The EVP Subsystem of COMPASS

EVP Task Descriptions

• EVPRNN - Event Processing with Neural Nets
  Most recent version - 14.
  This task performs the processing of COMPTEL event data, producing event data files (EVP). The event location method is based on a neural net algorithm (rather than a table look-up, as in the very earliest versions of this softare known as EVPROC).
  Software Users Manual | COMPASS Software Profile | Version History

• EVPSEL - EVP Data Selection
  Most recent version - 3.
  This task applies various user-specified selections and filters to generate a new smaller EVP datafile. In addition, EVPSEL has the capability to merge the individual members of an EVP list into a single EVP datafile.
  Software Users Manual | COMPASS Software Profile | Version History

• DDMEVP - EVP Data Display and Manipulation
  Most recent version - 6.
  This task can be used to generate various 1-d and 2-d plots of EVP data.
  Software Users Manual | COMPASS Software Profile | Version History

EVP Dataset Descriptions

This section describes the various datasets that are used as input or output within the tasks of the EVP subsystem.

• NL3 - Neural Net Location Tables for 3-Dimensions
  Standard Input to : EVPRNN
  The location of the event interaction sites in the D1 and D2 modules is determined from the relative pulseheights amongst the various PMTs within the given (D1 or D2) module. Within EVPRNN this process is achieved using a neural net lookup. In this case, the neural net maps were generated by a learning processed based on pre-flight "module mapping" data. The NL3 look-up tables were generated within the MMG (Module Map Generation) subsystem of COMPASS.
  • Standard NL3 Datasets

• MHT - Module Homogeneity Tables
  Standard Input to : EVPRNN
  • Standard MHT Datasets

• EVP - Processed Event Messages
  Standard Output from : EVPRNN, EVPSEL
  Standard Input to : EVPSEL, DDMEVP, others
  • EVP Data Format
  • EVP Dataset Representation History
  • EVP Quality Flag Definitions
  • EVP Rejection Flag Definitions
  • EVP ToF Corrections
  • EVP PSD Corrections
  • EVP Selection Set Functions

• EHR - Event History Record
  Optional Output from : EVPRNN
  Optional Input to : EVPRNN

  This data structure was defined as an option for intermediate processing of EVP events. The idea was that one could save data in partially processed form and return to them for final processing at some later time. This concept was introduced at a time when processing time per event was a principle concern. In practice, EHR data has virtually never been generated.

• ESR - Event Processing Statistics
  

• TCF -
  

• FPM -
  

• TCC -
  

Relevant Documentation

Here are some documents related to various aspects of COMPTEL event processing:

COM-ST-UNH-EVP-268 : issue26 : October 9, 1997
Summary of EVP Processing Gaps and Problems, by R. Freuder

COM-RP-ROL-DRG-37 : Issue 1 : November 26, 1992
ToF Selections, by B. Boer and R. van Dijk

COM-RP-ROL-DRG-44 : Issue 1 : April 14, 1993
The PSD Correction as a Function of PSD Selection and D1E, by R. van Dijk

COM-RP-ROL-DRG-45 : Issue 1 : March 30, 1992
ToF Selections, by B. Boer

COM-RP-ROL-DRG-53 : Issue 1 : June 25, 1993
ToF as a Function of D1E and D2E, by B. Boer

COM-RP-MPE-DRG-165 : Issue 1 : No Date
Studying the ToF Spectrum as a Function of Energy, Part I: the Forward Peak Region, by G. Weidenspointner and M. Varendorff

COM-RP-MPE-DRG-171 : Issue 1 : November 7, 1996 | addendum
Studying the ToF Spectrum as a Function of Energy, Part I: the Forward Peak Region, by G. Weidenspointner and M. Varendorff

COM-RP-MPE-DRG-173 : Issue 1 : June 5, 1997
Processing EVP-132 Datasets, by G. Weidenspointner and T. van Sant

COM-RP-MPE-DRG-174 : Issue 1 : October 9, 1997
Correcting the Correction of the Correction... ToF & PSD, by G. Weidenspointner, M. Varendorff and W. Collmar

EVP Processing

EVP Processing History

The COMPTEL collaboration has decided to reprocess all data which can benefit from reprocessing to handle failed PMTs. This includes data from TJD 8717 up through TJD 8956. Starting with TJD 8957 (1-Dec-92, the start of observation period 203.0), all routine processing has been peformed to process those modules with failed PMTs. This means that data between TJD 8717 and TJD 8958 can benefit from reprocessing.

In general, we can define five time periods of processing:

1. launch - TJD 8717 (5-Apr-92) : Obs periods 1-24.0
   These data were processed with incorrect EHORA information for each event. There are no plans to systematically process these data to correct this problem. These EVP datasets have QF = 120.

2. TJD 8717 (5-Apr-92) - TJD 8818 (15-Jul-92) : Obs periods 24.0-33
   These data were processed with incorrect EHORA information for each event. The failure of one or more PMTs for these data suggest that these data will benefit from reprocessing. Those datasets identified for reprocessing are given a QF = 119. Once properly processed, the new EVP dataset has a QF = 130.

3. TJD 8819 (15-Jul-92) - TJD 8956 (30-Nov-92) : Obs Periods 34-202
   These data were processed with the correct EHORA information for each event. PMT failures, however, suggest that these data can benefit from reprocessing to handle the failed PMT modules. Those datasets identified for reprocessing are given a QF = 129. Once properly processed, the new EVP dataset has a QF = 130.

4. TJD 8957 (1-Dec-92) - appx TJD 9247 (17-Sep-93) : Obs Periods 203.0 - 302.3.
   These data have been properly processed, with regards to both the EHORA information and the handling of dead PMTs.

5. appx TJD 9247 (17-Sep-93) - present : Obs period 302.3 -
   All data processed from 4-Oct-93 onwards incorporates an improved ToF correction. This includes all data from appx TJD 9247, but it also will include any data from earlier in the mission which has been reprocessed after 4-Oct-93.

History of Eratic and/or Failed PMTs

Modules with erratic PMTs cannot be properly handled within the COMPASS processing environment. Once a PMT has completely failed, however, it is possible for COMPASS to process the data for that module. The following PMT failures have now been noted:

• D2-13, PMT-2 TJD 8717 07:40 UT
• D2-11, PMT-6 TJD 8736 01:35 UT
• D2-14, PMT-4 TJD 8755 00:28 UT
• D2-02, PMT-1 TJD 8988 23:47 UT

John Macri has provided the following info on PMT failures. These data provide only a general overview of each case. Further details can be found in CRAPs.

D2-01, PMT #1:
Problem first noted by ITOG on TJD 9165 (27-Jun-93) at 18:42. HV has been left on. Erratic behavior still observed.

D2-01, PMT #2:
Problem first noted by ITOG on TJD 9279 (19-Oct-93). HV has been left on. Erratic behavior still observed.

D2-01, PMT#4:
Problem first noted by ITOG on TJD 9340 (19-Dec-93) at 9:48 UT. HV has been left on. Erratic behavior still observed.

D2-01 module has been shut off August 3, 1994, TJD 9567 17:41 UT.

D2-02, PMT #1:
Problem first noted by ITOG on TJD 8654 (2-Feb-92) at 21:06. HV was left on. Total PMT failure on TJD 8988 (1-Jan-93) at 23:47.

D2-11, PMT #6:
Problem first noted by ITOG on TJD 8652 (31-Jan-92) at 13:22. HV was turned off from TJD 8655 (3-Feb-92) at 2:37 until TJD 8734 (22-Apr-92) at 17:31. Total PMT failure on TJD 8736 (24-Apr-92) at 1:35.

D2-13, PMT #2:
Problem first noted by ITOG on TJD 8645 (24-Jan-92). HV was turned off from TJD 8645 (24-Jan-92) at 16:23 until TJD 8658 (6-Feb-92) at 18:56. Total PMT failure on TJD 8717 (5-Apr-92) at 7:40.

D2-14, PMT#4:
Problem first noted by ITOG on TJD 8398 (22-May-91). HV was turned off from TJD 8401 (25-May-91) at 1:11 until TJD 8755 (13-May-92) at 00:28. Total PMT failure on TJD 8755 (13-May-92) at 00:28. This PMT has been showing some signs of life from approximately TJD 9500.

EVP Datagaps

The following are a list of days when no EVP data is available due to reboost activities or other technical problems.

Routine EVP Processing Checks

The task EVPRNN is part of the routine data processing flow at MPE. UNH has responsibility for insuring the integrity of the EVP data that is produced. In practice, this means that UNH assumes the task of reviewing the outputs from each run of EVPRNN.

In practice, there are three separate runs of EVPRNN performed for each day of COMPTEL data. These correspond to the gamma-1, gamma-2 and cal-3 data. UNH performs its processing checks only on the gamma-1 and gamma-2 data. (SRON has responsibility for the cal-3 data.)

The routine processing checks are performed at UNH in the following sequence :

1. EVPRNN jobs are executed at MPE. For each EVP file generated, a run of DDMCHK is also performed.
2. All log files from EVPRNN and DDMCHK are sent to UNH in the form of e-mail messages to the UNH EVP management.
3. A procmail process automatically saves each log file message to disk.
4. A kron job ("scan_tem") runs once per day and scans each new TEM file. The TEM files from EVPRNN are checked for any limit violations. The resulting report on the EVPRNN jobs is sent by e-mail to UNH EVP management. Normally, the TEM files from DDMCHK are inspected only if a problem shows up in the visual inspection of the DDMCHK postscript file.
5. DDMCHK postscript output is sent be MPE to groftp at UNH.
6. A kron job moves DDMCHK postscript files to /home1/scratch.2 and sends e-mail notification of their receipt.
7. The postscript output is visually reviewed and then discarded. Hardcopy output is not routinely produced.
8. Quality flags are set by UNH personnel using "mpe-qual" form.

Last Update: February 2, 1998