Application on Sovol on mainline Klipper

KIAUH stack

Mon, 01 Jan 0001 00:00:00 +0000

The application stack via KIAUH#

Everything above Klipper — Moonraker, Mainsail, Crowsnest — is already stock upstream, so the fastest clean install is dw-0/kiauh. Clone it and install, in order: Klipper, Moonraker, Mainsail, Mainsail-Config, Crowsnest.

Point Klipper’s host at the same master checkout your MCUs were built from, with a venv built from its requirements.txt, and set [update_manager klipper] channel: dev in moonraker.conf so it tracks master.

Two post-install fixes#

Both come from experience on this board:

Klipper config

Mon, 01 Jan 0001 00:00:00 +0000

Translating the config to mainline#

The vendor printer.cfg doesn’t run as-is on upstream Klipper — a few sections depend on vendor-only firmware behaviour. Translate these off the vendor config; the rest carries over.

Eddy on software I2C#

The vendor ran the LDC1612 on hardware i2c2, but only because its F103 firmware carries STM32F1 hardware-I2C errata workarounds that mainline lacks; on mainline, hardware i2c2 throws START_NACK → shutdown. Use bitbang instead:

Plugins

Mon, 01 Jan 0001 00:00:00 +0000

Plugins and venv extras#

The add-ons this machine actually runs, on top of the stock KIAUH stack.

scipy — required by the eddy probe#

Install scipy into the Klipper venv:

~/klippy-env/bin/pip install scipy

Without it the eddy probe calibration fails. If PyPI stalls when downloading from the board, sideload the wheels — see troubleshooting.

sovol_codes — the knob-screen codes#

The vendor firmware shows numeric codes on the knob screen (101/103, the 60+ shutdown range); mainline shows human-readable messages. To keep the codes on upstream Klipper, install the opt-in sovol_codes.py plugin:

Calibration

Mon, 01 Jan 0001 00:00:00 +0000

Calibrate and verify#

With both MCUs on mainline, the config translated, and plugins in, calibrate the eddy probe and motion, then prove it with plastic.

Eddy and tap#

Run in order, SAVE_CONFIG after each:

LDC_CALIBRATE_DRIVE_CURRENT
PROBE_EDDY_CURRENT_CALIBRATE — builds the freq→height table
PROBE_EDDY_CURRENT_TAP_CALIBRATE — sets tap_threshold

Belt tension#

Aim for ~110 Hz plucking the XY belts over their 150 mm span. (The Z belt is factory ~137 Hz, but that’s a 21.5 cm span — a different length, so don’t compare the two numbers.) Tighter than that only cleans up the shaper graphs on paper and adds ringing. Klipper’s belt similarity test is unreliable here — the head-mounted accelerometer is noisy at standstill from motor hold-current buzz, so its readings swing run to run. Check tension by ear or a phone tuner instead (Android works; iPhone noise suppression corrupts the reading).

Extras

Mon, 01 Jan 0001 00:00:00 +0000

Extras — optional hardware and practices#

Additions that are not part of the core migration but earned their place on a real machine.

USB WiFi dongle (when the onboard radio dies)#

The onboard RTL8189FTV is a known single point of failure — on this unit it died of PSU-inflicted causes (mmc: error -110 whilst initialising SDIO card on every boot, on stock firmware too, through ~21 attempts across two kernel stacks; software revival — power-cycling the chip via its GPIO, longer post-power-on-delay-ms, lower SDIO clock — changes nothing). If you see that signature, stop debugging and plug a dongle into one of the board’s USB-A ports.