HW6 - Contribution for Part III: Sensors - Pixel Design (Anna Yu)

[anbananna]

[1] Usefulness

In the medical field, imaging extremely small things is an important and crucial resource for disease diagnoses and monitoring. Focusing on pathology scanners, where image quality is a requirement for diagnostic reliability, it becomes apparent that pixel architecture directly affects how the image will come out.

Pathology scanners are used to digitize tissue sections at very high resolution, usually at 20 to 40 times magnification and in the order of magnitude of Gigapixels. Color accuracy, imaging speed, high dynamic range, low noise, and high uniformity are other key factors in a pathology scanner. Considering these properties, pixel design becomes the key factor in the system’s ability to satisfy these requirements. It is common to see a 4T CMOS pixel architecture which will secure:

1. Color accuracy
   A pinned photodiode (PPD) drastically reduces dark current which produces a cleaner base signal and aids with better charge transfer efficiency for more stable color responses across the pixel.

2. High dynamic range
   A 4T pixel can support a dual conversion gain gate and a high full well capacity that allows for pixel capacitance switches: low conversion gains for highlights and high conversion gains for shadows.

3. Low noise
   Allowing for correlated double sampling (CDS), a pixel with this design can sample the floating diffusion before and after charge transfer to eliminate fixed-pattern, reset, and kTC noise.

4. High uniformity
   The combination of effects of a PPD, transfer gate (TX), CDS, and a layout that minimizes pixel cross talk results in a high uniformity image.

In comparison, CCD’s are avoided because of their lower frame rate, higher noise at higher imaging speeds, and worse uniformity. These are all a direct result of its architecture requiring components like high voltage clocks, off chip electronics, and mechanical shutters. The architecture’s lack of ability to support and allow other components, like transfer gates, dual conversion gains, etc., also hinders its ability to check all the boxes a pathology scanner needs.

Understanding pixel design enables one to understand how to maximize a specific component’s strengths while acknowledging physical limitations to then build intricate and high precision devices like pathology scanners.

[2] Proposed Improvement: Written descriptions of key modern pixel components

Modern CMOS pixels use a 4T pixel architecture which consists of four key components:

1. Photodiode (PD)
   The photodiode is the primary light-sensing element of the pixel, formed by a p-n junction that converts incident photons into electrical charge. When light enters the pixel and is absorbed within the diffusion volume, electron-hole pairs are generated. These photo-generated electrons are collected in a potential well where they accumulate over the exposure interval. The total stored charge is thus proportional to the number of absorbed photons, making the photodiode the fundamental signal integration component of the pixel.

2. Transfer Gate (TX)
   The transfer gate is a MOSFET positioned between the photodiode and floating diffusion, providing controlled isolation between the charge collection region of the pixel and readout circuitry. During the integration period, the transfer gate remains off, maintaining a potential barrier that prevents charge from prematurely flowing out of the photodiode. When a control pulse is applied to the gate, the barrier is lowered, enabling charge transfer from the photodiode’s potential well to the floating diffusion. This controlled transfer of charge is central to low noise readout and supports correlated double sampling which relies on precisely timed transitions between reset and signal states.

3. Floating Diffusion (FD)
   The floating diffusion is a small, isolated capacitance that serves as the charge-to-voltage conversion site. When photoelectrons are transferred from the photodiode through the transfer gate, they accumulate on the floating diffusion, producing a voltage change inversely proportional to the node’s capacitance. Because the floating diffusion is only connected to high impedance transistor gates during readout, its potential is able to accurately reflect the transferred charge until sensed by downstream circuitry. High conversion gains achieved by minimizing floating diffusion capacitance are important for reducing read noise and lowlight sensitivities.

4. Source Follower Amplifier (SF)
   The source follower amplifier is a transistor whose gate is connected to the floating diffusion source, acting as a voltage buffer between the floating diffusion node and the column level readout circuitry. The source follower provides high input impedance to ensure negligible loading on the floating diffusion and preserving the accuracy of the voltage representing the number of photons captured. Its output, taken from the source terminal, drives the column bus through the row select transistor with a low output impedance, which is essential for maintaining signal integrity along relatively long and capacitive column lines.

In addition to short descriptions of these key components in a modern CMOS 4T pixel, the diagrams and pictures shown in class were very helpful in visualizing the flow of photons, electrons, and charge throughout the whole pixel. Visualizations on the lecture slides would complement the content in this section well (eg. 02a Image Capture Sensor, slide 51 and 53).