SLIDE 1
Introduction Implementation Summary
Introduction Implementation Summary
Online Data Quality Monitoring in the PXD Peter Kodys and Peter - - PowerPoint PPT Presentation
Online Data Quality Monitoring in the PXD Peter Kodys and Peter - - PowerPoint PPT Presentation
Introduction Implementation Summary Online Data Quality Monitoring in the PXD Peter Kodys and Peter Kvasnicka Charles University, Prague Belle II PXD DAQ/Trigger Workshop Bayrischzell, 17 - 21 July 2012 Introduction Implementation Summary
SLIDE 2
SLIDE 3
Introduction Implementation Summary High-level PXD DQM
The PXD DQM
The PXD DQM system has to monitor the data-taking performance of the PXD. Low level: Peroformance of sensor data processing pixel signal quality (pedestal/CMN subtration, gain equalization, masking...) clustering . . . High level: Physics performance of the PXD Data reduction Alignment Timing . . .
SLIDE 4
Introduction Implementation Summary High-level PXD DQM
The high-level DQM
In this talk: Motivations for the high-level part of PXD DQM For impelementation, see Itoh-san’s talk on PXD DQM implementation within ExpressReco. High-level DQM needs VXD pattern reco + tracking We need to monitor the physics performance of the PXD. Performance of the ATCA data compression Correctness of alignment Timing . . . and any sudden changes thereof. To achieve this, we need sampled events (not just histograms)
SLIDE 5
Introduction Implementation Summary Data compression monitoring
Why monitor the data compression peformance?
Data compression monitoring Data compression performance is critical to the vertexing performance of Belle II. For tracks that only hit Si detectors we don’t have enough information to check data compression performance. Should be tuned to current experiment conditions. What is needed to monitor data compression performance Needs analysis of full-frame data Needs an independent (on ATCA algorithms) SVD+PXD pattern reco engine Needs access to configuration data
SLIDE 6
Introduction Implementation Summary Data compression monitoring
Requirements
Full frame PXD data Full frame PXD data will be saved during injection veto. Will they undergo normal processing?
This is needed for comparison with off-line analysis on uncompressed data. The goal is to estimate the sensitvity of the data compression algo.
Part of the analysis (for HLT-reconginzed tracks) can be done using compressed data. Independent pattern reco engine (Means independent on DATCON algorithms) Quite trivial requirement: The DATCON algorithms can hardly be used.
basf2 vs. FPGA (massively parallel) processing different (use of) SVD timing information ...
We could use Jakob Lettenbichler’s cellular automata or a combination
- f several pattern reco algorithms, standard basf2 tracking
SLIDE 7
Introduction Implementation Summary Data compression monitoring
Performance considerations
Performance Nearly real-time, will only process events with full-frame data
and also accumulate other useful data for PXD (pixel gains?)
The combinatorial complexity will be higher for full frame events, but the tools to manage it will be similar or better than those in the DAQ:
better use of SVD timing information higher algorithmic freedom than in FPGA.
The critical performance issua are low-Pt tracks with their complex signature in the Si detectors and long lifetime. Performance can be easily scaled by selection rate of analyzed events.
SLIDE 8
Introduction Implementation Summary ExpressReco
Implementation in ExpressReco
Itoh-san’s ExpressReco design
SLIDE 9
Introduction Implementation Summary ExpressReco
Implementation in ExpressReco
ExpressReco ExpressReco provide real-time processing of (sampled) event data from the recording nodes. Hardware: Several units of 20 PCs each. Software: basf2 clone similar to that used by HLT. ExpressReco feaures: Fully functional basf2 environment Raw data from all detectors available Configuration data (frozen) available The complete environment for pattern reco / tracking is available.
SLIDE 10
Introduction Implementation Summary
Summary
Open questions Can we have full frame data together with ROI data by DATCON / ATCA? Conclusions A pattern reco / tracking engine operating on PXD full frame data is required to monitor the performance of PXD data compression. The same will also be used for monitoring overal physics performance of PXD+SVD. The ExpressReco system proposed by Itoh-san provides a good environment for these tasks.
SLIDE 11
Introduction Implementation Summary