Why pipeline? (fwd)

[Elton Smith <elton@jlab.org>]

Elton Smith
Jefferson Lab
elton@jlab.org
(757) 269-7625

---------- Forwarded message ----------
Date: Tue, 16 Nov 1999 13:26:08 -0500 (EST)
From: Paul Smith <paul@xanadu.physics.indiana.edu>
To: elton@jlab.org
Cc: jonesrt@uconnvm.uconn.edu, dzierba@indiana.edu,
     teige@dustbunny.physics.indiana.edu, paul@xanadu.physics.indiana.edu
Subject: Why pipeline?

Hi,

Alex asked me to summarize the arguments for pipelining.  Here is what I wrote:

------------------------------------------------------------------------------


FAQ:  Why not use conventional, less expensive, front end electronics?

A:  For a 9 GeV tagged photon flux of 10**7/sec, the L1 trigger rate would be
about 9 KHz (tagged hadronic + accidentals, from figure 7.2 of the design
report) Assuming a 5 uSec conversion time gives only 4.5% deadtime.  However, at
3 * 10**7 tagged photons/sec, the tagged hadronic + accidentals trigger rate
climbs to 50 KHz, for a 25% deadtime.  The tagger also becomes ineffective in
reducing the trigger rate, which asymptotically becomes the total hadronic rate
of 100+ KHz for 3 * 10**7 tagged photons/sec, giving a 50% deadtime.  A
conventional electronics system would limit Hall D rates to less than 3 * 10**7
tagged photons/sec.  This also assumes all signals are split and are delayed by
cables.  The time available to form the level 1 trigger is fixed by the length
of these cables.  Note that there are ~30K channels in the Hall D detector.


Going to a pipelined electronics system replaces the delay cables with digital
memory.  The time available to form the level 1 trigger is now as long as the
memory depth, which could easily be 10 uSec or longer.  Rates of 10**8 tagged
photons/sec can be handled with no deadtime.  The electronics cost for a
pipelined system is estimated to be no more than twice that of a conventional
system, and this ignores the cost of the delay cables and splitters.


FAQ:  Even though the pipelined DAQ can handle high trigger rates, the tagger is
still useless for analysis since 20% or more of the events will have an
additional tagged photon for the high rate running.

A:  The tagger is not essential to the event analysis.  It only provides an
additional constraint.  Experience with the RadPhi experiment shows that events
with extra tagged photons can be analyzed.


------------------------------------------------------------------------------


It occurred to me that figure 7.2 in the design report is slightly misleading.  
The circles are labeled "Trigger Rate" but really are the sum of the Accidentals 
(triangles) and Tagged Hadronic (squares) rates.  For low rates the L1 trigger 
is this sum, but for high rates the L1 trigger approaches the Total Hadronic 
rate (diamonds).  Perhaps the circles could be relabeled and a new "L1 Trigger" 
curve added which follows the circles at low rates and follows the diamonds at 
high rates.  This is explained in the text at the end of 7.1.3, but isn't 
obvious from Figure 7.2.


Paul