Pipelined Image Sensor Readout for Speed and Quality
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image sensors face challenges in achieving faster processing speeds and better image quality due to insufficient time for pixel operations and excessive noise from incomplete correlated double sampling, particularly in high-speed applications.
Innovation Solution
The implementation of a pipelined readout method that extends the sig1 and sig2 readout phases to last for one row time combined, allowing for extended pixel transfer time, and removing column ADC circuit autozeroing from row operations to allocate more time for pixel operations and analog-to-digital conversion, with autozeroing performed only once per frame.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional row-by-row readout is used, then image quality is maintained, but processing speed is limited due to insufficient time for pixel operations
Solution Approach 1:
The readout process is segmented into multiple phases: first signal readout (sig1), second signal readout (sig2), and analog-to-digital conversion. This segmentation allows pixel operations to be completed within one row time while enabling overlapping operations across different rows, thereby increasing processing speed without sacrificing image quality.
Solution Approach 2:
The first signal (sig1) is readout and stored in column storage before the second signal (sig2) is readout. This preliminary action of storing sig1 allows the ADC conversion to start earlier and complete within the same row time, enabling faster processing while maintaining signal integrity for image quality.
2Measurement precision
If correlated double sampling is performed, then image quality improves, but noise increases due to incomplete sampling at high speeds
Solution Approach 1:
The pipelined readout method ensures continuous and complete correlated double sampling by reading both sig1 and sig2 signals fully within one row time. The column storage holds sig1 while sig2 is readout, and ADC conversion continuously processes the difference signal. This continuous action completes the sampling process without interruption, maintaining image quality while reducing noise through proper sampling completion.
3Productivity
If ADC conversion time is extended, then processing speed improves, but row readout time is insufficient
Solution Approach 1:
The patent utilizes the time dimension by overlapping row operations across multiple rows. While one row is being readout, ADC conversion for previous rows is simultaneously occurring. This dimensional approach allows extended ADC conversion time without increasing the apparent row readout time, as operations are distributed across time and row dimensions simultaneously.
Solution Approach 2:
ADC conversion for a given row starts before the row readout is completely finished by using column storage to hold the first signal (sig1) from previous rows. This preliminary preparation of ADC conversion allows the conversion process to extend beyond the traditional row time constraint while maintaining proper readout timing through the pipelined structure.
Data Source
AI summary
A pipelined readout method in an image sensor includes receiving one or more signals from a pixel of a row of a pixel array into a column storage at least partially during a time that a previously sampled amplified output of the column storage that is based on signals provided by a previous pixel of a previously read out row of the pixel array is converted from analog to digital by an analog-to-digital conversion circuit. The method further includes performing, by the analog-to-digital conversion circuit, analog-to-digital conversion of a sampled amplified output of the column storage that is based on the one or more signals from the pixel at least partially during a time that the column storage receives at least one signal from a another pixel of a subsequently read out row of the pixel array.