TL;DR: Kicad kicks ass.

I've spent most of my life staring at schematics in one form or another, but everything I've built to this point used some kind of point-to-point breadboard technique. Many years ago I discovered the RadioShack 276-159; it can handle a smallish DIP IC and a few peripherals and is readily available up the street. Great for LM3909s, 555s, 741s, and all manner of random interface boards and LED blinkers.

No good for SMD work, however. For the Neuron, I wanted to go full custom to minimize size and maximize awesomeness. Taking [Chris Gammell's] advice, I took my first plunge into circuit design using Kicad.

Circuit Design

I wanted Neurons to be incredibly simple. As such, they communicate by sending 5VDC pulses to each other--that's it. The interconnections between Neurons--I call them Axons in an attempt to be somewhat biologically accurate--require three conductors, VCC, Gnd, and signal. Taking a cue from servos, I put the VCC conductor in the middle so miswiring would most likely just ground out an input or output rather than short out a whole string of Neurons. As I mentioned in the last post covering the BOM, the only components beyond the seven board-level connectors, microcontroller, and LED are passives needed for signal filtering and current limiting.

A few notes:

• K1-K7 are 2mm spacing TE Connectivity 3-conductor headers
• C1 and C2 are the two filtering caps recommended in the ATTiny44A datasheet
• R1-R6 are pulldown resistors for the dendrites (inputs)
• R7-R9 are current limiting resistors for the RGB (OGB?) LED
• K8--ah, K8. You'll notice that I didn't include a JTAG header for programming--fortunately, all of the JTAG pins on the ATTiny are shared with various inputs and outputs, so I can program the boards using an adapter harness. All except the Reset pin, so I added a jumper. Simple enough.

Board Design

I wanted to minimize area--not just to save cost, but to make each Neuron as compact as possible. Auto routing is for suckers, so this is what I ended up with:

Again, notes:

• VCC and Gnd traces are 30 mil--pretty much the largest I could do. Neurons feed power to each other and I wanted to maximize the number per power input; based on some very fast web research, I think that should be good for 1A or more.
• I wish the Axon connector was centered. Oh well. At the least the LED is centered in one dimension.
• Tiny! Well, pretty tiny. I was shooting for <1" on each side but that wasn't in the cards.
• No vias. Vias are for suckers. Or something. That was fairly arbitrary, actually.
• K8, the Reset connector, is just a loop of wire.
• Red is Component Side. Green is Solder Side.
• Cyan is silkscreen on the component side. This was back when I planned for 3 Inhibitory inputs--not the case presently, but more on that later.
• Magenta is silkscreen on the solder side. Neuron v0.4, Salfred Labs.

Salfred Labs

That brings up a great point. What is Salfred Labs, and why is it silkscreened on the Neuron board?

Andrew Salveson + Zach Fredin = Salfred Labs! It's what we thought about calling our company, if we'd made a company. Probably won't happen now, but the legacy lives on in Neuron v0.4.

More to come! Happy turkey day!