[DAQ Page]
a)Number of Detectors
1) Run 19 - 6 BLIPS
2) Run 20 - Starting 9/1/99 - 6 ZIPS
3) CDMS-II - Starting 01/01/01 - 18 ZIPS
Final 42 ZIPS
4) Channel Count:
i)BLIP: 2 Ionization - 5 microsecond rise
50 microsecond fall
2 Phonon - 5 millisecond rise
50 millisecond fall
ii)ZIP: 2 Ionization - 2 microsecond rise
50 microsecond fall
4 Phonon - 2 microsecond rise
50 microsecond fall
b)number of veto elements
1) CDMS-I: 26 PMT's, grouped to 13 channels
1 ns rise, 10 ns fall, stretched to 1.2 microseconds
1) CDMS-II: 42 PMT's, grouped to 42 channels
1 ns rise, 10 ns fall, stretched to 1.2 microseconds
c)Sampling Rate, Resolution
(see also Ray Bunker's Digitizer Page)
1) CDMS-I:
* 12 bit, over -X to +Y V range - 2 Bytes/Sample
* BLIP
i)Ionization: 1000 samples, 0.5 microsecond/sample,
500 microseconds, 375 musec after, 125 musec before
2000 bytes
ii)Phonon: 2000 samples, 64.1 microsecond/sample,
128 milliseconds, 96 ms after, 32 ms before;
4000 bytes
* ZIP
i)Ionization: 10000 samples, 0.125 microsecond/sample, 1250 microseconds,
20000 bytes
ii)Phonon: 10000 samples, 0.125 microsecond/sample, 1250 microseconds,
20000 bytes
* Time History 12500 samples, 0.83 microseconds/sample, 10.4 milliseconds,
150000 bytes (96 Channels, 12 bytes/time slice)
2) CDMS-II:
* 12 bit, over -X to +Y V range - 2 Bytes/Sample
* ZIP
i)Ionization: 10000 samples, 0.125 microsecond/sample, 1250 microseconds,
20000 bytes
ii)Phonon: 10000 samples, 0.125 microsecond/sample, 1250 microseconds,
20000 bytes
iii)Veto: 450 kilobytes (a guess, use 3 time history units)
d)`data' Event Size
1) CDMS-I : 150 + 12*B + 120*Z kilobytes (B=6, Z=0 get 222 kilobytes)
(B=0, Z=6 get 870 kilobytes)
B= number of BLIPS
Z= number of ZIPS
2) CDMS-II: 450 + 120*Z kilobytes (Z=42, get 5.5 Megabytes)
Z= number of ZIPS
e)number of `slow control/monitoring' bytes/time interval
f)trigger rate (including calibrations)
1)Calibration: let's guess, for ZIP's, that one segments
each detector quadrant 4 by 4 by 4, or
64 bins, for 256 total bins. And,
in each bin, you want 100 good events, and
you need to take 500 triggers to get 100 good
events. So, that is 1.3 * 10^5 triggers per
detector, or 5.4 * 10^6 triggers for the whole
7 towers.
Now guess, you'd like to accomplish that in
4 days, or 3.5 * 10^5 seconds.
That is a trigger rate of
5.4/0.35 = 15 Hz,
or about 2.1 Hz per tower, or about 0.36 Hz per hocky puck,
or about 0.09 Hz per phonon channel.
To deal with this rate, we need to read out only a
fraction of the detector; in particular, only those
detectors that fire a minimum threshold. We can
work backward from a 100 BaseT ethernet bandwidth of
3 Mbyte/second; the information from one ZIP will
comprise 0.12 Mbyte; to stay above a livetime of
90%, we need to stay under 0.3 Mbyte/s. If we read
out 0.12 Mbyte at 15 Hz, then, that would be
15*0.12=1.8 Mbyte/second; we'd need to achieve another
factor of 6 on this, by, say, decimation of the trace.
The trigger must be reenabled in less than
6 milliseconds, in order to keep the front-end livetime
from falling below 90%.
2)Normal Data Taking: let's plan for a maximum trigger rate of
1/3 Hz; that is a factor of 3 higher than
that estimated in the proposal of 0.1 Hz.
To read out the entire detector at 1/3 Hz,
we'd need an available bandwidth of
5.5 Mbyte/event * 1/3 * event/s
= 1.65 Mbyte/second
The maximum 100 BaseT transfer rate is about
3 Mbyte/second; to keep a livetime of >90%,
we'd need to reduce the needed bandwidth by
a factor of 6 or so. That implies that we
must decimate the traces.
The algorithm to decimate the trace must go fast
enough to not add deadtime at the front end;
there is 3 seconds available on average between
events (from the 1/3 Hz), so the algorithm must
be done in 300 milliseconds or so. This is
actually a looser requirement than needed for
the case of calibration.
[DAQ Page]