Back in March, Headwaters quietly switched its reference brain from the Raspberry Pi 5 to the Compute Module 5 on a Waveshare base. We wrote up the why at the time. Some of it lands better on camera than on a page, so here is the video walkthrough: what was wrong with the old stack, what the new one looks like, and what it freed up underneath.
The short version
The Pi 5 was getting hard to source in the configurations Headwaters needed. Once you added an NVMe base hat for tileserver-grade storage and the ribbon cable to connect them, the bill of materials kept climbing and the assembly turned into a fiddly stack that nobody wanted to debug at the kitchen table. The CM5 on the Waveshare base solves all of that in one part order. NVMe is on the base. CAN is a clean hat. Power, USB, and storage land where you want them, and the carrier-board EDA we used to carry in the Headwaters repo is gone.
If you want the per-decision breakdown in writing, that is in Goodbye Pi 5, Hello CM5. The video is the visual companion to that post.
What is in the video
A walk through the old Pi 5 stack and where it kept costing time. The Waveshare base for the CM5 and what it lays out for you. CM5 module variants and why the choice matters less than it used to: any variant with 4 GB of RAM or more works, with or without onboard WiFi, with or without onboard eMMC. The NVMe slot, the CAN hat, and what the assembled brain looks like before it goes in the case. A few notes on what shipped in the image when we cut over, including a first-login splash screen, ASCII art that tells you which TrailCurrent device you just SSH'd into, a cleaner SSL cert generator, and cellular SMS notifications through the onboard router for trips off WiFi.
What has changed since the move
The CM5 is no longer the only board the Headwaters image runs on. The Radxa Dragon Q6A joined as a second reference platform with on-board NVMe, on-board 12 V input, and a cheaper CAN hat. The story of adding it is in Headwaters Gets a Second Platform, and the working confirmation is in Headwaters on Radxa: Confirmed Working. The image supports both, so you can pick the board you can actually buy and run the same Headwaters stack on top.
That is the pattern this video is really about. Where there is an off-the-shelf path that is faster, cheaper, and more reproducible than a custom carrier, we take it. Custom PCBs still earn their keep on modules where the I/O density and the connectors demand it. For the brain of the system, a catalog stack wins.
The behind-the-scenes channel
This one went up on @trailcurrentopensource, the second TrailCurrent YouTube channel. The main @trailcurrent channel covers what the system does and the value of running it on your rig. The open-source channel is for the work underneath: hardware reviews, firmware deep dives, design choices, and the kind of detail that does not belong in a product video. If that mix is what you came here for, that is the channel to subscribe to.
Where to go next
If you want to build one, Build Your Own Headwaters lays out the bill of materials, the print profile, and the steps from boards on the bench to a running gateway. The image and full source for Headwaters live in the trailcurrentoss GitHub organization. The ESP32 sensor modules that ride the CAN bus alongside it can be flashed from the browser at the web flasher.