Main INDEX
Monthly INDEX
PREV
NEXT
User name R. Michaels
Log entry time 10:09:50 on January13,2005
Entry number 139148
Followups:
keyword=deadtime problem solved
We did "blaster" tests from each crate (i.e. send data at high
speed from crate to adaql2 host). It showed that TS1 had a
severe bottleneck (~30 kB/sec) while all others were fine (8-11 MB/s).
This was traced to the network switch, whose duplex setting for
TS1 was "full" but according to my notes should be "auto".
I have no clue how that got changed. Anyway, I changed it to
"auto" and we saw the performance immediately improve to 11 MB/sec.
This was corroborated by running DAQ with ps8=1 (pulser trigger).
Before we fixed TS1 we had ~70% deadtime, after fixing TS1 we
had zero deadime at 1024 Hz T8.