Development of the European Scanning System

1
Progress Report
Development of the European Scanning System
Development of the European Scanning System
Development of the European Scanning System
Cristiano Bozza European Emulsion Group - LNF,
Oct 2002
2
Development of the European Scanning System
progress report
The Mikrotron camera with the Photobit CMOS
sensor can now work _at_ 87 fpsThe field of view
(excluding overlapping regions) is 350 270
mm2We can reach a speed as large as ?14 cm2/h
with 16 layers sampling
Obtaining such performance is not straightforward
Since the thread timeslice under Win2k is 10 ms,
and the camera cycle time is11.4 ms, 1/3 of the
fields is missed because of the software
architecture(the frame grabbing is started by
the host PC CPU) Actually, we only get 58 of the
87 fps...
3
Development of the European Scanning System
progress report
Frame 1
Frame 2
Frame 3
DAQ Thread
Other Thread
DAQ Thread
Frame 2 is missed because of the time structure
of the present DAQ software
This was expected, and we have been working on an
asynchronous frame grabbing model which has now
passed all technical tests An asynchronous
version of the scanning module (VertigoScan3)
is already being debugged, and is expected to be
ready for production within 2 weeks
4
Development of the European Scanning System
progress report
Old (Production version) time structure of the
DAQ cycle
Image grabbing (Camera)
Field N
Field N
Image Processing (Genesis)
Field N
Image Processing (Genesis)
Field N
Image Processing (Genesis)
Field N
Clustering (PC CPU)
Tracking (PC CPU)
Field N
XY Motion
Z Motion
The command to grab each frame is issued by the
PC CPU Clustering must stay in sync with
grabbing CPU idle for gt 50 of the time, whereas
tracking may slow down DAQ Avoid disk swapping
(cant use standard memory allocation) The thread
timeslice is 10 ms... Good for 60 fps (16.7 ms
cycle) but not for 87 fps (11.4 ms) Frames could
be missed Rough Z readout
5
Development of the European Scanning System
progress report
New time structure of the DAQ cycle
Image grabbing (Camera)
Field N
Field N
Image Processing (Genesis)
Field N
Image Processing (Genesis)
Field N
Image Processing (Genesis)
Field N-1
Clustering (PC CPU)
Tracking (PC CPU)
Field N-1
Genesis programming for next field (PC CPU)
FlexMotion programming for next field (PC CPU)
Trajectory data readout (PC CPU)
Field N
XY Motion
Z Motion
The CPU pre-loads the commands to grab frames and
to readout several Z samples At each field, only
a start command is issued by the CPU Longer
time available to perform tracking Disk swapping
does not cause missing frames No problem with
faster cameras Very precise Z readout Larger
memory occupancy on Genesis nodes The core cycle
needs reshaping (1 field of view delay)
6
Development of the European Scanning System
progress report
A brick was exposed in Sept 2002 to test vertex
location
In Salerno, we got 5 sheets 6 CS
Unfortunately, thickness ? 30 mm AGAIN!
First task Sheet to sheet intercalibration
Second task CS intercalibrationwith sheet 1
Third task CS intercalibrationwith Si-tracker
Fourth task Scanback of possibleinteraction
product tracks up tovertex
7
Development of the European Scanning System
progress report
Sheet to sheet intercalibration
4 zones 1 cm2 each were scanned on each sheet
Pattern matching was then applied from eachsheet
to its nearest neighbours
For each sheet, the 3 zones with the highest
matching numbers are selected out the 4 to
compute the alignment parameters
This work was successfully completed for all 4
zones on all 5 sheets by using the Salerno
prototype with provisional optics, no automatic
light controland a CHORUS camera (Hitachi
KP-F110 _at_ 30 fps) Scanning speed ? 4 cm2/h
8
Development of the European Scanning System
progress report
Sheet to sheet intercalibration
200?400 tracks match per each zone The resulting
precision is satisfactory, and even better than
needed forthe scanback phase
9
Development of the European Scanning System
progress report
Sheet to sheet intercalibration
OK
First task Sheet to sheet intercalibration
In progress
Second task CS intercalibration with sheet 1
Third task CS intercalibration with Si-tracker
Fourth task Scanback of possible interaction
product tracks up to vertex
Strategy under study... Scanback track
disappearance? Multitrack crossing? Both?