1/f Noise Modelling and Characterization for CMOS Quanta Image Sensors

Abstract

This work fits the measured in-pixel source-follower noise in a CMOS Quanta Image Sensor (QIS) prototype chip using physics-based 1/f noise models, rather than the widely-used fitting model for analog designers. This paper discusses the different origins of 1/f noise in QIS devices and includes correlated double sampling (CDS). The modelling results based on the Hooge mobility fluctuation, which uses one adjustable parameter, match the experimental measurements, including the variation in noise from room temperature to -70 • C. This work provides useful information for the implementation of QIS in scientific applications and suggests that even lower read noise is attainable by further cooling and may be applicable to other CMOS analog circuits and CMOS image sensors. fluctuation as the origin of 1/f noise. Different from the mobility fluctuation in Hooge's model, which is a bulk effect, this model considers the mobility fluctuation induced by scattering from the charge near the Si-SiO 2 interface.QIS is designed to be sensitive to single photoelectrons, due to the very low sense node capacitance. When single-trap or multi-traps-induced RTN is present in a QIS device, it can be easily observed even at room temperature, as will be shown later. However, there exists a lower-level of background noise in addition to the RTN. With higher sensitivity and smaller SF gate dimensions, the need for a mobility fluctuation model to explain the measured noise may be stronger than for other image sensor SFs, or perhaps there is more to SF noise than just charge trapping and mobility fluctuation in the SF transistor. This paper discusses different 1/f noise models and fits the measured SF 1/f noise from a QIS prototype chip using these models instead of the widely-used fitting model [10], which only shows the 1/f and area-scaling trend and includes all other effects into a process-dependent coefficient. We find that the mobility fluctuation-based model fits the best. However, as is discussed in the conclusion of the paper, there are a number of reasons why the physical interpretation associated with this model seems inappropriate for our device, and the fit may only mean that the mathematical form fits and other physical interpretations or explanations are needed.

Deep Sub-Electron Read Noise in Image Sensors Using a Multigate-Source-Follower

“…The well-known models, such as the McWhorter number fluctuation model [4], Hooge mobility fluctuation model [5], or the Berkeley unified 1/ f noise model [6], have been developed to model the 1/ f noise, although none of them is universally accepted. Experimental data previously obtained from a QIS prototype chip with small SF show that the mobility fluctuation model matches the best in that case [7].…”

“…A conventional MOSFET SF with the same total gate area as V1 and V2 is included for baseline reference (V0 SF configuration). Since the buried-channel SF generally has lower 1/ f noise [7], all MGSF configurations and the V0 SF are buried-channel devices.…”

“…Therefore, to achieve a higher CG, a smaller gate area is desired. Due to this tradeoff between 1/ f noise and CG, the input-referred read noise can only achieve a theoretically limited minimum level when the SF is sized at its optimum [7].…”

Modeling the photon counting and photoelectron counting characteristics of quanta image sensors

“…The bit error rate (BER) as a function of the exposure level and read noise for QISs was analyzed in Ref. [10], and the recommended input-referred read noise was lower than 0.15 e-r.m.s.. A modified 1/f noise model of the source follower in jots, which was fitted to the measured data, was established in Ref. [11].…”

“…On the one hand, for high dynamic range (HDR) applications, the overexposure latitude range of the D-logH response curve can be used to extend the DR of the QIS. On the other hand, for applications that need high accuracy photon and photoelectron counting, the non-linear response of overexposure latitude range will increase the counting error of the QIS in high light level, and the non-ideal characteristics, such as dark cur-rent and read noise, will increase the counting error of the QIS [10] in low light level.…”