Skip to content

build log /choosing-the-brain

Entry 02

Choosing the brain

pcb process

Every Rodeo instrument has the same heart: an Electrosmith Daisy Seed on the Rodeo motherboard. This log covers why I picked it, the version I started with, why I switched, and what the board around it ended up looking like.

Why the Daisy

This one was easy. I’d been using the Daisy for a while, and the honest answer to “why the Daisy?” is “everything about it”:

  • The audio. The codec is great, so the first thing you hear is clean.
  • The software. libDaisy and DaisySP are easy to use. DaisySP gives you oscillators, filters, envelopes and effects out of the box, so you’re making sound on day one instead of writing DSP from scratch.
  • The community. Plenty of people have built on it already, so when something breaks, someone has usually hit it before.
  • I already knew it. When the goal is an instrument, you don’t want to also be learning a new microcontroller.

I’ll probably do a Teensy version of the motherboard eventually. But the Daisy was the obvious place to start.

Starting on the DFM, and switching back

The first version of the motherboard was designed around the Daisy Seed2 DFM, the production-friendly version of the module. It was the newer part, and it seemed like the right choice.

It didn’t stay that way. The DFM has some awkward requirements for a board like this one:

  • It wants at least 6.1 V on VIN. I wanted to power the whole board from USB-C, which gives you 5 V. That’s a real problem when your main power source is below the minimum.
  • Its pins are split into five 2×5 banks instead of one familiar strip, and the audio pins come out in a way that needed untangling on the carrier.

Meanwhile I had a regular Seed sitting on my desk. The classic Seed footprint, a single 40-pin strip, is happy with 5 V on VIN and has simple single-ended audio. The newer Seed3 is pin-compatible with that footprint too. For this project the DFM just wasn’t adding much value, so I moved the design back to the classic footprint.

Straight into KiCad

There was no breadboard phase and no paper sketches. I started drawing directly in KiCad, and the first usable prototype was the Rev A board itself.

It’s a 4-layer board. I’d made a few 4-layer boards before, and for something like this it’s just the right way to do it. Routing is far easier with internal planes, and a board full of ADC traces benefits from solid ground underneath them.

What’s on the motherboard

Here’s what Rev A ended up with:

  • USB-C power. 5 V from VBUS feeds the Daisy, and an on-board LDO makes a separate, clean 3.3 V rail for everything else.
  • 24 analog inputs from two external ADS114S08 ADC chips over SPI. Each input is a 3-pin JST connector carrying 3.3 V, signal and ground, so a pot board just plugs in. The pots and the ADC reference run from the same 3.3 V rail, so the readings are ratiometric: if the rail drifts, the knobs don’t.
  • 12 digital I/O on 2-pin JSTs, for buttons and the like. Six of them can also read analog, so they double as spare analog inputs.
  • I²C, SPI and UART expansion headers. The I²C bus is where the encoder hub and the OLED connect.
  • MIDI in and out on 3.5 mm TRS (Type A), with an optocoupler on the input as the MIDI spec calls for.
  • Stereo line in and line out on 3.5 mm jacks.
  • A status LED, because every bring-up starts with “is it alive?”

Why JST cables, everywhere

The whole point of Rodeo is that you don’t have to wire controls by hand, so the cable choice mattered. I went with plain JST cables between the motherboard and every control board, not ribbon cables or board-to-board headers.

The reason is that I want Rodeo to be completely interface-agnostic. The motherboard doesn’t know or care where the controls are. Put them right next to it and you’ve got a tiny portable synth. Spread them across a huge panel and you’ve got a giant one. The only thing that changes is how long the cables are.

That choice is something I’m open to rethinking for Rev B. But it’s the reason Rodeo looks the way it does.

If I could tell day-one me one thing

Pay very close attention to JST orientation. It’s genuinely confusing: which end is pin 1, whether a cable is straight or reversed, and what the vendor’s label actually means. It caught me later on. One harness got built the wrong way round, and that’s a story for a future log.

Next: twenty-four knobs on one microcontroller. A closer look at the analog surface.

Follow along?

Every entry goes out as one email, roughly monthly. The Kickstarter is announced here first.

You will get one email asking you to confirm. Nothing else until there is something to say, and unsubscribe from any of it.