A Two-Stage Capacitive-Feedback Differencing Amplifier for Temporal Contrast IR Sensors

Abstract

This paper presents a small-area, ultra low-power, low-mismatch differencing amplifier for use in transient pixels of micro-bolometer based temporal contrast IR sensors. The two-stage capacitive-feedback amplifier works in the sub-threshold domain, has a voltage gain of 46dB, a 3dB bandwidth of about 10kHz and consumes 85nW of static power. The amplifier circuit has been fabricated in a 0.35µm standard CMOS process and consumes less than 2000µm 2 of silicon area, enabling a square pixel size of 50µm×50µm.

Event-Based Pixel Sensitive to Changes of Color and Brightness

“…The problems with charge injection in our cDVS pixel arise mostly from the flawed reset circuit. Although Posch et al have demonstrated that a two-stage architecture can provide high gain [28], we concluded for our circuit that two-stage switched capacitor amplifiers are not suitable, and therefore we have proposed circuit improvements which only use a single switched-capacitor amplifier stage.…”

“…SPICE simulations of the pixel before (and after) fabrication do not show this large injection effect, because in the simulations, the charge injected onto the input of from and from opening compensate. An easy fix to this problem is to split the generation of the two reset signals, as described in Section V and as used in the sDVS [34], or to use a circuit similar to the one used by Posch et al [28].…”

“…This change is easily made by duplicating the reset-and-slew circuit with separate biases for controlling the refractory period of and or using the scheme proposed by Posch et al [28]. However, we now believe that a two-stage amplifier is probably not necessary.…”

“…The voltage on nReset1 is higher than the voltage on nReset2, which ensures that is held longer in reset than so that charge injection from opening can settle. Posch et al [28] generate nReset1 with two inverters in series from nReset2. Here we use a different implementation with the intention to save area (two transistors versus four transistors) and avoid the large power consumption caused by the slowly switching inverters.…”

“…Second, they largely compensate the factor , so that is nearly independent of the difference between and . The summing amplifier is implemented as two consecutive capacitive-feedback inverting amplifiers and [28], where has two summing input capacitances and . in is used to limit the bandwidth of the pixel for noise filtering and to balance possible bandwidth differences in the two photoreceptors.…”