Table of Contents
Reduce the number of events in an EventWorkspace by grouping together events with identical or similar X-values (time-of-flight).
Name | Direction | Type | Default | Description |
---|---|---|---|---|
InputWorkspace | Input | EventWorkspace | Mandatory | The name of the EventWorkspace on which to perform the algorithm |
OutputWorkspace | Output | EventWorkspace | Mandatory | The name of the output EventWorkspace. |
Tolerance | Input | number | 1e-05 | The tolerance on each event’s X value (normally TOF, but may be a different unit if you have used ConvertUnits). Any events within Tolerance will be summed into a single event. |
WallClockTolerance | Input | number | Optional | The tolerance (in seconds) on the wall-clock time for comparison. Unset means compressing all wall-clock times together disabling pulsetime resolution. |
StartTime | Input | string | An ISO formatted date/time string specifying the timestamp for starting filtering. Ignored if WallClockTolerance is not specified. Default is start of run |
The algorithm can be run in two modes, with and without pulsetime resolution. This switch is made by specifying a WallClockTolerance. The StartTime is ignored unless WallClockTolerance is specified.
This algorithm starts by sorting the event lists by TOF (or whatever the independent axis is) and ignoring the pulsetime. Therefore you may gain speed by calling SortEvents v1 beforehand. Starting from the smallest TOF, all events within Tolerance in a spectrum are considered to be identical. A weighted event without time information is created; its TOF is the average value of the summed events; its weight is the sum of the weights of the input events; its error is the sum of the square of the errors of the input events.
Note that using CompressEvents may introduce errors if you use too large of a tolerance. Rebinning an event workspace still uses an all-or-nothing view: if the TOF of the event is in the bin, then the counts of the bin is increased by the event’s weight. If your tolerance is large enough that the compound event spans more than one bin, then you will get small differences in the final histogram.
If you are working from the raw events with TOF resolution of 0.100 microseconds, then you can safely use a tolerance of, e.g., 0.05 microseconds to group events together. In this case, histograms with/without compression are identical. If your workspace has undergone changes to its X values (unit conversion for example), you have to use your best judgement for the Tolerance value.
Similar to the version without pulsetime resolution with a few key differences:
While the algorithm can be run with arbitrary values of
WallClockTolerance, it is recommended to keep in mind what
resolution is desired for later filtering. The pulsetime is
effectively filtered on a bin of the form:
The StartTime property is only used in pulsetime resolution mode. Any events that occur before it in a run are ignored and do not appear in the OutputWorkspace. If it is not specified, then the Run.startTime is used. An example ISO8601 format for the StartTime is 2010-09-14T04:20:12. Normally this parameter can be left unset.
Example
ws = CreateSampleWorkspace("Event",BankPixelWidth=1)
print("The unfiltered workspace {} has {} events and a peak value of {:.2f}".format(ws, ws.getNumberEvents(), ws.readY(0)[50]))
ws = CompressEvents(ws)
print("The compressed workspace {} still has {} events and a peak value of {:.2f}".format(ws, ws.getNumberEvents(), ws.readY(0)[50]))
print("However it now takes up less memory.")
Output:
The unfiltered workspace ws has 1900 events and a peak value of 257.00
The compressed workspace ws still has 1900 events and a peak value of 257.00
However it now takes up less memory.
Categories: Algorithm Index | Events
C++ source: CompressEvents.cpp (last modified: 2018-10-05)
C++ header: CompressEvents.h (last modified: 2018-10-05)
Python: CompressEvents.py (last modified: 2019-02-26)