3. For quick, general use of the GPU corr for such things, the following need to be streamlined : (i) a table of values of recommended LO settings for GAB that we can use for any of the normal bands being used for the GTAC observation. (ii) a slightly more independent set-up between GSB and GPU (Jitendra is cc'ed here for this aspect) : I found that, at present, situation is such that if you don't start the GPU corr at the time when the GSB is started for the GTAC obs, then it is not easy to add the GPU later on (that's why today's tests had to be done using standby online) -- can this issue be addressed? (iii) ability to extract corresponding data (e.g. matching frequency) from GSB and GPU files for a quick comparison. 4. It is not clear why the GPU corr goes into the mode of fringes only between antennas on same Roach board : it does not necessarily go away with reset of the boards (may be a hard power cylce is needed?) or is it some other control level problem?