Minutes of the 15 June 2006 T474 Phone conference ------------------------------------------------- - Bino reported on the progress that was made in the BPM analysis and the DDC optimisation. The key was finding that the algorithm in the lastest version of the code had a bug. The IQ plots had a lot of phase jitter. The reason for this was incorrect treatment of the filtering algoritm in timedomain, where the filter reference time (tfilter) was put to the starttime (startFit, picked up from calibration file) and the end time, tstop, was put to that value +dt, where dt is calculated from the filter bandwidth. This technically speaking reduces the bandwidth of the filter as well by a factor of 2. This issue is extensively supported by plots in the electronic logbooks at : https://www.hep.ucl.ac.uk/elog/Bino's+logBook/5 https://www.hep.ucl.ac.uk/elog/Bino's+logBook/4 https://www.hep.ucl.ac.uk/elog/T-474/16 A new version of the code (NanoModules version yury-13jun06a) was sent around by Yury, with the following changes : * the value t0Ref picked up from the calibration files now refers to the filter time (tfilter) * tstart = tfilter - dt, tstop = tfilter + dt This code produced good IQ plots again so optimisation could continue. - For external clock data, the optimal bandwidth and t0 Bino and Yury agree upon to be t0 = 68.0 filtBW = 0.20 For t0Ref Yury finds an optimal value of 86, which produces after some fudging resolutions on BPMs 3-5 of 500nm in y and about 800 nm in x, Bino finds 89 as optimal value with slightly worse resolutions in y at 800 nm and x ~ 1 um. Having the tfilter value a little further (89) filters the waveform only at it's end, thus conveniently bypassing any saturation effects but having a slightly worse resolution. Having tfilter at 86 takes more of the beginning of the waveform into account, thus having better resolution, but more sensitive to saturation effects. So this is also the reason why Yury had to apply some 'fudging' to account for saturation effects that might slip into the calibration. So basically this boils down to the fact that we need to find a way to treat saturation better in the DDC algorithm. - The idea there is to have a look to the downconverted waveform, try to fit the gamma's and use those then. However problem here is that for some strange reason the phase has a slope across the pulse. This slope can be changed by tweaking the frequency assuming that this dependence is linear with time. This is however not the case, so the phase slope changes with amplitude (eg. in calibration runs) The reason for this is either maybe influence from a monopole component or some other pathology (possibly hardware related). Note that there appear also 2 peaks for e.g. 417-419 in the FFT spectra, which give possible indications for hardware problems... So the play here is to use a macro called "plotDDC.C" in NanoMacros (version yury-13jun06a) and tweak the frequencies manually in order to make the phase plots flat across the pulse, next step is to fit the gamma's in the downconverted waveforms and use those in the DDC algoritm. ---> In future we will tabulate this variation of the phase across the pulse and correct for it by means of a lookup table.... - Yury brought up the issue of runs 960, which have external clocking and where the terminators were removed. This effectively changes the frequency and the width of the cavity, so one should take care of that when optimizing the data. This will be taken care of. - Chit produced a new set of waveform info files for the entire runrange having the correct settings depending on internal or external clocking. Yury will send these around as a tarball. Bino produced new values for the fixed t0 for external clock, he will send these around as well. - Yury will commit a new version of the t0 finding routine for internal clocking, which essentially removes the code which 'goes back a couple of samples' to do the linear fit of the rising edge of the trigger pulse. This should hopefully improve t0 determination for internal clock runs and resolution. - Bino brought up an issue about calibrations where the IQ phase is around pi/2. Yury suspects this is probably due to saturation and not so much the actuall IQphase value. A quick check by slightly changing t0 (which changes the IQ phase) should bring clarity on this issue. - Bino was wondering why the phase difference between q9Phase and q10Phase is more or less Pi. This is purely by coincidence Yury informed :). It brought up however the issue of looking to the Phase resolution. Bino will put up a plot in the eLog so we have an idea about it. - The conversion of the calibration files to have the real frequencies and widths of the cavities (so to make them uniform between external and internal clocking) has been postponed as everyone seems to be happy with the current version. Might implement it once the entire optimisatio/understanding of the data is finished. - Plans for future : o split calibration / reprocessing according to internal & external clocking... o Implement a correction table for the phase change with time in the DDC algoritm... o Check whether the ADC values for the corrector magnets are in the root files and write some code to automatically have this do the calibrations, possibly having calibrations 'on the fly'... - A mailing list was setup called 'esa-t474@hep.ucl.ac.uk' for mass mail concerning T474 analysis. To subscribe, people should send an email to "esa-t474-request@hep.ucl.ac.uk" with an empty subject and in the message body "subscribe address=" - Time for phone conference next week will be decided via email exchange as Yury is away.