This post was first shown on Zach's Hackaday Projects page and is available here.
More background from this past Spring.
The next phase of Neuron development was a bit faster. In terms of hardware, it was more of a 'tock' compared to the 'tick' of Neuron v0.2. Having said that, we still made a few physical changes to our setup, in ascending order of importance:
We purchased and assembled 5 additional Arduino Pro Mini 5vdc modules from Sparkfun. While we did consider moving to a different platform, the Arduino gave us tons of power and flexibility (compared to our ultimate plan of a relatively bare-bones processor); also, we already had 5, so it was a cheap way to get 10 Neurons. That means, at the very least, an interesting single loop.
Andrew got his hands on a monster breadboard; I think it's got 4 rows with a separate power bus. More than enough room to fit ten Neurons with a few support components.
We ditched the NeoPixels. Personally, I was pretty set on the 'pretty blinkenlights' effect; however, Andrew brought up the excellent point that we needed to get something that worked reliably, and RGB fading LEDs might have to wait for now. Reaching back towards the color scheme of Neuron v0.1, we ended up using a pair of diffused T 1 3/4 LEDs, one red (for trigger indication) and one green (for internal potential level).
Andrew took the lead on the firmware for this generation and we ended up with a much more stable design; without the timing and overflow issues, we were free to build networks and run tests on various neural configurations. Once the basic functionality was validated, we began to experiment with the coefficients developed in v0.1. I learned that the v0.2 program had another issue: the decay time was far too long, at least for building a small self-sustaining loop.
I pulled that video link out of an email Andrew sent my way on Feb 23, 2014. My exact response was:
"Oh man, that is insanely awesome. Man. Man that is exciting. Wow.
We need to make a whole crap ton of these things."
Neuron v0.3 didn't stick around for long; the Pro Minis got re-purposed into various other projects, including a reflow oven controller, a MIDI interface, GimbalBot's test rig, and a few other random projects. But the proof of concept worked well enough; we could definitely see the potential for awesomeness with v0.3.
Time to scale up production.