\(\renewcommand\AA{\unicode{x212B}}\)

GetAllEi v1

../_images/ImageNotFound.png

Enable screenshots using DOCS_SCREENSHOTS in CMake

Summary

Analyze the chopper logs and the signal registered by the monitors to identify energies used as incident energies in an inelastic experiment.

See Also

GetEi

Properties

Name

Direction

Type

Default

Description

Workspace

Input

MatrixWorkspace

Mandatory

The input workspace containing the monitor’s spectra measured after the last chopper

Monitor1SpecID

Input

number

Optional

The workspace index (ID) of the spectra, containing first monitor’s signal to analyze.

Monitor2SpecID

Input

number

Optional

The workspace index (ID) of the spectra, containing second monitor’s signal to analyze.

ChopperSpeedLog

Input

string

Defined in IDF

Name of the instrument log, containing chopper angular velocity. If ‘Defined in IDF’ option is specified, the log name is obtained from the IDF

ChopperDelayLog

Input

string

Defined in IDF

Name of the instrument log, containing chopper delay time or chopper phase v.r.t. the pulse time. If ‘Defined in IDF’ option is specified, the log name is obtained from IDF

FilterBaseLog

Input

string

Defined in IDF

Name of the instrument log, with positive values indicating that instrument is running and 0 or negative that it is not. The log is used to identify time interval to evaluate chopper speed and chopper delay which matter. If such log is not present, average log values are calculated within experiment start&end time range.

FilterWithDerivative

Input

boolean

True

Use derivative of ‘FilterBaseLog’ rather then log values itself to filter invalid time intervals. Invalid values are then the values where the derivative of the log turns zero. E.g. the ‘proton_chage’ log grows for each frame when instrument is counting and is constant otherwise.

MaxInstrResolution

Input

number

0.0005

The maximal energy resolution possible for an instrument at working energies (full width at half maximum). Peaks, sharper then this width are rejected. Accepted limits are: 1e^(-6)-0.1

MinInstrResolution

Input

number

0.08

The minimal energy resolution possible for an instrument at working energies (full width at half maximum). Peaks broader then this width are rejected. Accepted limits are: 0.001-0.5

PeaksRatioToReject

Input

number

0.1

Ratio of a peak energy to the maximal energy among all peaks. If the ratio is lower then the value specified here, peak is treated as insignificant and rejected. Accepted limits are:0.0 (All accepted) to 1 – only one peak (or peaks with max and equal intensity) are accepted.

IgnoreSecondMonitor

Input

boolean

False

Usually peaks are analyzed and accepted only if identified on both monitors. If this property is set to true, only first monitor peaks are analyzed. This is debugging option as getEi has to use both monitors.

OutputWorkspace

Output

Workspace

Mandatory

Name of the output matrix workspace, containing single spectra with monitor peaks energies together with total intensity within each peak.

Description

Algorithm finds the estimate for all incident energies allowed by chopper system of an inelastic instrument and returns a workspace, with the estimates for positions, heights and width of incident energies provided by the choppers and registered on monitors. These estimates can be used as guess value for GetEi v2 algorithm or as inputs for a peak fitting procedure.

Algorithm performs number of steps to identify the values requested:

  1. It takes appropriate log names from instrument definition file (IDF), namely chopper-position component and calculates last chopper speed and delay as average of the filtered log values. Guess chopper opening times are calculated from chopper speed and delay time. The “chopper-position” component with appropriate properties has to be present in IDF for this algorithm to work. See ISIS MARI or MAPS instrument definition files for example of “chopper-position” component.

  2. Algorithm uses estimate for the minimal energy resolution of an instrument and searches for real peaks around guess values obtained earlier within 4 sigma of this resolution interval.

  3. If peaks are found, the algorithm performs running averages over signal in the appropriate time interval until first derivative of the signal has only one zero. The position of this zero is returned as the guess energy and the distance between closest to the guess energy zeros of the second derivative are returned as the guess values for the peak width. The peak amplitude is estimated from the total intensity of the signal within the search interval, assuming that the peak shape is Gaussian.

  4. Similar procedure is performed for second monitor. The peak is accepted only if the peak width lies between the minimal and maximal instrument resolution values and the distance between peaks positions on two monitors (on energy scale) is smaller then two sigma.

Algorithm returns matrix workspace containing single spectrum, with x-values representing peak positions, y-values: peak heights and the error: peak width. X-values are sorted according to energy in peaks (peaks with maximal energy are returned first).

Used Subalgorithms

The algorithm uses Unit Factory and ConvertUnits v1 algorithm to convert units from TOF to energy.

Example: Find all incident energies for test workspace

# BUILD SAMPLE WORKSPACE
# Build sample workspace with chopper and in energy units to
# have defined peaks in defined energy positions
wsEn=CreateSampleWorkspace(Function='Multiple Peaks', NumBanks=1, BankPixelWidth=2, NumEvents=10000, XUnit='Energy', XMin=10, XMax=200, BinWidth=0.1)
# convert units to TOF to simulate real workspace obtained from experiment
ws = ConvertUnits(InputWorkspace=wsEn, Target='TOF')
# find chopper log values would be present in real workspace
l_chop = 7.5  # chopper position build into test workspace
l_mon1 = 15. # monitor 1 position (detector 1), build into test workspace
t_mon1 = 3100. # the time of flight defined by incident energy of the peak generated by CreateSampelpWorkspace algorithm.
t_chop = (l_chop/l_mon1)*t_mon1
# Add these log values to simulated workspace to represent real sample logs
AddTimeSeriesLog(ws, Name="fermi_delay", Time="2010-01-01T00:00:00", Value=t_chop ,DeleteExisting=True)
AddTimeSeriesLog(ws, Name="fermi_delay", Time="2010-01-01T00:30:00", Value=t_chop )
AddTimeSeriesLog(ws, Name="fermi_speed", Time="2010-01-01T00:00:00", Value=900 ,DeleteExisting=True)
AddTimeSeriesLog(ws, Name="fermi_speed", Time="2010-01-01T00:30:00", Value=900)
#-------------------------------------------------------------

# FIND GUESS PEAKS
allEiWs=GetAllEi(ws,Monitor1SpecID=1,Monitor2SpecID=2)
# Analyze results
allEi = allEiWs.readX(0);
peakHeight = allEiWs.readY(0);
peakWidth  = allEiWs.readE(0);

# Check if peaks positions are indeed correct:
#-------------------------------------------------------------
resEi=[]
for ei_guess in allEi:
   nop,t_peak,monIndex,tZero=GetEi(InputWorkspace=ws, Monitor1Spec=1, Monitor2Spec=2, EnergyEstimate=ei_guess)
   resEi.append((nop,t_peak));
print("! Guess Ei ! peak TOF ! peak height ! peak width !")
for ind,val in enumerate(resEi):
   print("!  {0: >6.1f}  !  {1: >6.2f} !   {2: >6.2f}    ! {3: >6.2f}     !".format(allEi[ind],val[1],peakHeight[ind],peakWidth[ind]))
#
# NOTE: incident energy of GetEi is calculated from distance between monitor 1 and 2, and this distance is not correct in
# the test workspace. The tested point is that getEi can find energies from guess values and TOF for peaks is correct.

Output:

! Guess Ei ! peak TOF ! peak height ! peak width !
!    67.0  !  4188.03 !    34.68    !   2.35     !
!   124.1  !  3079.09 !    14.01    !   4.35     !

Categories: AlgorithmIndex | Inelastic\Ei

Source

C++ header: GetAllEi.h

C++ source: GetAllEi.cpp