\documentstyle[11pt,a4wide]{article}
\title{Singles in Eurogam}
\author{Ian Lazarus}
\date{February 1990}
\begin{document}
\maketitle
 
\section{Assumptions and definitions}
I will assume that Eurogam will be built in VXI with a common deadtime over
the whole system of 15$\mu$seconds to cover shaping time, conversion time and
readout time. I will also assume that recovery time (i.e. time when it has
returned as close to zero as required by the system accuracy specification)
of the shaping amplifier is 10$\mu$seconds from the start of its input.
 
I will define two types of singles: `true' singles and `pseudo' singles. True
singles means that every detectable event on a channel is collected in a 1-d
histogram. This will
be limited by either the physics of the reaction or the
deadtime. Pseudo singles means that 1-d spectra are built up from
multiparameter event data but the system is not allowed to collect any singles
only events. The limit is the multiparameter event rate which will normally be
defined by the fold used in the trigger's multiplicity check.
 
\section{Which type of singles?}
 
\subsection{What's different?}
The difference between the two types is that true singles counts at a higher
rate per channel (10K plus background
 on each of 70 Ge channels) and has slightly
different data. The difference in data will arise because pseudo singles will
be biassed towards those parts of the array favoured by gamma rays from the
reaction with a consequent higher inclusion rate in multi-parameter events
while other parts of the array which detect low fold (e.g. doubles) or singles
only events will be cut out of the singles spectra so detectors which are
never involved in a large multiparameter event will never be histogrammed.
I don't know whether or not this will be significant in physics terms.
{\em Comments from physicists?}
 
\subsection{Rates}
The pseudo singles will histogram typically 3 or 4 different 
channels out of 70 for each collected event (assuming minimum fold of 3) at a
rate of around 30KHz. Making the simplifying assumption that we can ignore
directional bias in counting rates, we find that the total 
counting rate for true singles across the whole array is a factor of seven
greater than that for pseudo singles.\\
True singles rate is $10K \times 70 = 700 KHz$\\
Pseudo singles rate is $30K \times N = 100 KHz$\\ 
(where $N$ is fold, typically 3 or 4)\\
This means that the average rate per channel for pseudo singles (again ignoring
directional bias) will be between 1 KHz and 1.5 KHz. 
 
For setup with a source instead of beam the multiplicity in the trigger could
be cut to 1 which would allow a pseudo singles
rate over the whole array of up to 67 KHz, the limit imposed 
by deadtime. The common deadtime makes the system susceptible to
the influence of background; I assume that using a source and no beam this is
negligible anyway. I would guess that the strength of the source would limit
counting rate in this case. {\em Comments from physicists?}
 
 
 
\subsection{Quality of data}
Background counts  are random occurences which are completely
uncorrelated with the reaction under investigation and arise from such effects
as coulomb excitation of the target. True singles will collect good data and
background indiscriminately, mixing the two at a rate governed by physics.
The only exception would be when
the background rate is several times the 10 KHz singles rate from the main 
interaction, in which case the rate is limited by deadtime.
 
Pseudo singles are part of a coincidence event and so should produce data
virtually free from background contamination when run at folds of 3 and above.
 
\section{Implementation}
 
It is possible to design a system to collect either true singles or pseudo
singles. Technically there is little difference in complexity
between incrementing a
histogramming memory for every pulse received in a Ge card (true singles) or
incrementing a set of histogramming memories in a readout controller through
which all multiparameter data pass. 
 
The pseudo singles in the readout controller is slightly easier because the Ge
cards are sensitive to noise (limits readout of histogram to time the card is
not doing any analogue work) and board area is also limited by the need to
include as many Ge channels as possible per card. 
The readout controller will be a much less crowded card and will be all
digital (apart from possibly buffering the analogue sumbus for the trigger
system) so noise problems are either removed or reduced.
It is easier for the spectrum
readout processor to get all spectra from one source per crate
than to read many Ge cards although in either case a VME block transfer should
be used to read 256 bytes at a time so as to minimize the effect of bus
overheads like arbitration.
 
On the other hand it can be argued that true singles collected on the Ge cards
will provide better information for diagnostics, giving a true picture of what
the Ge detectors are seeing, including background.
 
\section{Conclusions}
 
It seems to me that pseudo singles with spectra built in the readout
controller is the better of the two options because it supplies cleaner data
than true singles with only a factor of 7 difference in rate, the actual rate
still being over 1 KHz per channel. It is slightly easier to implement
technically too.
 
The argument against it is that in the current version of the Ge electronics
the shaping circuit will process pulses from the detector while the trigger
and validation are being generated. The effect of the trigger decision will be
to either abort the event or allow the ADC to convert the shaped pulse.
The amplifier has a fixed recovery time after peaking
which is usually masked by the ADC conversion time, but is typically
10$\mu$seconds. When no trigger validates the input pulse
the amplifier still requires this
recovery time before it can accurately shape its next input, so it could be
argued that once the pulse has been shaped that we might as well convert it
and increment a local histogram while the amplifier is recovering.
Incrementing a local histogram takes much less time than reading out an event
because everything happens on the same 
card, typically in under 200 nanoseconds\footnote{More recently it has been
suggested that we use gated integrators instead of the semi-gaussian
shaping amplifier and in this case it is no longer true that true singles
incurs no time penalty; the penalty is the conversion time plus the update
time which is over 5$\mu$s.}.
This gives true singles. Since there will be no trigger unless the detected
pulse was part of a multiparameter event, the system will not be dead during
this time, so if the count is caused by a random gamma ray and is quickly
followed by a good event then the good event will not be affected.
 
Nevertheless it seems to me that on balance it is better not to add complexity
to the Ge card and use board area to collect true singles on every channel
even though this is possible with little or no effect on the throughput of
multiparameter events\footnote[1].
The advantages of collecting singles in one place 
(which has no board area problems) per crate
and of having singles data without background seem to be the most important
factors and point to the use of the pseudo singles method.
 
\end{document}