Choose timezone
Your profile timezone:
Developer's meeting
→
Europe/Zurich
present: Milo, Dan, Frank, Roland
Compression (Milo):
- using Eric's MicroBooNE data, reformatted to conform to ProtoDUNE WIB format
- started testing several compression algorithms: gzip, lz4, lzma, Huffman (the latter needs customisation but in principle seems most promising)
- compression achieved thus far is not yet reaching a factor 4; but the present status is the result of a few days' work only, so confident that it can be improved
- Huffman option:
- one single "pedestal" value common to all channels (seems appropriate enough for the MicroBooNE data used here)
- compression scheme works with essentially the full noise distribution being contained in a range of 32 ADC values
- error information is the biggest unknown -- not sure what's contained here exactly
- Dan provided some information concerning the error words etc. (and will point Milo to more detailed information):
- the WIB trailer CRC should be valid; but the COLDATA checksums are not (but Dan's firmware code should leave sufficient information to trace errors)
- the 16-bit error word in the WIB header should be 0 (if there are no errors); there are 2 bits to indicate errors for each physical COLDATA stream
- discussion:
- it may be appropriate to allow for other Huffman schemes: per-channel pedestal, record difference w.r.t. previous ADC value, etc.
- the question of what to do with CRCs, checksums, and other (header) error information is not specific to FELIX; a discussion with RCE seems useful
Next meeting: in principle, on June 14. Roland will let us know if he is ready to discuss his investigation of QuickAssist already next week.
There are minutes attached to this event.
Show them.