Drive the purifier fans through stock [fan_generic] objects

[justinh-rahb]

Summary

Moves the purifier's exhaust and inner fans onto stock Klipper [fan_generic] objects, so they are addressable like every other fan on the machine — SET_FAN_SPEED FAN=exhaust_fan, M106 P3, the control/generic_fan endpoint, printer.fan_generic circulation_fan in the object query API, and the Fluidd/Mainsail fan panels.

Today both fans are driven by PurifierFan/PurifierFanTachometer, private classes inside purifier.py that reimplement what fan.Fan already does (PWM setup, kick-start, tachometer). Nothing outside SET_PURIFIER and the control/purifier webhook can see or drive them.

Net effect on klippy/extras/purifier.py: -114 lines (1197 → 1083).

Approach

1. [purifier] no longer takes raw fan pins. The pin/PWM/tachometer config moves into [fan_generic exhaust_fan] and [fan_generic circulation_fan], and [purifier] references them by name via exhaust_fan_name / inner_fan_name. Purifier.__init__ resolves them with load_object(config, 'fan_generic ' + name).
2. PurifierFan and PurifierFanTachometer are deleted.
3. PurifierFanRouter wraps the fan.Fan each [fan_generic] object holds and overrides only set_speed_from_command(), delegating everything else through __getattr__.

Step 3 is what makes step 1 safe, and is the part worth reviewing.

Why the router is required

[fan_generic] registers its own SET_FAN_SPEED mux command and control/generic_fan endpoint, and extras/fan.py reaches through fan_obj.fan.set_speed_from_command() for M106/M107 fan-id mapping (fan.py:192, fan.py:206) and for power-loss resume (resume_all_fan_speed(), fan.py:178).

All of those drive fan.Fan directly, bypassing set_exhaust_fan_speed()/set_inner_fan_speed() and everything they own:

• power_enable_pin — the fan power rail. Never asserted, so the fan gets a PWM duty cycle with no power.
• the delay-off timers
• the _power_detected presence gate
• inner_work_time, which tracks filter life

Every one of those entry points converges on set_speed_from_command(), so routing that single method covers all of them, with no changes to fan_generic.py, no dependency on config section order, and no reliance on gcode/webhooks internals. Purifier.set_*_fan_speed() calls the real fan's set_speed(), which the router does not intercept, so there is no recursion.

Measured without the router, on hardware: a circulation fan started with SET_FAN_SPEED keeps spinning at 4652 rpm through both SET_PURIFIER FAN=inner SPEED=0 and SET_PURIFIER_MODE MODE=0, because set_inner_fan_delay_turn_off() returns early on FAN_STATE_TURN_OFF.

Fan state enum removed

With one write path per fan, _exhaust_fan_state/_inner_fan_state no longer carry anything not already in fan.Fan.last_fan_value:

old	derived
state == FAN_STATE_TURN_OFF	last_fan_value == 0
state != FAN_STATE_TURN_OFF	last_fan_value != 0
state == FAN_STATE_TURN_ON	clamped speed != 0

FAN_STATE_TURNING_OFF only ever meant "not off" — the fan is still spinning through the delay-off window — so the derived checks give the same answer. The speed clamp is kept, which is what keeps the accrual trigger equivalent.

Unchanged

SET_PURIFIER, GET_PURIFIER, SET_PURIFIER_MODE, WAIT_CHAMBER_TEMP and the control/purifier webhook are untouched, as are the chamber modes, dynamic cooling ramp, RPM fault detection, presence detection and the purifier status object.

Config change

[purifier] sections carrying raw pins will no longer load — exhaust_fan_name/inner_fan_name are required. lava/printer.cfg is updated to match; see the diff for the exact split.

The module and the config ship together in a firmware image, so the two always move as a pair and there is no migration path to support. The required-with-no-default choice is deliberate on that basis: a [purifier] missing either key is a packaging error worth failing loudly on, not a machine that should come up with a silently absent fan.

M106/M107 fan-id mapping

extras/fan.py already carries an exhaust_fan/exhaust_fan_id pair under a comment reading "exhaust fan / purifier fan", and Snapmaker's own factory configs use it (factory2.cfg sets exhaust_fan: purifier, exhaust_fan_id: 3). It was unusable from printer.cfg because the purifier's fans were not [fan_generic] objects and fan.py resolves the mapping with lookup_object("fan_generic <name>"). lava/printer.cfg now points it at the exhaust fan, so slicer-emitted M106 P3/M107 P3 drive it.

This is also the case that makes the router necessary rather than merely tidy: cmd_M106/cmd_M107 call fan_obj.fan.set_speed_from_command() directly (fan.py:192, fan.py:206) and never touch the SET_FAN_SPEED mux command, so intercepting at the dispatch layer would not have covered them.

Testing

On a U1 with a purifier attached, Klipper 1.6.0.267_20260815150420:

check	result
SET_FAN_SPEED FAN=circulation_fan SPEED=0.6	4645 rpm
SET_PURIFIER FAN=inner SPEED=0 after the above	0 rpm
SET_FAN_SPEED FAN=circulation_fan SPEED=1.0	6294 rpm
DELAY_OFF=5 against a SET_FAN_SPEED-started fan	armed, fired at ~5s
control/generic_fan S=70, then SET_PURIFIER ... SPEED=0	0.7 → 0
SET_FAN_SPEED FAN=exhaust_fan SPEED=0.5	purifier reports exhaust_fan.speed: 0.5
SET_PURIFIER FAN=inner SPEED=0.3	reported as 0.3 / 3520 rpm in the generic view
presence gate on the generic path	SET_FAN_SPEED refused while power_detected is false
M106 P3 S128	speed 0.502 (128/255)
M106 P3 S255	speed 1.0
SET_PURIFIER FAN=exhaust SPEED=0 after M106 P3	0 — purifier can stop what M106 started
M106 P3 S200 then M107 P3	0.784 → 0

The delay-off row is the meaningful one for the state removal: delay-off only arms when the fan reads non-zero, so a timer firing on a fan that SET_FAN_SPEED started shows the derived check matches the old enum.

Credit

The [fan_generic] migration and the PurifierFan removal are @paxx12's design, from paxx12-snapmaker-u1/SnapmakerU1-Extended-Firmware#670. This PR carries the same approach against a stock 1.6.0 tree, adds the router, and drops the state enum. The two implementations converged independently and are functionally equivalent.

Offering it upstream so the community firmware and mainline do not diverge on this, and so the fix lands where the exhaust_fan_id hook already lives.

---

🤖 Generated with Claude Code

[justinh-rahb]

Screen.Recording.2026-08-26.at.10.21.19.AM.mp4

You can now set exhaust fan speeds per filament profile from here: