Concepts and terms of event-by-event data sorting
Sunsort is designed to sort what is referred to as event-by-event
data. In this style of data, an experiment records a sequence of
independent events. Each event is independent. A simple example
of an experiment which generates event-by-event data is a a coin tossing
experiment. A coin is tossed many times. After each toss the result (heads
or tails) is noted. In this experiment, an event is the action of a coin
being tossed and the event-by-event data is the list of heads and tails
that is being recorded.
Note that not everything that occurs in the experiment is necessarily an
event and not every event may make it to the data stream. Using the same
example, suppose that sometimes when the coin is tossed it misses the
table. The experiment could be set up to exclude these occurrences.
Alternatively, as often happens in real experiments, the rate of events
could be so high that the results may not be noted down as fast as the
events occur. In this case some events could be missing from the event
stream. If the events are independent, as they would be in a coin tossing
experiment and as they are in nuclear physics experiments, then missing
events will not affect the data analysis except by reducing the statistics.
The two verbs most commonly heard in relation to data analysis are
sorting and filtering. It is the former from which
Sunsort takes its name, and from which the term sort code
comes. In practice they mean the same thing. Probably the biggest problem
with data analysis is that you're looking for things which are right on
the limit of detectability (if it's not in the noise, it's not real
physics). One of the main things a sort code has to do is to
identify those events in which something interesting happened. The code
then sorts, or filters, the interesting events from the uninteresting
events.
Strictly speaking the verb filter is used when the interesting
events are being written out to create a data file consisting solely
of interesting events.
A spectrum is just a graph, to be precise, it's a histogram showing how
many times a particular value occurred. Sunsort allows both one and two
dimensional spectra. The values that are being graphed are divided into
bins or channels. For example, a
user may be making an energy spectrum. They might decide to make the bins
100 keV wide so that all energies between 34.35 and 34.45 MeV go into bin
344 (in Sunsort, bins are centred on integral values). The scaling applied
is called the dispersion and, in this case, would be quoted either as `100
keV per channel' or `10 channels per MeV'.
The word channel is one of the most over used words in physics. As
well as meaning a bin number in a spectrum it also
is used for the value returned on a particular signal from an experiment
as well as for the number of the signal, and in addition a reaction
channel is a particular way a reaction can proceed. This can lead
to odd sentences such as:
I was analysing the beryllium-neutron channel when I noticed an odd
effect in channel 32 - all signals above channel 2000 were missing.
Steven M. Singer
Last modified: Tue Aug 31 13:56:52 BST 1999