Engineering
Viterbi Decoder Chip Set
While at Microtel Pacific Research in the mid-1980's, I built a Viterbi Decoder chip set for a satellite communications system. The chips were fabbed by NEC on a 2-micron gate array process, and worked as planned.
The chips had some really interesting self-test features, including loopback test modes and an on-chip pseudo-random data generator, so that it could generate data, encode it, decode it, compare the decoded stream to the original, and report whether or not it was working.
In the process of planning to test the Viterbi decoder, I also developed a simple technique for generating low-probability events with good randomness properties. Many engineers are surprised to find out how tricky this can be. I ended up supervising a Simon Fraser University co-op student, Dion Horvat, who ended up getting a master's thesis out of the project.
Speech Coding
For my master's work at Simon Fraser University, I worked on speech coding with Vladimir Cuperman. Together with Rob Pettigrew, we developed a low-delay 16 kb/s speech coding algorithm that vied with an algorithm by AT&T for the CCITT G.728 standard.
AT&T won out, needless to say, but things worked out OK for us. We filed a patent that covered the backward adaptation of the predictor coefficients, which was an idea also used by the AT&T team, but we had precedence. Thus, the CCITT standard infringed on our patent, and so our patent became a valuable commodity as a portfolio play in the late 1990's when internet telephony became important. The bottom line is that Cisco bought the rights to the patent in 1999 for ...
... One Million Dollars ...
But don't worry, I didn't get much of it! I found out the hard way that when you divide a million dollars by two about five times, there isn't much left...