Analog VLSI Adaptive, Logarithmic, Wide–Dynamic-Range Photoreceptor

We describe a photoreceptor circuit that can be used in massively parallel analog VLSI silicon chips, in conjunction with other local circuits, to perform initial analog visual information processing. The receptor provides a continuoustime output that has low gain for static signals (including circuit mismatches), and high gain for transient signals that are centered around the adaptation point. The response is logarithmic, which makes the response to a fixed image contrast invariant to absolute light intensity. The 5-transistor receptor can be fabricated in an area of about 50 by 50 μ m2 in a 2μ m single-poly CMOS technology. It has a dynamic range of 1–2 decades at a single adaptation level, and a total dynamic range of more than 6 decades. Several technical improvements in the circuit yield an additional 1–2 decades dynamic range over previous designs without sacrificing signal quality. The lower limit of the dynamic range, defined arbitrarily as the illuminance at which the bandwidth of the receptor is 60 Hz, is at approximately 1 lux, which is the border between rod and cone vision and also the limit of current consumer video cameras. The receptor uses an adaptive element that is resistant to excess minority carrier diffusion. The continuous and logarithmic transduction process makes the bandwidth scale with intensity. As a result, the total A.C. RMS receptor noise is constant, independent of intensity. The spectral density of the noise is within a factor of two of pure photon shot noise and varies inversely with intensity. The connection between shot and thermal noise in a system governed by Boltzman statistics is beautifully illustrated. THE CIRCUIT A simple nonadaptive source-follower logarithmic receptor is shown on the left of Figure 1. The gain is V T =kT/q=25.4 mV per e-fold intensity change. The sourcefollower receptor is lacking in two respects: The offsets are as large as signals produced by typical scenes, and the response is too slow. The adaptive receptor remedies both problems. A lowpass-filtered output signal is fed back to the gate of the feedback transistor in the source follower receptor, shown on the right of Figure 1. A conceptual way of thinking about the operation is indicated by the shaded areas. The circuit uses an internal model to make a prediction about the input signal. The output comes from a comparison of the input and the prediction. The loop is completed by using learning to refine the model so that predictions become more accurate. The adaptive receptor, with its level adaptation, uses perhaps the simplest type of learning. The output of the source-follower receptor is amplified by the inverting amplifier consisting of Q n, Q cas and Q p . The voltage gain –A amp is typically several hundred. Bias voltage V b determines the cutoff frequency for the receptor, by setting the bias current in the inverting amplifier. Q cas is a cascode that nullifies the Miller capacitance from the gate to the drain of Q n and also doubles the gain of the amplifier. The feedback loop is completed when the output V o is fed back to V f through the resistor-like adaptive element and the capacitive divider formed from C 1 and C 2 . Receptor Gain The high gain in the feedback loop effectively clamps V p . When the input current changes an e-fold, V f must change by V T / κ to hold V p clamped, where is the back-gate coefficient. Hence, the small-signal steady-state gain is given by (steady-state closed-loop gain) For transient signals, where the output must go through the capacitive divider, the gain is (transient closed-loop gain) κ 0.8 ≈ vo VT ⁄ i Ibg ⁄ ----------------1 κ -= Acl vo VT ⁄ i Ibg ⁄ ----------------1 κ -C1 C2 + C2 -----------------= ≡ Vp Vo Vf