Make New Entry, Make Followup Entry
User name R. Michaels
Log entry time 20:47:19 on February26,2011
Entry number 348899
keyword=DT mystery resolved (NO deadtime)People have been worrying about DAQ deadtime ...
I came in and investigated.
From end-of-run GUI, with cosmics runs at 5 Hz, the DT is
often reported as 12 to 25% for prescale = 1 for short runs.
After several runs I realized the problem is that the
scaler for T1, T2, T3, etc are counting a bit at the start
of run and also at the end, while L1A is obviously zero rate
until DAQ is running. See for example fig 1. Historically
we had precise gating of all scalers by the Trig. Super.,
but apparently not for this experiment.
To correct this it will be necessary to compute the DT
differentially during the run using scalers.
Example: run 1959 with ps3=1, others zero
Scaler counts for T3 and L1A, obtained from server.
Done "by hand":
T3 .... L1A .... DT
514 ... 395
568 ... 447 ... 3.8% (54 T3 and 52 L1A since last count)
641 ... 519 ... 1.3% (73 T3 and 72 L1A)
708 ... 583 ... -4.5%
821 ... 698 ... zero
etc
So the DT is zero, although the end run GUI says 12% for run 1959.
I observed that the halla_scaler_process.tcl is correct
in some details (finding correct data, processing the prescales,
computing DT) but I did not check about correlations.
I'm assuming Vince took care of that.
However, I ran with single triggers enabled only, so correlations
are irrelevant for my tests.
The L1A scaler is correct because it always agrees with number
of events in the datastream (I checked this for several runs).
As an example to correct the DT for run 1959 :
T3 has an offset of 514 - 395 = 119 at start of run
because the normalization scaler is not properly gated.
T3 continues to count for about 6 sec after, so at 5Hz = 30 counts.
There were 1250 physics events (plus assorted EPICS and
scaler events).
The raw T3 count was 1423.
The GUI computes DT = 1- 1250/1423 = 12.2 % which is WRONG.
The corrected result is DT = 1 - 1250 / (1423 - 119 - 30) = 1.9%
which is normal for cosmics.
We may want to fix the gating of these scalers, like we
have done historically. But as I said the DT can easily
be corrected using the full information from scalers
recorded in the datastream.
Bottom line: NO DEADTIME, and we were chasing an ILLUSION.
FIGURE 1
FIGURE 2
