Future lithography systems must produce microchips with smaller feature sizes, while maintaining throughputs comparable to those of today's optical lithography systems. This places stringent constraints on the effective data throughput of any maskless lithography system. In recent years, we have developed a datapath architecture for direct-write lithography systems, and have shown that compression plays a key role in reducing throughput requirements of
such systems. Our approach integrates a low complexity hardware-based decoder with the writers, in order to
decompress a compressed data layer in real time on the fly. In doing so, we have developed a spectrum of lossless
compression algorithms for integrated circuit layout data to provide a tradeoff between compression efficiency and
hardware complexity, the latest of which is Block Golomb Context Copy Coding (Block GC3).
In this paper, we present a modified version of Block GC3 called Block RGC3, specifically tailored to the REBL direct-write
E-beam lithography system. Two characteristic features of the REBL system are a rotary stage resulting in
arbitrarily-rotated layout imagery, and E-beam corrections prior to writing the data, both of which present significant
challenges to lossless compression algorithms. Together, these effects reduce the effectiveness of both the copy and
predict compression methods within Block GC3.
Similar to Block GC3, our newly proposed technique Block RGC3, divides the image into a grid of two-dimensional
"blocks" of pixels, each of which copies from a specified location in a history buffer of recently-decoded pixels.
However, in Block RGC3 the number of possible copy locations is significantly increased, so as to allow repetition to be
discovered along any angle of orientation, rather than horizontal or vertical. Also, by copying smaller groups of pixels at
a time, repetition in layout patterns is easier to find and take advantage of. As a side effect, this increases the total
number of copy locations to transmit; this is combated with an extra region-growing step, which enforces spatial coherence among neighboring copy locations, thereby improving compression efficiency.
We characterize the performance of Block RGC3 in terms of compression efficiency and encoding complexity on a number of rotated Metal 1, Poly, and Via layouts at various angles, and show that Block RGC3 provides higher compression efficiency than existing lossless compression algorithms, including JPEG-LS, ZIP, BZIP2, and Block GC3.