Research On Feature Extraction And Low Power Consumption Of TDI Image Sensor

CMOS image sensors are widely used in the fields of digital photography,medical imaging and spatial remote sensing because of its advantages of low power,low cost and easy integration.In order to obtain good quality images in low-light environment,researchers introduce time delay integration(TDI)technique in CMOS image sensor to extend the exposure time to enhance the signal-to-noise ratio of output signal.TDI technique requires different pixels to expose the same physical scene,and the exposure signal of the same physical scene must be accumulated before reading out.With the development trend of pixel array,when designing the CMOS-TDI image sensing system,we must consider the increasingly severe power consumption,especially the power consumption of analog-to-digital converter(ADC).In this paper,the power consumption of the ADC is considered from quantization and input.At the quantization level,the SAR-ADC architecture with low-power consumption potential is selected and a low-power switching sequence suitable for CMOS-TDI system is designed.At the input level,we use hardware to implement feature extraction as a way to eliminate a large number of irrelevant image data,reduce the burden of the SAR-ADC,and reduce the power consumption of the SAR-ADC.The main contents of this paper are as follows:First,at the quantization level,two typical SAR-ADC low-power switching algorithms in CMOS image sensing systems are analyzed and compared.Combining the working characteristics of CMOS-TDI system,based on the idea of data prediction readout,a switching algorithm with a lower power consumption potential is proposed,and the theoretical power consumption analysis of the switching sequence is completed.Second,at the input level,we use feature extraction to achieve the distinction between image key features and unrelated pixel data,reduce the amount of data sent by the front-end readout circuit to the ADC,and reduce the energy consumption of the ADC.In this paper,a feature extraction algorithm for CMOS-TDI system is designed,and the feature extraction algorithm is simulated and analyzed.The simulation results show that the rejection ratio of unrelated pixel data is more than 80%,which means that 80% of the ADC's quantization tasks are reduced.Third,in order to ensure the accurate implementation of the feature extraction algorithm for low-power consumption,an analog accumulator with offset error correction method is designed.By adjusting the internal structure of the op-amp,the sign of the buffer offset voltage is reversed.The correction of the offset error is implemented by the method of accumulating the error caused by the offset voltages during two accumulation stages.The theoretical analysis and simulation verification of the accumulator are carried out.The simulation results show that the analog accumulator has correct circuit function and the accumulator precision reaches 99.2%.In the case of process mismatch,the correction circuit can effectively correct the offset error and ensure the accumulator's accumulative accuracy is above 99.1%.