Application of discrete-wavelet transform compression in wafer fabrication

This paper reports the application of the discrete wavelet transform (DWT) compression and its performance in wafer fabrication process control data. Both single-channel time series and multi-channel image data were examined. We found that with a proper choice of the mother wavelet, decomposition level, and thresholding values, DWT compression can effectively compress data while maintaining satisfactory signal authenticity to meet the needs for process control in wafer fab equipment. In particular, the Daubechies-3 (db3) and Haar wavelets are appropriate mother wavelets for data compression in single channel time series measurements. Our compression analysis results have shown satisfactory signal authenticity and storage space savings as high as 95%. Fabrication data obtained from optical emission spectrum that contains thousands of recording channels were treated as images and compressed accordingly using the Set Partitioning in Hierarchical Trees (SPIHT) algorithm. Satisfactory results were also achieved with more than 99% of data storage savings.