Make New Entry, Make Followup Entry
User name Bob / Alex
Log entry time 10:52:29 on March 4,2011
Entry number 350046
This entry is a followup to : 350012
keyword=modification to bb_sfi3.crlWe modified bb_sfi3.crl so that it has a flag "onlytriglatch"
# if onltriglatch==1, then only read the TDC with trigger latch info (slot 23)
const onlytriglatch = 1
To reverse this, set it =0.
Using this trick, the DT on ROC10 is 30 usec.
Running buffered, see datamon = 7%. Not surprised it didn't change.
Raw data dump of ROC10 in fig 1. Run 2304.
Note 0xb8000000 means slot 23, which is the trigger latch TDC.
Somebody should double-check this by analyzing some data.
Return to Unbuffered mode. config UB2ArmNTS11.
DT = 26% with 2 kHz. Note, it was 40% before, so this is
an improvement for the unbuffered mode, which is directly sensitive
to this deadtime source.
Note 0.40 / 2000 = 200 usec (previuos DT, dominated by ROC10)
Now, 0.26 / 2000 = 130 usec -- roughly consistent with ROC4.
Per discussion with Doug Higinbotham, we decide to run ROC10
with the full readout (onlytriglatch = 0) and use buffered mode.
This should be fine, since buffering makes you somewhat immune
from the readout deadtime.
Now we compile with
const onlytriglatch = 0
and use the config B2ArmNTS11.
Run 2307 is with these conditions.
We will leave the DAQ like this.
Last observation was 7% DT at 2.5 kHz DAQ rate.
---------------- Appendix ------------------
What was "wrong" with ROC10 readout ?
The event size was quite small, but the readout was inefficient.
Instead of a MULTIBLOCK readout, the CRL code addresses
each TDC individually. This can lead to a factor of 2 or 3
increase in the readout time. It would be a lot of work to
change this because the TDCs are spaced out with 1-slot gaps
in between, and since the SFI is in the middle of the crate
we would need two MULTIBLOCKs if we didn't move the SFI.
FIGURE 1
