Investigation of Avalanche Gain in the APS
Measured output of a pixel in the typical APS, Monolithic CMOS detector shows a fixed ramp due to leakage current. Eventually the ramp gets steeper after the voltage drops sufficiently on the pixel diode. This rapid fall has been reported in published reports as due to avalanche charge generation at the drain of the source follower. Stuart Kleinfelder pointed out that one might be able to exploit the avalanche behavior to amplify the signal. This is a report on measurements investigating this issue.
The measurements were done on APS3 – sector 1, the standard pixel configuration.
The chip was clocked to connect a specified pixel to the output buffer and the output was observed on an oscilloscope. An LED was used to inject short light pulses (300 ns) at 3 different points in time during the cycle. Clocking of the chip and the pulsing of the LED was done directly with a single FPGA. The light injection points are recognized as steps in the output (see figure 1). The underlying signal without the injection steps is a long leakage current ramp that after ~ 150 ms from reset starts to increase in slope. The increase in slope is interpreted as the onset of the avalanche condition. The timing of the 3 identical LED pulses has been chosen to have two on the straight ramp with the third falling on the steep slope of the avalanche fall off.
See Fig. 2 See Fig. 4 See Fig. 3
Fig. 1 Output voltage of a single pixel in APS3 sector 1. The 3 step points are where an LED was flashed, injecting signal into the silicon. The slow ramp is due to leakage current. The steeper slope was believed to be from avalanching in the source follower. For magnified views of the steps see Figures 2-4
It was expected that the third LED pulse, the one that occurs during the avalanche phase would have a larger step because of the additional gain in a proportional avalanche. The results, however, contradict this expectation. Magnified views of the three light injection steps are shown in figures 2-4. The first two steps, the ones during the constant leakage current ramp, have (as expected) identical amplitudes, 24.2 mV. The third step, however, is only 22.0 mV, 9% smaller than the first two. These results are not understood. It has been suggested that the reduction from expected gain is due to reduced voltage on the diode with it’s increased diode capacitance. An examination of this possibility does not explain our observation. The gain change from slopes is 11.4 while the gain reduction from voltage change is between 1 and 0.8.
Fig. 2 Magnified view of the first LED injection step.
Fig. 3 Magnified view of 2nd LED step.
Fig. 4 Magnified view of the third LED step