Aim
To find out any remaining problems with SctRodDaq which could compromise our ability to do stand-alone readout.
Prerequisites
To do before Monday 23 Jan:- Endcap area DCS machine needs to be working - currently being repaired (Jo P)
- Create at SR1 version of SctRodDaq updated to tip of TDAQ14 branch
- Done, setup script at /daqsoft/sct/setup_tdaq14_branch.sh
- Check various modifications are in:
- Analysis checks for empty triggers (done AJB)
Tests with H8 Testbox (7 modules)
Calibration Mode
Ideally to be done before 23 Jan- efb and router firmware to be updated to latest versions
- Test sequence to be performed for each of the following module groupings:
- 7/0/0/0
- 2/2/2/1
- 1/2/4/0 (or 0/1/2/4 to make sure no dependence on group 0? BJG)
- Test sequence to comprise:
- NMaskTest
- PipeLineTest?
- FullBypassTest
- This test fails about 30% of the time at bin 20, when the input masks are first changed. Under investigation.
- StrobeDelayTest
- ThreePointGainTest
- NoiseOccupancyTest
- DoubleTriggerNoiseTest
- Problems when no modules in group 0. Scan completes, but histograms are empty. Applies to new firmware only (?). Under ivestigation by John Joseph.
- SynchTriggerNoiseTest?
- Was broken by tim enables setting in setRunMode command. Fixed by setting tim register 0x18 to 0x600 - Alan
- SynchDoubleTriggerNoiseTest?
- Trigger rate revised. Now OK.
- PreTriggerNoiseTest? ok if proper NIM signal is sent
- could do with aborting if no triggers are coming through!
- NMaskTest
- Used to fail after any scans with TIM triggers, in fact this was due to the failure of the preceeding test. Now OK.
- Restart
- Run "Full mode" test. PreTriggerNoise? is an example of one
- Full mode now ready for analysis
- Run "Full mode" test. PreTriggerNoise? is an example of one
Tests with Barrel Sector (15 modules)
To do during the testing week 23 Jan to 27 JanCalibration mode
- efb and router firmware to be updated to latest versions (subject to results with H8Testbox modules)
- Test sequence to be performed for each of the following module groupings:
- 15/0/0/0
- 12/3/0/0
- 6/7/0/0
- 3/4/4/4
- Test sequence as described above.
Tests with either Barrel Sector or H8 testbox
To do during the testing week 23 Jan to 27 Jan- A first attempt was made, the sector ran but slowly. This led to a workaround for the well known "Suspiciously Precisely" message, which involved the uneceesary tranfer of 32k blocks of trash data between ROD and host.
- A second attempt has now been made. The four ABCD2 modules were quickly removed from the configuration, as was one module which cannot be biassed, however the noise occupancy scan remained very slow.
- Module ...07 was found to have an incomplete set of trim target information. This was in turn causing 100% occupancy for 11/12 chips of this module. Once corrected, this module group runs at a reasonable speed.
- Module ...16 has four dead chips which are usually masked off.
- If one runs NO by itself, the speed is now up to the expected level. If one runs it as part of a sequence, it is still very slow. This was traced to module ...16, which has four dead chips. These are masked off in the configuration, but it seems that after the pipeline test all channels are enabled, even the ones which are masked in the starting configuration. ***Theoretical bugfix by BJG installed in TDAQ_14 version. To be tested with Sector.
Physics mode
- Is the TIM busy logic correct?
- Are there remaining ROD busies?
- L1ID and BCID correct during physics running?
- Document work arounds
Tests with Imaginary Detector (lots of modules)
To do during the testing week 23 Jan to 27 Jan- Run state transitions and BOC scans for dummy config containing several hundred modules
- Use "cosmics" xml config file
- Run state transitions
- BOC scans
- Calibration controller - reports that this fails with dummy configuration - seems to be related to a 30 s corba timeout -- Alan
- Work-around - set > 30 second timeout for BOTH CalibrationController and SctApiServer?
- Complete fix - make initialise a rapid-return method, and then poll on isInitialised()
This part signed off by Alan 23/01/2006
DCS
To do during the testing week 23 Jan to 27 Jan- Preliminary investigation of DIM
General
- Does the ErrorMask? work?
- Running analysis and fitting on different machines
- pmg agent wont start on pcphsctr03
- Fixed by opening ports in firewall? (Ok now - AJB)
- pmg agent wont start on pcphsctr03
Other Things
(Some of which border on functional enhancements...)High Priority
- Analysis of full mode scans
- done AJB 30-1-2006
- no OPE - since full mode scan
- occupancy seems higher in central time-bin - is this expected?
- was caused by modules being in X1X mode. Fixed in any-hit mode. AJB
- Permit more than one error per scan before abort?
Low Priority
- zipped histograms
- read out errorInfo structs
- mask faulty streams (only after above completed)
Post Script
- Sector non-zippy speed check (PWP)
- OK
- Sector zippy check (PWP)
- OK
- Masking of channels in CAL loop scans
- fixed in PWP dev dsp and zippy
- MSR config send (AJB)
- STN bin 0 (AJB)
- Not a problem. Log file is misleading as ROD bin count does not increment, but SctApi bin counter seems to be OK
- GUI Analysis names (DR)
- fixed
- GUI DCS display (DR)
- Physics as above (AJB+DR)
- Check mappings (PWP+DR)
- FBT (with and without zippy) (PWP)
- BJG investigating issues with link masking
- With the Barrel sector, 10 modules in one group can cause the FBT to abort on bin 35 / event 99. A subtle change was made in the dsp code, but this may not be entirely successful. To be confirmed.
- zippy-meme + on-rod boc (BG)
- RxThr? Lester improved (ok AJB)
- OPE with full (PWP)
- CC signal handling (CGL)
- TDAQ versions
- groups in real config
- fibre rods
- fibre s-links + test H8 test box with s-link
- relook at analysis pass/fail
- Physics mode with SLINK in dosent work even with no_slink API option AJB
- FullMode