Threshold Shifts

To measure the threshold shift of the external DEPFET gate, I-V curves were recorded by varying the Gate-On voltage VG and recording full frame memory dumps. While this method is not as precise as a measurement of the drain-source currentID with an external device, it has the advantage that IV-curves of all 192000 DEPFET cells are recorded simultaneously.

As the dynamic range of the DCD is limited to 256 ADU, the measurement was re-peated with different VnSubIn settings on the DCD, which steers the current that is subtracted from the drain-source current inside the DCD (see also section 4.5.3). This effectively ”shifts” the I-V curve and allows to record a different part of it. The individ-ual curves (of the same pixel) are later put back together to form a complete I-V curve.

Using this approach it is also possible to restrict the analysis to the central part of the dynamic range of the DCD in which the linearity of the ADC is expected to be best [95].

Figure 7.10 shows an example of this I-V curve ”stitching” for one pixel. Instead of

Figure 7.10.: Example I-V curve of one pixel. The different colours in the top figure indicate the VnSubIn setting that was used during the recording of the curve. The bottom figure shows the merged I-V curve.

extracting the absolute threshold voltage from the recorded curves, only the relative shift was calculated by comparing the curves to a reference measurement that was done before the irradiation. For this comparison, which is done for each pixel individually, the curves are binned along theID axis and the mean gate voltage for each bin is calculated.

The binning is chosen such that most bins only contain one or two data points. The mean voltages of each bin are then compared to the corresponding values of the reference measurement and the difference is calculated. The threshold voltage shift is defined as the mean over all bins.

7.5. Threshold Shifts

V_G ID

Figure 7.11.: The threshold shift of each irradiation step is calculated by comparing the I-V curves before (red) and after (black) the irradiation step. The shift is measured for several bins along theID axis. Then the average is calculated.

Care has to be taken here not to compare the ”flat” part of the I-V curves. A visuali-sation of this technique on hypothetical I-V curves is shown in fig. 7.11. In fig.7.12the evolution of the I-V curves during the irradiation for one pixel is shown.

As mentioned in section 7.3, the gradient of the X-ray beamspot has to be taken into account when the dose for pixel is calculated. A major consequence of this gradient is that one I-V curve measurement is not associated to a single dose value but rather a dose range. This also means that, if the step size is chosen accordingly, two measurements can partly overlap in terms of received dose and a quasi continuous dose distribution is created when the individual measurements are combined. If the step is too large, a gap in the dose continuum occurs.

Figure7.13shows a boxplot of the threshold shifts extracted from the I-V curves. The data has been binned along the dose axis and all pixels from all measurements were used to fill the bins. Many bins contain therefore data from different measurements.

The boxplot shows the median as well as the 5%-95% interval of the ensemble of threshold shift values for each bin. Nearly all of the dose bins contain at least one data point except for two bins around50kGy. Here the irradiation step size was too large. A separate inset within the figure shows the low dose range with a finer binning. Here it can be seen that the slope of the threshold shift curves changes. Until≈2kGy the curve is very steep and then seems to hit a plateau. Around 4 kGy the slope increases again and the curve then follows the expected shape. The maximum threshold shift measured at the end of the irradiation is ≈10.5V at a dose of≈266kGy.

Not only the external DEPFET gate is affected by the radiation, but also the common clear gate (CCG). Recording an I-V curve is not possible for this gate, because of the PXDs design. It is, however, possible to measure the threshold shift indirectly. When the clear gate voltage is lowered to a certain point, a secondary channel between drain and source of the DEPFET gate is opened. This can be seen directly as an increase

Figure 7.12.: I-V Curves of a single DEPFET pixel at different irradiation steps.

Reprinted with permission from [87].

of the source current I˜_D measured by the LMU power supply. Additionally, a small current starts to flow over the drift contact, where under normal operation conditions no current flow is registered by the LMU power suppy². The shift of the threshold voltage is directly correlated to this point which is why it is used to determine the shift.

During the measurement, the CCG voltage is lowered in small steps and the drift current is recorded. The reference point is defined by a current of 2 mA flowing over this line.

Similar to the threshold shift calculation of the DEPFET gate, the voltage of this point is compared to a measurement done before the irradiation. The evolution of the shift can be seen in fig.7.14.

The dose in this plot corresponds to the dose at the central part of the sensor’s matrix.

A correction for the beam spot is not possible as only a common shift can be determined using the method described above. The shape of the curve is very similar to the one of the DEPFET gate. The major difference is the absolute scale. At the end of the campaign the shift was ≈ 6V. This is significantly lower (almost 50%) than the shift seen in the DEPFET gates. As the CCG has a different design and a different oxide than the DEPFET gate, this is to be expected.

Not only the threshold shifts, but also the effects of annealing were studied. Before the DUT was switched off, it remained powered for an additional period of 10 days after the last irradiation step. During this time additional measurements of the DEPFET threshold shift were done and one measurement for the CCG shift. As can be seen in fig. 7.14, the CCG shift had decreased by ≈ 0.7V. The DEPFET threshold shift was

2The resolution of this channel is only±1mA

7.5. Threshold Shifts

Figure 7.13.: Evolution of the threshold shift during the whole irradiation. Data of all pixels from all measurements is binned based on the TID of the pixel during a measurement. Bin width is 5 kGy for the main figure and 6 kGy for the inset except for the first 4 bins (0.5 kGy width). The x-axis shows the mean dose of each bin. Reprinted with permission from [87].

measured six times during this annealing phase. Figure 7.15 shows the evolution of the shift. The figure shows the evolution of three different ”regions” defined by their received dose. All pixels that belong to these regions are considered to form ensembles of shifts of which the median is calculated, similar to the process used to visualise the shift during the irradiation. As the figure shows, the trends of all three regions are similar. During the first 100 hours, the shift drops by about 1V and then shows only a very slight de-crease over the next 140 hours. After≈120hours, the temperature of the water chiller was increased from 5^◦C to 10^◦C too check whether the temperature increase affects the annealing. As one can see from the data, this was not the case and no acceleration of the annealing was seen.

Figure 7.14.: Threshold shift of the common clear gate during the irradiation. Reprinted with permission from [87].

Figure 7.15.: Evolution of the DEPFET threshold voltage shift during the ten day an-nealing phase at the end of the irradiation campaign. Three different TID regimes are shown. All pixels within the respective regime are taken into account for the calculation of the median and spread. Reprinted with per-mission from [87].