Comparison of Simulation Data
with Calibration Data
Derek Hullinger
7 Jan 2004
Bottom Line:
The number of counts in the calibration data is about 24% higher than
the number of counts in the simulated data.
However, the predicted number of counts (from calculation) is only about 8%
higher than the number of counts in the simulated data.
(The values and histogram plots given below were generated with this
IDL routine.)
Comparison of Spectra:
- Both data sets were processed using sum_spectra.
- For both data sets, the following detectors were excluded:
- "bad" detectors (i.e., hot or
disabled detectors)
- edge (i.e., non-center) detectors
- detectors for which good linearity correction
factors are not known
- detectors for which μτ values are not known
- Before comparing the two, the calibration data was divided by a factor of 1000
(to compensate for
the fact that the simulation source was 1000 times weaker) and by a factor of
1.17 (to compensate for the fact that the simulation source stength was based
on the activity of the source on 8/18/03 whereas the calibration data was taken
on 6/19/03).
- The mask-weighting factors for the simulated data set were generated using:
- maskoffx = 0.107
- maskoffy = -0.232
- maskoffz = -1.014
whereas the mask-weighting factors for the calibration data set were
generated using maskoff[xyz] = 0.
This is because of differences between the simulation mask position,
the actual mask position, and definitions of the mask position in
batmaskwtimg.
The simulation offset values were found by running batfftimage and
batcelldetect
several times until values were found which reproduced the
correct source position.
Results:
Calibration Data
Calibration Data (Scaled to 122 keV peak of Sim Data)
Simulation Data
At the 122 keV peak, the simulation data is about 30% lower than the calibration data.
Comparison of Total Counts (above 20 keV)
- For the simulation and the calibration data, the total counts were found by
multiplying by the exposure time (1073 s) and adding up the counts in all bins
corresponding to energies greater than 20 keV.
- The total calibration data counts was divided by a factor of 1170 (see above).
- The predicted counts above 20 keV was calculated for each detector as:
I122 / (4πr2) *
f122
Cos(θ) A CZT122 +
I136 / (4πr2) *
f136
Cos(θ) A CZT136
where
I122 is the number of 122 keV photons/s emitted by the source
in 4π
f122 is the fraction of 122 keV photons that pass through any
lead tile that may be between the source and the detector.
CZT122 is the fraction of 122 keV photons that are absorbed in
a CZT detector.
Each value was then multiplied by the appropriate mask-weighting factor and all values
were added together (excluding those detectors that were excluded during the processing
of the simulation and calibration data).
The result was then divided by the sum of the squares of the
mask-weighting factors. These are the same steps that sum_spectra follows.
Results:
- Total Simulation Counts (above 20 keV):
- 14.2 counts
- Total Calibration Counts (above 20 keV):
- 18.7 counts
- Total Predicted Counts (above 20 keV):
- 15.5 counts
The total counts in the simulation is 24% lower than the total counts
in the calibration data. This is consistent with what is seen in the plot.
The total predicted counts falls in between the simulation result and the
calibration result, being 18% lower than the total
calibration counts and is 8% higher than the total simulation counts.
Back-of-Envelope Calculation:
The source is located at x=-7.769216, y=4.365952, and z=294.4 (all in cm).
A fully illuminated detector directly below the source would be at r=294.4 cm.
The absorption efficiency in CZT of 122 keV photons is 0.65, and the absorption
efficiency in CZT of 136 keV photons is 0.55.
The cosine correction factor is 1.
Using a source photon rate of 1.41 x 105 122 keV photons/s and
1.79 x 104 136 keV photons/s (where were the rates on 8/18/03),
the predicted counts is:
[(1.41 x 105)*/(4*π*294.4*294.4)*(0.16)*(0.65)+
(1.79 x 104)*/(4*π*294.4*294.4)*(0.16)*(0.55)]
*1073
=16.0 counts.
This agrees very well (within 3%) with the predicted total counts above.
Comparison of Histograms:
Simulation Data
Calibration Data
Predicted Data
The peak positions in these background-subtracted histograms match the
totals above quite nicely.
The peak of the simulation data histogram is not as well-defined, due to the
low statistics. The total counts found by sum_spectra is a little
higher than the position of the peak, but this might be simply because of the
low statistics.
Conclusions:
The difference between the simulation and the calibration data is large, but it
is interesting that the difference between simulation and prediction
is not so large--only 8%. Any materials included in the simulation
(other than the lead tiles) that lie
between the source and the detectors
would attenuate the beam and decrease the total number of counts. Could
8% of the photons be removed by scattering? If so, that could explain the
difference. There's still the concern over why the calibration data gives
a much larger value, though. That, I don't have an explanation for yet.
Return to main BAT response page