1. 1 xCAL monitoring Yu. Guz, IHEP, Protvino I.Machikhiliyan, ITEP, Moscow

2. 2 Histogram analyzing algorithms The prototypes of the following algorithms foreseen are being prepared: ✜ gain monitoring (or calibration) procedure: follows gain variations, updates CondDB when necessary. Works on “summary” histograms, relatively few events per saveset (≤100) sufficient ✜ histogram analyzer: analyzes positions and shapes of signal and pedestal peaks, issues warnings. Works on histograms for individual channels with several K events The calibration farm and some software components are not available yet (coming soon)  still algorithm development on stored LED data. Should be transformed then into “farm-ready” versions. Histogram Analyzer Histogram Analyzer Histogram Database ROOT histograms Histogram producer (ORWELL task) Calibration Farm Histogram adder Saver Histogram presenter Alarms to ECS CondDB updates

3. ECAL LED DATA available  only timing scan data available (according to Twiki)  “default” (bad) time alignment, too few signal events (20÷30) (see the next page)  ok for pedestal studies

4. ECAL LED DATA (cont) Prev2Prev1T0 Next1Next2

5. 5 HCAL LED data available Several short runs with LEDs taken, in order to study the gain monitoring (“calibration”), with “almost” nominal HV (Et_max=15GeV) and LED intensity corresponding to ~3000 ph.el. in each PM. The amplitude varies from ~200 ADC counts in the centre to ~3000 at the periphery. Only LEDs #1 of each module were used. Half of LEDs were fired in each event: the TSB patterns 101010… and 010101… TAE events with 5 BX. Runs with slightly deviating settings in some cells (namely, module 14) were taken, in order to study the gain monitoring procedure.

6. Histogram analyzer - I  Tasks:  determine pedestal and LED signal position and width;  classify channels according to their spectra;  produce summary DQ histograms;  Preliminary algorithm has been developed (using root files produced by Orwell v1r3) and now under polishing

7. Histogram analyzer - II HCAL: pedestal position NB note the difference between Inner and Outer parts HCAL: LED signal amplitude bad channel

8. Histogram analyzer - III ECAL: pedestal positionECAL: pedestal width NB crate 8 FEB 3

9. ONLINE Histogram binning  ONLINE processing: minimize amount of data to transfer;  Instead of one 4096-bin histogram per channel, Orwell produces three, ~100…200 bins each:  xCal/Signal (detailed spectrum, 1 ADC count/bin)  xCal/SignalBis (full ADC range / ~100 bins)  xCal/Pedestal (detailed spectrum, 1 ADC count/bin)  ONLINE processing: histogram parameters have to be taken from DB for each individual channel; ✔ A set of histogram parameters for ECAL/Pedestal, HCAL/Signal and HCAL/Pedestal was produced;

10. ONLINE Histogram binning HCAL: pedestal range HCAL: Signal range ECAL: pedestal range

11. Histogram analyzer - IV  Current classification (both for PM and PIN): 1. Channel OK 2. Empty Channel (no ADC readings or no histogram available) 3. Bad number of peaks in the distribution (bit errors, timing lost, etc) 4. Bad pedestal distribution shape (by Chi^2 of the fit) 5. Bad pedestal position / width 6. Absent LED signal 7. Bad LED distribution shape (by Chi^2 of the fit) 8. Bad LED position / width (i.e. ADC overflow) Control sum 2 3,4,5 6 7,8 %

12. Histogram analyzer - V  χ 2 /n.d.f. of the Gaussian fit of pedestal and signal peaks turned out to be a good parameter to monitor HCAL Pedestal (all C-side cells)HCAL LED (all C-side cells)

13. Histogram analyzer - VI  Signal in ECAL: no reliable χ 2 /n.d.f. distribution, very limited statistics ECAL, χ 2 /n.d.f for pedestal Worse fit than in HCAL, because of the noise tails

14. Histogram analyzer - VII HCAL Pedestal fit Chi^2 / ndf Pedestal spectrum in the bad channel (see prev. slides) is not Gaussian (ATI connector problem, now fixed)

15. Histogram analyzer - VIII ECAL Pedestal fit Chi^2/ndf R/O crate 8, FEB 3 – all pedestal distributions have non-Gaussian shape

16. Histogram analyzer - IX ECAL: it may be useful to monitor the noise sweep one ‘noisy’ power line is clearly seen (plot from ‘C-W noise’ note)

17. 17 Outer cell mod #14 Inner cell mod #14 The procedure was tested on the sequence of all runs (06,07,08,09) (for the moment, as an algorithm in Orwell). The amplitude was averaged over 50 events; PINs: “T0”+”Next1”; PMs: “Prev1”+”T0” (emergency measure, because of not perfect time alignment). “Update” threshold was 4% (for the moment, just types to the screen). PIN mod #14 run 06 run 07 run 08 run 09 Outer cell mod #12 run 06 run 07 run 08 run 09 LED calibration procedure: test on HCAL

18. 18 In the meanwhile, a very unstable PM was found (very high rate effect, to be fixed), and few less unstable ones, to be investigated LED calibration procedure: test on HCAL

19. 19 Check of the PIN correction for the LED intensity variation: 26%  2.7% -- not perfect, but maybe satisfactory for HCAL (for ECAL – to be studied) 26% 2.7% run 06 run 09 run 06 run 09 LED calibration procedure: test on HCAL

20. 20 Pedestals: nonzero in crate 22 (Outer section + PINs)  crosstalk ~1% ; negligible in crate 23 (Inner section) : to be investigated (study the crosstalk pattern) Outer cells crate 22 PINs crate 22 Inner cells crate 23 LED calibration procedure: test on HCAL

21. 21 ● Take more data for ECAL and HCAL: ● long runs (several hours), to study the stability, and measure the stabilization time after HV ON in various conditions ● vary PM HV and LED flash intensity, in order to take regulation curves ● vary LED firing sequence, to study the crosstalk pattern; choose optimal sequence ● etc… ● Prepare “farm-ready” versions of the procedures. Plans for the nearest future