An Electronic Photoreceptor Sensitive to Small Changes in Intensity

Tobi Delbrück, Carver MeadNeurIPS 1988

Introduction

Standard electronic light sensors often struggle with a classic trade-off: they either provide high sensitivity over a narrow range of light levels or a wide dynamic range with poor sensitivity to small fluctuations. In nature, biological retinas solve this beautifully, detecting tiny changes in contrast whether in a dim room or under bright sunlight.

Authored by T. Delbrück and Carver Mead, this research describes a bio-inspired electronic photoreceptor that mimics the adaptive behavior of retinal bipolar cells. By utilizing a feedback loop with a hysteretic element, the circuit achieves high sensitivity to transient changes in light while maintaining a wide dynamic range, effectively replicating Weber’s Law seen in human vision. The goal was to create a sensor that ignores steady-state background illumination (the DC component) and amplifies rapid changes (transients).

Circuit Design and Methods

The researchers designed a CMOS circuit consisting of two primary stages of amplification to capture this biological "gain control."

Circuit diagram of the electronic photoreceptor
  1. Logarithmic Conversion: A phototransistor ($P$) converts incident light into a current. This is paired with an exponential feedback loop (transistors $Q_1, Q_2, Q_3$) that produces a voltage ($V_p$) proportional to the logarithm of the light intensity.
  2. Hysteretic Feedback: A transconductance amplifier ($A$) takes the output of the first stage. Its output passes through a "hysteretic element"—a pair of diode-connected transistors ($Q_4, Q_5$).

The key innovation lies in the feedback loop. For slow changes, the circuit has low gain. However, for fast changes, the hysteretic element acts as a form of memory, preventing the feedback from responding immediately. This allows the amplifier to operate at full capacity, providing high transient gain. Interestingly, the authors found that the Early effect (a phenomenon where transistor resistance varies with current) naturally provides a gain control mechanism that keeps the circuit's response proportional to "contrast" ($\delta I / I$) rather than absolute intensity.

Evaluation and Key Insights

The authors evaluated the circuit using direct electrical characterization of the chip's output in response to LED flashes, human psychophysical data (flicker-detection trials), and biological comparisons using historical data on retinal bipolar cells.

  • The Early Effect as Gain Control: One of the most elegant findings is that the parasitic resistances in the transistors are inversely proportional to the current. This mathematical relationship perfectly cancels out the absolute intensity, ensuring the voltage change is proportional to the percentage change in light.
  • Transient vs. DC Gain: The circuit provides a transient gain roughly 200 times higher than its DC gain. This allows it to be incredibly sensitive to movement or flickering while remaining stable under changing ambient light.
  • Weber’s Law in Silicon: The circuit naturally obeys Weber’s Law—the principle that the minimum detectable change in a stimulus is a constant proportion of the original stimulus, a fundamental characteristic of human perception.

Key Results

The circuit successfully demonstrated human-like performance. When compared to human flicker detection, the circuit’s thresholds and temporal contrast sensitivities were nearly identical.

Circuit response to small transient changes in light intensity

Specifically, the circuit achieved a peak sensitivity of 1.2% at an incident intensity of $660 \mu W/cm^2$ (roughly half the brightness of a typical office). Furthermore, the circuit's DC output remained valid over four orders of magnitude of light intensity. However, its high-gain transient "window" was narrower (about 0.1 log units) compared to biological bipolar cells, which typically span 1.5–2 log units.

Limitations and Future Work

While the circuit is a successful proof-of-concept, the authors noted several areas for improvement:

  • Asymmetry: Unlike biological cells, which respond symmetrically to light increases and decreases, this circuit responds much more strongly to increases.
  • Time Constants: In the circuit, the response time constant is inversely proportional to light intensity (it gets slower in the dark). In biological bipolar cells, the time constant remains remarkably stable across light levels.
  • Sensitivity Gap: The authors ultimately intended to use these sensors for neurophysiological optical dye recording, which requires detecting changes of 1 part in $10^3$ or $10^4$. The current circuit is still 1 to 2 orders of magnitude short of that goal.

To address these limitations, the authors suggest moving toward hybrid bipolar/FET technology to further improve sensitivity and overall performance.

Summary updated

BibTeX
@inproceedings{delbruck1988electronic,
  author = {Tobi Delbrück and Carver Mead},
  title = {{An Electronic Photoreceptor Sensitive to Small Changes in Intensity}},
  booktitle = {Advances in Neural Information Processing Systems},
  year = {1988},
}
Figure