Variegated Coffee

2025-10-29: State of the union

Posted on Oct 29, 2025 by Magnus Nordlander

It’s been about a month since my last entry, and quite a bit has happened in that time. I’ve successfully validated gear pump operation (with some caveats), the TFT display is now functional, and - last but not least - just received my order of the EV2 revision of the GS3 carrier board. Time for a comprehensive status update on where things stand and where they’re heading.

Gear Pump Testing

I’ve successfully tested a Fluid-o-tech FG304 gear pump with the GS3, and I’m happy to report that it works. Pressure profiling is now a reality on this machine. However, “it works” doesn’t mean “it’s perfect” - the pressure control PID tuning is currently pretty rough, and that’s high on my list to improve.

I also added an over-pressure valve as a safety measure. While the pump shouldn’t overpressurize with normal control, the pump is perfectly capable of reaching pressures of 15 bar or higher and the rest of the machine simply isn’t made for that.

The next steps for the gear pump integration are both straightforward and fiddly:

  • Dial in better PID tuning for smoother pressure control
  • Design and manufacture a proper mounting bracket that handles noise damping (the pump is small, but vibration isolation is critical)
  • Drill mounting holes in the drip tray for an additional outlet for the OPV
  • Actually remove the rotary pump and permanently install the gear pump
  • Find a home for the 24V power supply

I might start on some of this before the EV2 boards arrive, since the current board is functional enough for testing.

Display Progress

The TFT display is working well. It’s still running a pretty basic UI at the moment, but having a proper high-resolution color display opens up a lot of possibilities that the 16x2 LCD simply couldn’t support. Performance can always be improved - display code tends to be a constant optimization target - but it’s responsive enough for daily use.

Gravity Board

The Gravity scale board still hasn’t made it into the machine, but there aren’t any real blockers at this point. It’s more a matter of prioritization - getting the gear pump dialed in has felt more pressing for now.

EV2 Revision

I’ve begun initial validation on the EV2 revision board. This revision incorporates everything I learned from the EV1, plus adds some capabilities I’ve been wanting for future features.

The headline improvements include:

  • Two 24V PWM-capable outputs for proportional solenoid valves (particularly for that steam wand upgrade)
  • On-board LEDs to show output states, making debugging significantly easier
  • User-assignable LEDs and an additional button for future UI improvements
  • More broken-out IO for additional sensors - potentiometers, flow meters, whatever you might want to add
  • PCB changes to improve EMI performance
  • A UART connector wired to the ESP32-C6, which is electrically compatible with the La Marzocco IoT board (though reverse engineering their protocol is an entirely different challenge I’m not tackling right now)

There’s one trade-off worth mentioning: the SD card is now SPI-attached rather than SDIO. It’s a bit slower, but much easier to interface with, and being able to free up the IO for other things made it worthwhile. Beyond that, it’s mostly fixes for things that weren’t quite right in EV1 and general improvements to manufacturability.

Short term software Roadmap

Looking ahead at the firmware development, I’ve got a pretty clear picture of what needs to happen in the short-ish term.

  • Better status UI for the TFT - making better use of that screen real estate
  • Performance counters for tracking system behavior
  • Shot log entries to SD card
  • Display code refactoring - currently this isn’t modular
  • Shot graphs on the TFT during and after extraction
  • Basic settings and configuration UI for both TFT and the stock 16x2 LCD
  • Sensor fusion combining pump flow, inlet flow, backpressure, and output flow to create rock-solid output weight measurements
  • Debug logging to SD card and over WiFi

Long-term hardware plans

On the hardware side, I’ve got some more ambitious plans as well both for machine mods and for new parts to manufacture:

  • Removing the gicleur to allow the gear pump fuller control over flow
  • Installing a 24V proportional solenoid valve for the steam wand
  • Designing and manufacturing new front fascias with:
    • An integrated shot mirror
    • Proper mounting points for load cell magnetic connectors
    • Mounting for a larger touchscreen HMI board
  • Actually designing and building that larger HMI board
  • Adding potentiometers, possibly both an EP-style paddle and some other potentiometer for steam

These longer-term items will require significant CAD work and probably some iteration to get right, but the EV2 board is already prepared for them with the additional PWM outputs and IO breakouts.

Looking Forward

As per usual, I’m hesitant to make promises regarding the timeline but I am making steady progress, and I am keen to get to a state where using the new board feels like a nice experience. I don’t think it’s unreasonable to expect at least the software tasks identified above to be done some time around January.

As always, the Discord is a great place to chat about the project, and I try to post some photos and short video clips as I make progress on the project Instagram.