Real-time neural data acquisition with Soma
In graduate school, I attempted to rebuild our entire data acquisition system from scratch -- needless to say, after years of 80-hour weeks, it didn't work out. It was a large ambitious project that collapsed under its own weight due to the second-system effect.
I did get the analog mixed-signal front end up and working through production -- 8+ channels of ultra-low-noise differential acquisition and digitization with a bunch of DSP in a FPGA, all sent out over electrically-isolated, inexpensive plastic optical fiber.
I hope this design can serve as a reference design for others' mixed signal work, and I hope the project can serve as a warning to graduate students everywhere:
Hey graduate students, don't take on 5-year projects outside your lab's core competency that don't obviously translate into publishable units.