708
8.1. Gain Calibration
709
The low level charge calibration converts raw ADC response of the electronics to photoelec-
710
tron units. It is performed in three steps: pedestal subtraction, correction for the electronics
711
non-linearity and the relative low/high gain response, and correction for the MPPC gain varia-
712
tions.
713
The pedestal, i.e. the baseline response of the MPPC and electronics without input signal,
714
as well as the MPPC gain, is measured using non zero-suppressed dark rate noise (Figure 17).
715
The pedestal peak in the dark noise spectrum is fit to a Gaussian function in each integration
716
cycle separately to account for the small variations among the cycles. The mean of the Gaussian
717
33
gives the pedestal constant used for the pedestal subtraction. The MPPC gain is measured as the
718
separation between the pedestal and the 1 p.e. peak after combining the dark noise spectra from
719
all 23 integration cycles which were corrected for individual pedestal shifts. The two peaks are
720
fit to a double Gaussian and the difference in their means is used to measure the photoelectron
721
unit in ADC values. Since the MPPC overvoltage, as well as the pedestal, is quite sensitive to
722
the temperature at a fixed bias voltage, the gain and pedestal require continuous monitoring and
723
updating.
724
ADC count
1500 160 170 180 190 200
1000 2000 3000 4000 5000
Figure 17: Typical digitized dark noise spectrum of an MPPC with a double Gaussian function fitted to the pedestal and 1 p.e. peaks.
Before converting the signal into photoelectron units, the raw ADC response needs to be
725
corrected for the non-linearity of the electronics and the relative gain difference between the
726
high and low gain response. The response of each input channel is measured using the internal
727
TFB calibration circuit as a series of 174 signal levels that covers both the high and low gain
728
dynamic range. This measurement is performed in situ when the MPPCs are powered since the
729
capacitance of the photosensors and the mini-coax cables connecting the sensors to TFBs repre-
730
sent a significant additional capacitance on the input, altering the effective electronics gain. The
731
measured response as a function of the calibration level is fit to a bi-cubic polynomial with nine
732
free parameters. This parametrization is used to correct the raw ADC values during the offline
733
calibration of the data. The bi-cubic function allows an adequate representation of the Trip-t non-
734
linearity with residuals typically smaller than a few percent (Figure 18). The electronics gain and
735
non-linearity are fairly stable, requiring only occasional checks, and therefore the constants are
736
updated only if there is a hardware change to the front-end electronics.
737
34
High Gain ADC
0 100 200 300 400 500 600 700 800
Input Charge (arbitrary units)
0 50 100 150 200 250 300 350 400 450
Figure 18: Charge versus ADC for a high-gain Trip-t channel fit to a bi-cubic function.
8.2. MIP Light Yield
738
A useful characteristic of minimum ionizing particles (MIPs) is that their energy loss is only
739
weakly dependent on their energy. Therefore, for high energy muons passing through the detec-
740
tor, the mean energy deposition per unit length is a constant. With a sample of through-going
741
muons, this constant can be determined. The first sample used was the cosmic checkout data,
742
selecting tracks with cosθ >0.8 [16] (whereθis the angle that the track makes with the z axis of
743
the detector) but this was for individual SuperPØDules only.
744
After the PØD was installed in the basket, it became possible to calibrate all PØDules with the
745
same data sample. The best sample was through-going muons from beam neutrino interactions
746
in the wall or sand and rock upstream of the PØD. After reconstruction, events were selected
747
with a single 3D track entering the front face of the PØD and exiting out the downstream end.
748
These events were analyzed to show each layer’s detection efficiency. Due to the triangular
749
design of the PØD’s scintillator bars, a normally incident MIP is most likely to pass through two
750
bars, as demonstrated in Fig. 19. However, depending on the path taken, there is a chance that one
751
bar is untouched, or that the signal is below the noise threshold cut applied by the reconstruction.
752
The results, shown in Fig. 20, indicate the probability of finding 0, 1 or 2 hits in each x or
753
y plane. The tracking efficiency is 100% for all but the first three scintillator planes, which is
754
explained by the selection criteria allowing a small number of first layer neutrino interactions
755
into the sample.
756
Figure 21 shows the summed charge deposit for the two hit sample, after calibration and
757
path correction. The plot has been fit with a Gaussian-Landau distribution, and returns a most
758
probable value of 37.9 p.e./mip/cm. This value provides a known point, which each channel of
759
35
Figure 19: Illustration of a singlet and doublet, as a MIP passes through a PØDule layer.
the PØD can be calibrated to, ensuring a constant response for the detector.
760
8.3. LIS Operation and Performance
761
The LIS system simultaneously illuminates the entire PØD and is read out at in bursts of 20
762
Hz interspersed with other trigger types. The current settings give the LIS system an effective
763
rate of 1.5Hz. The LIS system cycles through a set of ten amplitudes, each with 500 flashes,
764
taking about one hour for a complete cycle. Figure 22 shows the average ADC signal produced
765
by each of the four pulser boxes during a typical run. Each plateau corresponds to a single
766
amplitude. The sequence of amplitude was purposefully chosen to produce a clear step structure
767
in the response to enable easy visual separation of the groups from each other.
768
Besides providing the ability to quickly determine the correct functioning of all PØD pho-
769
tosensors, the LIS provides a tool to monitor the stability of the photosensor signal, shown for
770
a portion of a physics run is shown in Fig. 23. The variation over short periods of time can be
771
attributed to changes in the photosensor gain. Shifts that are different with respect to each pulser
772
can be evidence for malfunctions in the PØD readout electronics.
773
8.4. Water Target Filling and Monitoring
774
The depth sensors were found to have fluctuations of±1 mm but had a±15 mm calibration
775
offset before insertion into the PØD. This offset was reduced by using the fixed binary wet-dry
776
level sensors to provide calibration reference points in situ. We expected the water level to drop
777
36
X Layers
0 5 10 15 20 25 30 35 40
Proportion (%)
0 10 20 30 40 50 60 70 80
Y Layers
0 5 10 15 20 25 30 35 40
Proportion (%)
0 10 20 30 40 50 60 70 80
Figure 20: The PØD layer detection efficiency. The plots show the proportion of tracks with 0 (green circles), 1 (blue triangles) and 2 or more (red inverted triangles) hits in a layer, for both x and y. The small excess of 0 hits in the upstream x layer is due to neutrino interactions in the first y layer passing the cuts.
in some layers due to deflection of the plastic scintillator. As shown in Fig. 24, the largest change
778
in water level is closest to the downstream end of the PØD which is not directly supported by the
779
basket.
780
Geometry and the measured dimensions of the PØD constrain the uncertainty on the total
781
mass of water in the fiducial volume to approximately 3%. The addition of measurements from
782
the WL400 depth sensors and the external tank volume measurements reduce this uncertainty to
783
less than 1%.
784