Hardware validated unified model of multibit temporally and spatially oversampled image sensors with conditional reset

Abstract

Abstract. We describe a photon statistics-based theoretical model of the response to incident light of an image sensor and show that conditional reset and multibit temporal oversampling increase the dynamic range significantly. This photon-based modeling approach describes the full image sensor design space of temporal and spatial oversampling either with a binary comparison or with a multibit read of each sample. We find excellent quantitative agreement between measurements on custom hardware and our theoretical predictions. We then use this model to show what improvements in dynamic range and low-light response can be achieved by oversampling and what the limits of improvement caused by pixel size and lens parameters are. © The Authors.Published by SPIE under a Creative Commons Attribution 3.0 Unported License. Distribution or reproduction of this work in whole or in part requires full attribution of the original publication, including its DOI.

Non-destructive threshold assessment of a pinned photo diode pixel with correlated double sampling

“…However, their approach requires pixels with single-photon sensitivity and sizes far below 1μm and is not feasible with current pixel technologies. We have shown [3] that sampling with a multiple electron binary threshold comparison followed by a conditional multibit read with an ADC if above threshold is equivalent to binary sampling. Our proposed method can be implemented using today's pixel technology.…”

“…The architecture is mostly identical to a conventional 4-transistor pixel with the exception of an additional transistor to provide column as well as row control of the transfer gate. The oversampling method of [3] requires that the photo diode of a pixel that is below the threshold is not read out or reset and continues to accumulate photo electrons. In a normal photo diode read, the charge is transferred from the photo diode to the floating diffusion and its value is read out through the source-follower and row select transistors.…”

“…This charge is sensed by the sense-amplifier, causing a per-column circuit outside the pixel array to apply an on-voltage to TGc, thereby transferring the remaining charge to the floating diffusion. The total transferred charge is then measured with an ADC and used in the multi-bit oversampled reconstruction described in [3]. The corresponding timing diagram is shown in Fig.…”

A 99.95% linearity readout circuit with 72 dB dynamic range for active pixel sensors

“…Due to the reduction in the supply voltages under CMOS technologies downscaling and the resulting linear decrease in the analog voltage swing, the DR enhancement becomes crucial for high‐performance CMOS image sensors in deep submicron technologies. Accordingly, the high DR CMOS imaging has become a leading research topic in the area . The increase of dynamic range shall be achieved while preserving the linearity.…”

Images from Bits: Non-Iterative Image Reconstruction for Quanta Image Sensors

“…In the past few years, research groups at the University of Edinburgh (Edinburgh, UK) [14,15,16], as well as EPFL [17,18] have made new progresses in QIS using binary single-photon avalanche diode (SPAD) cameras. In the industry, Rambus Inc. (Sunnyvale, CA, USA) is developing binary image sensors for high dynamic range imaging [19,20,21]. For the purpose of this paper, we shall not differentiate these sensors, but refer to them generally as the QIS, because their underlying mathematical principles are similar.…”