Jan Conrad, Uppsala University.
Arvid Pohl,  Kalmar University

Last Update:

Oct 30: added ADC selection results.

Update History:

Oct 22: List of DEAD OMS ADDED
 

Introduction

This website intends to describe the selection of bad oms and bad runs for data analysis of 1999.
The algorithm is described and all necessary ingredients are provided. A list of bad OMs and bad  runs is given.

A template script providing the appropiate recoos flags (for noise rate based selection)  is given

NOTE : THE SELECTION IS FOR B 10 ONLY

Comments and suggestions are of course wellcome.
 

dBTreeBuilder

The data diagnostics for (1998)/1999 utilizes the packagde dBTreeBuilder which has been developed in Zeuthen by Mathieu Ribordy.
It builds a root tree from either monitoring  files or reader hbook files.
Get the package from here
 

OBS: The Trees are available ON REQUEST  for the moment (not enough disk space) :

The old tree:   99Tree300.root
A more complete tree :  Tree2484.root
The most complete tree:  dbtree99moni.root
 

Selection Criteria

At this point the selection is based on Noise Rates only:

The selection is performed in two steps:

a) a Gaussian fit is made to the global noise rate distribution (String 1-4, 5 - 10
      seperately)

b)  a loop is performed over all RUNS and all OMS:

      - if in a given run an OM has a 5 sigma deviation from its own mean (over the year) or from  the
        global mean (String 1-4 and 5- 10 seperately) it is flagged.

      - if in a given run an OM has a 10 sigma deviation from its own mean (over the year) or from
        the global mean it is flagged.

      - an OM is considered BAD if it has a 5 sigma deviation in 50 % of all runs. A BAD OM should be excluded.

     - a RUN is considered BAD if it has more than 50 % OMs with more than 5 sigma deviation.
         or if it has more than 20 %  OMs with 10 sigma deviation.

     - a RUN is considered SHORT if it is shorter than 2000 s.

     - a RUN which is either BAD or SHORT is considered as NASTY (and should be excluded)

      - an OM which has a 10 sigma deviation in a given RUN is considered as EVEN WORSE (for that RUN) and
        should be excluded (for that RUN).

In addition the ratio TDC-Rate / Noise Rate is calculated.  An OM is flagged if its ratio of this quantity  is smaller than 1. (It is not clear at the moment
how this variable was used in the Zeuthen analysis, so we do not use it for the time being).

The ROOT script, that does it:  noise_rates.c

The script produces also a map of oms with 10 sigma deviation and a map of Oms with TDC/Noise >=1.

 List of BAD OMs

Following OMs should be excluded from the reconstruction, independent from run number:
 

OM #	Status
3, 28, 32, 36*,39, 40, 41, 42*, 44, 47, 50, 57, 61*, 62,  78, 81-86, 94, 96, 117, 143, 169, 172, 186, 189+, 190+, 195, 197, 199, 215, 227, 229, 231, 232, 235, 243, 244, 265, 267	BAD

 * these are NOISY with respect to the average String 1-4 OM.
 + these have a sudden jump around run 350.
 

 Following  OMs are DEAD (< 10 Hz in 90 % of all runs). They should be included in the
 Monte Carlo:
 

OM #	Status
28 32 39 40 44 47 50 57 62 78 81 82 83 84 85 86 94 96 117 143 186 195 197 227 231 232 235 265 267	DEAD

 

 ...... and a map (20++ ps pages) : OM with 5 sig deviation
 

 ...... and if your twisted mind wants the whole map as a table (90 ps pages):  om.ps
 
 

Following OMs should be excluded from reconstruction for certain runs:

We decide to exclude OMs with very high noise rates for certain runs:  for
 certain runs which have a deviation greater  than 10 sigma and are not allready excluded by the bad run or
 bad OM list:
 

OM # (RUN)

Status

254 (11),  264 (11),    16(97),   263(112), 234(160), 249(160), 258(160), 234(161) 249(161), 258(161), 234(162), 249(162), 258(162), 302(215), 302(216), 263(217) 302(218), 302(219), 302(220), 302(221), 302(222), 302(223), 302(224), 263(308), 198(450), 263(450), 263(477), 302(661), 302(664), 302(667), 302(668), 302(669), 302(670), 277(674), 277(675), 289(675), 277(676), 289(676), 302(678), 302(682), 302(683)

EVEN WORSE  ..... but not allways.

 
 

List of BAD RUNs

Following Runs should be considered to be BAD runs:
 

RUN  #	Status
7, 16, 17,  21 - 30,  39-47, 71, 73 - 91, 106, 149 -155   346, 412-426, 549, 686-700	BAD

The green runs are within the "normal" data taking  period , which is defined by:
The black runs are excluded because of the 5 sigma criterion.
The blue runs are excluded because of the 10 sigma criterion.

START:  the first entry in  the  99 run log book (hardcopy) by the WOs.
END    :  the first entry of summer activities in the  99 run log book (hardcopy).

In addition  we add all Runs which are short (t < 2000 s), so we arrive at a  "complete" run selection:
 

RUN  #	Status
1-6, 7, 16, 17,  21 - 30,  33,  35, 39-47, 67,  69, 71 ,73-91, 102,106, 108, 109, 111, 112 ,116, 149-155, 156-157,  243, 251, 276, 291, 326, 345, 346, 379, 389,  412-426, 428, 500 -511,  518, 531, 549, 685 , 686-700	BAD + SHORT (= NASTY),

 Runs in  red  runs are excluded solely based on run length.
 

 Cleaning Script

 A  template script with the appropiate recoos flags:  template.pl

 The script has not been tested yet, so use with caution !!!!
 

 ADC results (Arvid Pohl, Kalmar)

 The following OM/run-selection comes from the ADC-calibration, where the 1pe-peak has been determined.

This is the list:
 

 Bad OMs

 OM  264: Highly unstable, keeps jumping between 1pe = 30mV and 1pe = 90mV.
 
 

Problematic OMs

OM 276:  Loses 1pe-peak in runs 159-162 and 435-528.

OM 302:  Disturbed TOT and large-scale ADC from run 667 and forward. The disturbances have large ADC and TOT and will probably not disappear in hit  cleaning.
 
 

Bad runs

 Runs 494-499, 517, 519-521: Dramatic changes in the ADC spectra, due work on the adc-crates.
 

Bad OM/run-combinations:

 OMs 34 and 66 in runs 107 and 159: No ADC.

 OM 238 in run 72: No ADC

 OM 171 in runs 542-544: Loses 1pe-peak
 

 The flariest runs

 In many runs the detector suffers from disturbances clearly seen in the ADC spectra, possibly wind related.

 Most of the flare related hits are discarded by crosstalk cleaning.

 However, the number of cleaned hits is also seen to change during flares, so to some extent the analysis is affected by flares.

 The flare phenomenon is currently (October-November 2002) being investigated further.

 Presented here is a list of runs with many bad ADC fits, mostly due to flares:

 Runs with at least 100 OMs with disturbed ADC spectra:

 217-219, 283, 284, 307-309, 317, 318, 320, 321, 348-350, 375, 376, 443, 449, 450, 464, 472-477, 480, 481, 522, 590-592, 594, 595, 637, 638