Audio / FreeCAD / measurement / embedded control
Three-way active speakers based on the SEAS KingRO4Y MkIII design. Plywood cabinets, routed baffles, Hypex FA503 plate amps.
CAD, woodworking, DSP presets, USB HID reverse engineering, BLE control work, driver repairs, and one very VERY loud mistake I would rather not repeat.
Speakers built and playing. USB control is working on the bench. The physical knob is still more notes and prototype code than finished object.
FreeCAD
The whole cabinet as a parametric model: the assembly, the spreadsheet the dimensions come from, section views through the enclosure, and the baffle drawing the router work was cut from. Under two minutes, with the commentary.
Cabinet build
Front baffle from the drawing into plywood. Two big circles, one coaxial driver, one woofer, and not much room for casual error.
Cabinet
FreeCAD model for the enclosure and baffles, then drawings for the parts that actually needed cutting. The front panel is the awkward one: driver rebates, bevels and screw positions all need to agree before the router starts.
Plate amp
FA503 amp plate in the back of each cabinet. Audio, mains, USB and cooling all meet in one expensive rectangle, so the rear baffle got its own drawing.
Finish
Plywood left visible, with a gloss finish that shows every sanding decision. Looks good when the light catches it. Also shows where I got impatient.
Finish
Assembly
First listening
Routing
Rear panel
Rear baffle
Finish
Physical object
The funny part of active speakers is how ordinary they look once the drivers are in. Most of the project is now hidden in the rear plate, DSP files, wiring and measurement notes.
Plate amp
Driver
Wiring
Driver
Measurement rig
Log sweep from the PC, digital audio into the speaker, calibrated mic back through the Scarlett. The first job was not tuning for taste. It was finding out whether both speakers were alive, wired the same way, and safe after the mistake.
Fault finding
Measurement showed the left speaker failing above about 2.5 kHz. Later repair work found a tweeter polarity problem that made a deep cancellation notch around 1.6 kHz.
Loud incident
While reverse-engineering the Hypex USB packet I zero-filled bytes that turned out to include persistent startup volume. The amp came back at 0 dB. Nasty, educational, and fixed with strict read-modify-write plus a startup-volume tripwire.
Afterwards
Re-ran the sweep against the saved baseline. The right speaker survived. The notes are now more paranoid about packet tails, pacing and volume slew limits.
USB protocol
FA503 control is USB HID with fixed 64-byte packets. Volume is little-endian dB times 100, presets and input source are fields in the same state packet, and some opcodes are best left alone unless reflashing the amp sounds like a good evening.
Knob
The plan is deliberately boring to use: turn knob, both speakers follow. The prototype path is BLE from knob to ESP32-S3 dongles, then USB from each dongle into its own Hypex amp.
Bench status
The ESP32-S3 can read amp status and write volume changes. The useful version now preserves unknown bytes, spaces out requests and refuses silly volume jumps.
