Mini Mac Classic

The machine is a Raspberry Pi Zero 2W running minivmac against a Mac II ROM, booting directly into System 7.1 inside a miniature Macintosh shell. Raspberry Pi OS sits underneath it with labwc in place of a desktop environment.

The goal was never an emulator on a Pi in a case. It had to feel like a real vintage machine, which meant instant boot into the OS, a physical brightness dial, real shutdown and reboot buttons, clean audio, and no visible Linux anywhere. Almost every component in that list fought back at least once, and those fights turned out to be the actual project.

minivmac ships as source, so it had to be compiled for the Pi before anything could run. It then boots from a disk image you build yourself, out of several virtual floppy images. That part is straightforward once you know how it works. Getting to that point, with no prior experience of minivmac and none of early Mac OS either, took the longest.

Much of the software for that process is still archived as .sit StuffIt files, and those became a problem of their own. Some wouldn't extract on a modern machine and some needed older tools entirely, so I ended up moving between three environments depending on the archive: a modern PC, my G4 iMac, and an extractor running inside minivmac itself. No single tool handled everything, and working out which environment could open a given file became part of the workflow.

The bracket started from ThisDoesNotCompute's design, and I wanted to keep the glass front panel over the display. It's a small thing that does most of the work in making the screen read as authentic instead of a bare LCD bolted into a shell.

Committing to a working dimmer was the harder route, and it decided the bracket. The original design carried more than the display, and there was no longer room for the encoder and the front-panel buttons alongside it. So the bracket had to be redrawn in TinkerCAD to hold the display and nothing else, while still keeping the glass in front of the panel. Three revisions to get the fit right, with Gerry Gonzalez printing every one.

The dimmer and the front-panel buttons both started from code Sideburn shared. I wanted a tactile way to dim the screen rather than a setting buried in a menu: a rotary encoder wired to two GPIO pins, read continuously by pigpio, driving the backlight over PWM.

Making the encoder behave meant a glitch filter on both pins and getting the direction logic right, so one way reliably brightens and the other reliably dims without the reading jittering back and forth on every detent.

minivmac is compiled to talk directly to ALSA, the low-level Linux audio API. My first instinct, and the internet's usual advice, was to fix sound through PipeWire and WirePlumber. None of it did anything, because minivmac never touches that layer. Old AUDIODEV environment variables were just as useless, an OSS-era relic the emulator doesn't read.

The fix was a hand-written /etc/asound.conf: a plug device with a route layer folding stereo down to mono for the single speaker and sending it to the USB card. My first attempt used ALSA's dmix, which looked like the standard approach and returned "invalid sample format" errors instead. Dropping it for the simpler chain solved it outright.

A vintage Mac shouldn't show you a kernel boot log. Getting there was a stack of small fixes: stripping the scrolling console text out of cmdline.txt so nothing but the splash is ever visible, masking PackageKit so it stops raising update notifications, and setting a black wallpaper as a clean fallback.

The boot sequence needed choreography rather than just a splash screen. labwc's autostart launches minivmac in the foreground immediately, bypassing the file manager and panel completely. Linux only resurfaces if the emulator is closed, which under normal use never happens.

Four things this machine will not do, and no amount of work was going to change two of them. They are worth stating plainly, since the workarounds are what's actually running.

The machine boots, displays, and plays sound, the GPIO controls work, and the emulator produces the expected Mac II chord on startup.

The recurring lesson arrived over and over in different clothes: fix problems at the layer that's actually responsible, not the layer that's easiest to blame. PipeWire looked like the audio culprit. dmix looked like the correct ALSA approach. The USB dongle looked like the source of the hum. None of them were. Every real fix came from tracing the signal or the code path back to where it started instead of patching the nearest visible symptom.

It's a small machine, and none of it was easy. Making it behave like a genuine miniature Macintosh Classic, a dedicated appliance that looks like it always worked this way, came down to the small details that were worth taking the extra time for.

This build leans on five people. Without the work they shared, it would either not exist or would have taken far longer to get running.