Optimization of a CCD-in-CMOS TDI detector's operating clock voltages by Taguchi based Grey relational analysis

TL;DR

This paper presents an optimization method for CCD-in-CMOS TDI detector clock voltages using Taguchi's DOE and Grey relational analysis to enhance multiple performance parameters efficiently.

Contribution

It introduces a systematic experimental design approach combining Taguchi's method and Grey relational analysis for multi-parameter optimization of detector voltages.

Findings

Optimized clock voltages improved detector performance.

The method reduced experimental trials needed for optimization.

Enhanced multiple parameters simultaneously with a single optimal voltage set.

Abstract

In recent years, CCD-in-CMOS TDI image sensors are becoming increasingly popular for many small satellite missions to assure a fast and affordable access to space for Low Earth Observation. Our monolithic CCD-in-CMOS TDI imager features a specifically developed technology which combines the benefits of a classical CCD TDI with the advantages of CMOS System-On-a-Chip (SoC) design. Like CCD, this detector is also controlled by a large number of clock voltages. Optimizing these voltages allows to increase the performance of the detector by improving multiple characteristic parameters, such as full well capacity (FWC), dark current, linearity, dark signal non-uniformity (DSNU) and charge transfer efficiency (CTE). Traditionally, it has been the standard practice to adjust the CCD voltages by trial and error methods to get a better image. Because of the large parameter space, such subjective procedures may yield far from the optimum performance. This paper reports a design of experiments (DOE) technique applied on the clock voltages to improve the multiple performance parameters of the detector. This method utilizes the Taguchi's orthogonal arrays of experiments to reduce the number of experiments with different voltage combinations. Finally, optimal combination of clock voltages is obtained by converting the multiple performance parameters of the detector into a single Grey relational grade. In this process, the sequences of obtaining parameter values are categorized according to the performance characteristics. The condition Higher-the-better is used for parameters like FWC and CTE whereas condition Lower-the-better is applied for parameters, such as dark current, linearity error and DSNU.