Abstract
This work explores the benefits of linear-mode avalanche photodiodes (APDs) in high-speed CMOS imaging as compared to different approaches present in literature. Analysis of APDs biased below their breakdown voltage employed in single-photon counting mode is also discussed, showing a potentially interesting alternative to existing Geiger-mode APDs. An overview of the recently presented gated pinned avalanche photodiode pixel concept is provided, as well as the first experimental results on a 8 × 16 pixel test array. Full feasibility of the proposed pixel concept is not demonstrated; however, informative data is obtained from the sensor operating under −32 V substrate bias and clearly exhibiting wavelength-dependent gain in frontside illumination. The readout of the chip designed in standard 130 nm CMOS technology shows no dependence on the high-voltage bias. Readout noise level of 15 rms, full well capacity of 8000, and the conversion gain of 75 µV are extracted from the photon-transfer measurements. The gain characteristics of the avalanche junction are characterized on separate test diodes showing a multiplication factor of 1.6 for red light in frontside illumination.
Highlights
An image sensor that is suitable for low-light imaging on one hand needs to have good optical properties to collect as many incident photons as possible, and, on the other hand, requires low readout noise levels to bring the signal-to-noise ratio (SNR) close to the Poissonian limit
It can be seen that an avalanche photodiodes (APDs) with k = 0.3, biased below breakdown having a readout noise level of σR = 5 e− can achieve a photon detection probability (PDP) as high as 35%, which is comparable to the performance of the state-of-the art CMOS single-photon APDs (SPADs) [17]
A gated pinned APD (PAPD) pixel concept was proposed to explore the possibility of combining standard CMOS image sensors (CIS) technology with avalanche signal multiplication, and at the same time providing good optical and speed properties due to full-depletion of the epitaxial layer [14,31]
Summary
An image sensor that is suitable for low-light imaging on one hand needs to have good optical properties to collect as many incident photons as possible, and, on the other hand, requires low readout noise levels to bring the signal-to-noise ratio (SNR) close to the Poissonian limit. To compare the low-light performance of different image sensors, one can define a figure of merit (FOM) as the minimum incident photon count I that is required to achieve signal-to-noise ratio (SNR) equal to one. This approach is similar to luminance-SNR of 10 (YSNR10) FOM first proposed for mobile-phone cameras [5]. Taking into account the following parameters: quantum efficiency QE, fill factor f f , multiplication factor M, excess noise factor F, readout noise σR in electrons rms, and the number of dark electrons per second D, where the factor a describes the effective fraction of amplified dark current carriers In this model, the FOM depends on the frame rate only through the dark current.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
