Similar story. I got my first job assignment at to do a comparison of hardware for the EUMETSAT METOP ground station. This was 2004 and the leading options were custom hardware ASICs in the $1-10M range. There was a snow storm and I had ridden my bike to work, and actually got hit by a car on the way, so I didn’t really feel like trying to make it home, and $10M seemed like a lot of money when I didn’t even know how fast my new Pentium 4 desktop could do it. I slept in my office for the next three days while most of the lab took snow days, trying to write the decoder in C. How hard could it be, right? I’d never taken a CS course, and assumed it was just a little bit of bit-banging to decode a structure or whatever. It was, until I came across this error-correction thing called Reed-Solomon. I knew about CRC, and thought that was the end of the story, but this used a bunch of totally unintelligible ‘new math’ called Galois fields, and it just happened to have been invented at the place I was working at, and by the way, the guy who gave me to assignment was named Solomon, unsure of any relation to Gustave, and I felt I was setting myself up for great embarrassment. Well anyway, this place had a great library, so I read the original paper, still getting nowhere, and then a book on the subject, also unintelligible, except that it included lots of very detailed pseudo code. When the lab returned, I had something working really slowly, but the hardware engineer thought it was cool enough to keep going, and built a PCI DAQ card to read the I/Q inputs directly into the computer. Eventually I made it fast enough, order of 10Mb, and that’s how NOAA made our weather reports during the decade when NPOESS VIIRS was delayed. I adapted it to VIIRS too, but only stuck around long enough to see it used in testing. I was very proud of myself, and was recognized with an award of zero dollars and the non-exclusive right to post a reply about it on Hacker News when the topic finally came up.