Cluster Charge Analysis

One type of analysis that does not require information from other subsystems is a study of the cluster charge distributions measured by the PXD sensors and recorded with the local PXD DAQ.

Hot pixel masking: Before the raw data of a run is clustered, a hot pixel masking is performed. The definition of a hot pixel is not precise and varies from analysis to analysis. In general, a pixel is considered hot if it gives a hit signal more frequently than the majority of all other pixels. In case of the PXD sensors this can be due to an electric damage of a pixel or a wrongly uploaded pedestal value. In the analysis presented here pixels were marked as hot in the following way:

1. The occupancy defined as the number of hits divided by the number of data frames (events) for all pixels is calculated.

2. The median of this occupancy distribution is calculated.

3. All pixels that have an occupancy larger than ten times the median occupancy are marked as hot and masked for further analysis.

This calculation is done for each sensor individually.

Clustering After the calculation of the hot pixel mask the hits of neighbouring pixels (including diagonal neighbours) are grouped into clusters. For each cluster the seed pixel is defined as the pixel with the highest hit value of all pixels belonging to the cluster. The charge of the cluster is the sum of the charges of the individual

hits. For the clustering a minimum seed charge of 7 ADU is required. This cut is applied to further remove noise hits from the data sample. All further analyses are based on the created cluster database.

A typical cluster distribution of a run during phase 2 can be seen in fig.6.9. As expected a Landau shaped distribution is visible due to the charged particles traversing the PXD.

In addition to this, a second Gaussian shaped component centred around 20ADU can be seen. When looking at multi-pixel clusters the ratio of the these two components is changed significantly as can be seen in fig. 6.9b. Due to its shape and because of the change in the ratio when comparing multi and single-pixel clusters, this component was identified to be caused by photons. To extract the MPV of the Landau component, a Landau function was fitted to the cluster charge distribution for each run and each PXD sensor. As can be seen in fig. 6.9, these fits describe the rising part of the distribution and the region around the MPV well, but underestimate the number of clusters with larger charges.

This is caused by two effects. First of all the dynamic range of the DCD has to be taken into account when comparing the distributions. Depending on the pedestal value pi of the pixel that gives a signals, the maximum signal is given ass_max= 256−pi. The second and much more important effect is the angle of the incoming particle. When a particle goes through the PXD under a large angle (compared to perpendicular transi-tion), it deposits more energy on average. As the particles in Belle 2 hit the PXD under a wide range of angles, the cluster charge distribution cannot be described accurately by a single Landau function.

However, the fit and the extracted MPV can be used to evaluate the evolution of the system gain, which changed due to the shifted threshold voltageV_thr. If the gate voltage V_Gateis not adjusted, the drain currentIDdecreases and by that the overall amplification g_q:

(V_Gate−V_thr)∝p

I_D ∝g_q∝MPV (6.1)

Figure 6.10shows the evolution of the MPV for all four modules as a function of time.

The figure shows that the MPV slowly decreased over time with two pronounced jumps upwards. These jumps occurred whenever the threshold voltage was readjusted and the working point was set back to its original value.

While the threshold shift of the sensors was not measured directly by an I-V curve mea-surement, the shift can be inferred from the readjustment of the gate voltages. When using the dose measured by the diamond system, the threshold shift can be compared with previous irradiation measurements. This comparison can be seen in fig.6.11. The comparison shows that the measured threshold shift of the phase 2 sensors is signifi-cantly larger than in the previous measurement with a prototype sensor. Even though previous measurements only used DEPFET prototype sensors, the discrepancy seen here

6.4. Data Taking Experience and Analysis

(a) All clusters

(b) Only multi-pixel clusters

Figure 6.9.: Cluster charge distribution for all four phase 2 sensors. In fig. 6.9a the cluster charge of all clusters is shown and in fig.6.9b only those with a size larger than one. For the outer-backward module (red), the overall entry count is lower because no data was recorded for the third DHP-DCD pair (25% less data)

.

is not expected. The most plausible explanation is that the diamond system was not sensitive to some part of the irradiation that affected the PXD and therefore the total

Figure 6.10.: Cluster charge MPV as a function of time for all four PXD sensors:

W41_IF (top left), W46_IB (top right), W37_OF1 (bottom left), and W37_OB1(bottom right). Changes of the applied gate voltages and other changes of sensor conditions are marked by vertical lines.

Figure 6.11.: Threshold shift of the phase 2 sensors compared with a previous PXD irra-diation measurement. The dose for the phase 2 data points was measured by the diamond system.

dose is underestimated. A naive scaling of the dose of the inner PXD half-ladders by a factor of 100 and 50 for the outer ones aligns the curves reasonable well to the reference

6.4. Data Taking Experience and Analysis

measurement.