E895 Meeting Minutes 25 March, 1997 Attending: Gulshan Rai, Nathan Stone, Dieter Best, Heng Liu, Mike Lisa, Jim Draper, Mark Gilkes ------------------------------------------------------------------------ News & Status Report: NS - Tasgen is installed in CVS. - We have two listserv lists at BNL, half of the collaboration has subscribed to them. For those people who haven't subscribed to it, please do so now. - There are some nice utilities which can convert the archived messages to html, so that people can see them on the browser. I'll try to get this working for us. JD - argue that a column of positive charges along the path of gold beams due to ionization could explain the deviation that beam height is off by 2cm from normal events. DB - working on the tracker. Marvin suggested to do a second pass of the tracker to pick up the unidentified hits, this improved the track removal efficiency by about 20%. - looking into the helicity problem because we have seen a clear anti-proton band even at lower energies with a few hundred events. - track fit routine fails a lot, we have to find out why. Spore finder seems to have a problem too. I put track merger back into CVS. - doing simple Geant simulation of a few particles, the tracker should come out with the same as what we have put in. ML - talk about the hitfinder. TRFF=1 seems to find more hits but less tracks, TRFF=2 is not doing better. New hitfinder does find more tracks and track length is also longer, the hit removal efficiency is about 70% (with double tracking) compared with 30% for old hitfinder (with double tracking). - The old hitfinder has a bug that the distortion correction flag is not passed correctly. General Comments on Mike's new 2D hitfinder, (from Nathan's log book) - hitfinder_2d has several running "modes" (explained below) - which hitfinder and mode is chosen via the (obsolete) switch HIT_SW.LPRT - LPRT > -99 chooses old (standard) 1D hitfinder - LPRT <= -99 chooses Mike's 2D hitfinder - if (LPRT<=-99) then (hitfinder_2d mode = 99 - LPRT) AND... - one must also set a switch defining the maximum number of hits per cluster (HIT_SW.UNDERSHOOT, explained below) - hitfinder_2d takes roughly the same CPU time as the 1D hitfinder running with TRFF=0/1 - modes fall into two groups: first-pass and second-pass - 1st pass modes calculate fit parameters for identified peaks "from scratch", and feed those to the tracker - 2nd pass modes are called after 1 full pass of HIT/TRACK-finding, and use dip and crossing angles of the tracks to modify the initial guesses made at peak shape fittings. The new hit parameters are then given to the tracker again for a second tracking. - 5 fit parameters characterizing peaks are: A : amplitude x0 : x coordinate y0 : y (t) coordinate alpha : crossing angle lambda : dip angle Explanation of modes: mode 0: (1st pass, LPRT=-99) ---------------------------- 1) eliminates electron clusters (vertical spirals extending all the way down to the pads). 2) cluster-finder (a la STAR) 3) mountain-finder (a la STAR) - even handles mountains with "flat tops" (saturated ADC) 4) peak fitter - "local" fitting - this calculates the 5 fit parameters for each peak considering only its 8 nearest neighbors in x/t-space mode 1: (1st pass, LPRT=-100) ----------------------------- mode 0 plus... - peak fitter does "global" fitting (multi-peak simultaneous fit) - fit determining 5 parameters for each peak is done accounting also for the "volume" under nearby peaks BUT - it only allows the AMPLITUDE to vary in this respect * mode 1 is about as fast as mode 0, and gives better dE/dx mode 5: (2nd pass, LPRT=-104) ----------------------------- mode 0 plus... - peak fitter does "global" fitting, allowing ALL 5 fit parameters to vary (not just amplitude) - it uses (A, x0, y0) from the 1st pass hitfinder_2d but uses (alpha, lambda) from the 1st pass tracker as initial guesses mode 6: (2nd pass, LPRT=-105) ----------------------------- mode 5 plus... - keep (alpha, lambda) fixed during the second fitting (don't let the 2nd hitfinder change the values returned from the 1st tracker) mode 10: (1st pass, LPRT=-109) ------------------------------ - global fitting of all identified peaks, varying all 5 parameters * quite slow FOR ALL MODES: (setting HIT_SW.UNDERSHOOT...) --------------------------------------------- - The peak-finder will look for up to [HIT_SW.UNDERSHOOT] hits per cluster. If more than this number of hits are found in a cluster, the peak-finder will temporarily raise the threshold value (for THAT cluster) until the number of hits found no longer exceeds the value defined by UNDERSHOOT. - one _typical_ example quoted for this value was 4 or 5 - to "turn off" this feature, simply set UNDERSHOOT to 1024 * This "searching" for the best threshold (varying of number of peaks) does require some additional time, but it saves time in the end, since peak-fitting takes longer than peak-finding. HL - showed the study of bad pads by looking at the normal interaction events with pad_monitor. The number of bad pads is about 170 which is much smaller than 400 claimed by Paul. ML - The pulser runs show different "bad" behavior than the real (physics) data. From the pulser events, many of the pads that I would call bad (like, ADC value is negative, always overflow or something) are the first few pads of a stick. This seems like a DSP problem. My guess is that the DSP in the receiver cards sometimes screwed up in pulser runs (format 1) differently than in physics runs (format 10). HL - The drift velocity calculation for 2 and 4 GeV data is underway. I used the top edge method with the right parameter of global t0 offset and Yoff as input, then by knowing the geometric location of the top edge we can deduce the drift velocity. GR - The codes that need to be checked into CVS must go through librarian Nathan (in his absense Dieter is the deputy). - Mike's new hitfinder code is in to CVS. - propose the next collaboration meeting is the first week after July 4th at Ohio State. please let me know if there is any conflict for anybody. What we have decided: o Dieter will focus on debugging the tracker. o Nathan will do systematic evaluation of new hitfinder and double tracking. o Heng will get the bad pad table from noraml event and also get PID routine going. o Mike will decouple dE/dx from tracker. o Mark and Arun will tune the fast simulator and have some efficiency study from it. Prepared by Nathan Stone and Heng Liu