A CLOG DETECTED pause is usually a real blockage: on 2026-11-17 a nozzle
blocked right after a swap, the purge check passed (a blocked nozzle also
builds pressure) and the automatic resume only printed air until the next
CLOG pause. CLOG messages are still cleared after a resume, so a leftover
one cannot hide the next NOT FEEDING message.
The purge check now also fails when bd_pressure's delta is still above
pressure_max_delta (15) when the check stops: 16 good purges -3..+10, two
long purges into the blocked nozzle +19/+21 (the short ones did not show it).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Moved from PRINT_END to _PRINT_CLEANUP, which PRINT_END and CANCEL_PRINT
both run, so a cancelled print's filament is deducted too.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The SpoolSense middleware then deducts the print's filament and writes it
to the spool's NFC tag at the next scan. Skipped when spoolsense.cfg is
not included.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
PETG and PETG-CF are too viscous for the 25 mm3/s test flow of the 0.4
nozzle; 12 matches their slicer flow limit. First run at 12 (spool 35,
PETG-CF, 250C) gave PA 0.058.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A NOT FEEDING / CLOG pause during a print runs RESUME by itself after 5 s,
so the lane check and the purge check decide; a failed check keeps the print
paused. At most 2 per print, none within 10 min of printing of the previous
one, only between _PRINT_START_FINISH and the print end.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
aldis (https://github.com/mjonuschat/aldis, GPL-3.0) replaces
scripts/update_all_mcus.sh: it discovers the MCUs from Moonraker and builds
and flashes only outdated ones with the Kconfig each MCU reports. Its source
is a submodule in external/aldis pinned to v0.4.0, the release installed on
the Pi in ~/aldis (Moonraker does not fetch submodules, so it stays empty
there). The old script moves to scripts/archive/; the Katapult script stays,
aldis does not update bootloaders. scripts/README.md: install, setup, use
and the checks for this printer (all MCUs report mcu_kconfig, all use
Katapult). README and CLAUDE.md point to it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Infinite spool (SET_RUNOUT) is set up per print; after a print end or
cancel every lane with a runout lane gets SET_RUNOUT RUNOUT=NONE. Lanes
without one are skipped, so the console stays quiet.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Both now live in the fork https://github.com/MangaValk/bd_pressure
(klipper/, branch main, upstream PR markniu/bd_pressure#13), installed by
its klipper/install.sh and its own Moonraker update_manager entry. The fork's
bdpressure_check.py also reads the v2 firmware diagnostics. bdp_touch.py
stays here. Docs and comments point to the bd_pressure repo.
install.sh only links the files it has, so after this update it no longer
overwrites the bd_pressure links. Run ~/bd_pressure/klipper/install.sh
before updating, or the two links are left dangling.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
At the configured 8 the buffer faults after 30 mm in one state; during the
test lines it needs far more to reach its other sensor, so every lane1
calibration on 2026-11-07 paused with AFC NOT FEEDING. With detection off
the buffer switched after ~80 s and the test finished.
_PA_CFG.test_error_sensitivity (2) is set before PA_CALIBRATE and the
[AFC_buffer] config value is restored after it through
_FIRST_LAYER_FAULT_DETECTION ON=0 (lowered=True), which _PRINT_CLEANUP also
calls when the test stops with an error. Nothing changes when detection is
off in the config.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- _PA_CFG.test_z (default 5): LANE_CALIBRATE_PA lowers the nozzle to this
height instead of Z0 before PA_CALIBRATE, so the test lines are extruded
in the air.
- BLOBIFIER: purge_spd_nozzle {"0.4": 150, "0.2": 75} mm/min, picked by
the pa_nozzle save variable (SET_NOZZLE); other sizes use purge_spd.
150 mm/min is about 6 mm3/s, too much for a 0.2 nozzle.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The slicer already retracts on the last move; 5 mm on top was too much.
The soak retract in _LOAD_FOR_CHAMBER keeps the 5 mm default.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
An AFC swap pauses and resumes the file without the PAUSE and RESUME macros
and only stops on AFC's own error state. A failed check called PAUSE, so the
swap went on and restarted the print while the printer counted as paused.
During a swap (file paused, pause_resume not paused) the check now raises an
error at the end of BLOBIFIER_CHECKED, after the temperature and sensor
cleanup. CHANGE_TOOL catches it and pauses through AFC_error. AFC's remap is
skipped, so the message names the SET_MAP command. Other failed checks still
call PAUSE.
Read from AFC 5ea84ea and Klipper 0cee841; render tests only, not yet run on
the printer.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A PA value only holds for the nozzle it was measured with, so values are
saved per nozzle and spool (pa_value_0_4_spool_<id>). SET_NOZZLE sets the
fitted size; a different NOZZLE= from the slicer on PRINT_WARMUP stops the
print before anything heats. The existing values are copied to 0.4 once at
the first start; the old variables stay in the file, unread.
Each nozzle has its own test flow and PA range (0.2: 5 mm3/s, 0-0.196,
starting estimates, about 9 min a lane). The test primes at the test flow
instead of 120 mm3/s.
Remove the Spoolman write: it went to localhost and never arrived, and
nothing read the values back.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
With the chamber open the bed fans only blow room air past the bed. PRINT_WARMUP
sets print_fans_off for CHAMBER 30 or less, _PRINT_CLEANUP clears it. Chamber
prints, the cooldown and heating the bed outside a print are unchanged.
The PID values in PRINT_WARMUP never took effect: the SET_HEATER_TEMPERATURE
override in bed_fans.cfg only passes on the target. The bed keeps its config
values; the misleading console messages are gone.
Tested with rendered macros only, not yet on the printer.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
BED_SCAN_CHECK levels, scans and warns, without heating or touching the bed.
extras/mesh_check.py fits a smooth surface to the mesh and reports areas more
than 0.03 mm from it, and a range over 0.30 mm.
In PRINT_START a failed check pauses before the nozzle heats. Resume sets the
print file back to the PRINT_START line (M26), so it runs again as normal file
G-code. BED_CHECK_IGNORE lets the next failed check only warn.
Tested on saved scans and rendered macros only, not yet on the printer.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A finished soak called RESUME, which goes through AFC_RESUME. HEAT_SOAK pauses
with PAUSE_BASE, so AFC has no saved position and restores 0,0: the toolhead
went to X0 Y0 at Z5 at up to 350 mm/s (AFC.log, 2026-11-04), and _Z_CHECK then
moved straight back. Both nozzle guard stops in z_check (delta -34 and -33)
happened in runs like this, with the toolhead front left. The exact source of
the force is unknown.
Call _RESUME_PRINT directly: it runs RESUME_BASE and skips the purge, because
PAUSE did not park. A manual Resume during a soak still goes through AFC.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Plastic left on the nozzle sets the contact Z too high and squeezes away
during the run. The first real Verify (2026-11-04, about an hour after a PLA
print) saw bd_pressure and Beacon contact sink by about 0.004 mm per point
while proximity held within 0.006 mm, so it asked for a re-run. That cause
is likely but not yet confirmed.
Same order as PRINT_START: QGL, G28 Z, BLOBIFIER_CLEAN with the guard in log
mode, then the contact step. Refuses to start without BLOBIFIER_CLEAN or
_GUARD_ON. Tested against the simulated printer only.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A closed terminal made print() raise and any unexpected exception skipped the
cleanup, leaving the nozzle low and hot. Both now end in the normal stop.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
scripts/twist_check.py runs on the Pi and measures or verifies the nozzle vs
Beacon offset; windows explain each step, show progress and give a verdict,
with an HTML report in config/twist_data/. It refuses to start during a print,
stops on unexpected readings and never changes printer.cfg.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
At 2.2 force per 0.01 mm (measured 2026-11-02) the old default of 80 was
out of reach within FLOOR from Z0.15 (peak 70 at -0.35); 60 is reached at
about 0.27 mm into the bed and still leaves ~100 readings for the fit.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Stepping down 0.005 mm at a time found no touch even 0.25 mm into the bed
(peak 16): the sensor sets a new baseline by itself while the force looks
steady, and the slow ramp looked steady. A quick press gave about 25 per
0.1 mm. BDP_TOUCH now lowers the nozzle in one drip-mode move (SPEED,
default 0.3 mm/s) while a timer reads the force every few ms and stops the
move at STOP; Z per reading comes from stepper_z's past position, the halt
position is recovered like homing.py does. The threshold needs 3 readings
in a row, and the curve is logged also when no touch was found.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
bd_pressure's endstop output is not wired, so BDP_TOUCH lowers the nozzle
in small steps and reads the force after each one, up to a force limit
(default 80) or 0.4 mm below the start, then lifts back. It reports the
threshold height and the zero-force height from a line fit and logs the
curve to klippy.log. TOUCH_COMPARE heats the nozzle to 150C, homes, and
alternates Beacon's contact probe with BDP_TOUCH at one spot, reporting
mean, spread and range of each and of the difference. Refuses while the
bdp_guard is armed or the sensor is in PA mode.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Beacon meshes every 3 min on 2026-11-02 showed the 10 mm plate settles
~33 min after the bed reaches 100C with the chamber at 50C, and ~11 min at
60C with the chamber open. The old soak ended far earlier (bed_core within
5C, or the chamber reaching its target). HEAT_SOAK gets MIN_TIME=<minutes>
(default 0); PRINT_WARMUP soaks on bed_side with RATE 0.05 over 120 s,
SOAK_TEMP 70% of the bed target, MIN_TIME 35 (chamber) / 12 (no chamber),
and skips the soak without a bed target.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The AM312 case light and nozzle LEDs waited for long commands (heat waits,
AFC loads, QGL) because [gcode_button] runs through the G-code queue. The
new module reads the pin itself, keeps the timeout on a reactor timer and
sets the case light with output_pin's send_async_request and the nozzle
LEDs directly on the led_effect object. CASELIGHT and _LIGHTS_ON keep
working; _STATUS_SHOW reads printer.motion_lights.near.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A large, slow first layer tripped AFC's buffer fault detection ("AFC NOT
FEEDING"). _FIRST_LAYER_FAULT_DETECTION ON=1 at the end of PRINT_START
sets first_layer_sensitivity (4: pause after 70 mm instead of 30 mm);
ON=0 restores the [AFC_buffer] config value. The slicer calls ON=0 at
the start of layer 2:
{if layer_num == 1}_FIRST_LAYER_FAULT_DETECTION ON=0{endif}
_PRINT_CLEANUP calls it too, for prints that end in the first layer.
ON=0 only restores what ON=1 changed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
With TPU in a nozzle heating to 150C the pressure kept changing for
minutes (+43, then -56 as it oozed), faster than the sensor's own
baseline update, and the guard stopped the printer at the start of
z_check and home_xy with the nozzle in the air. Every ARM (also a phase
switch) now writes reset_probe (register 52) and waits 0.2 s without
reading, so only force during the guarded moves counts.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
PA_CALIBRATE takes PA_STEP (default 0.002, 0-0.098 over 50 lines).
_PA_CFG.material_pa_step uses 0.004 (0-0.196) for tpu/tpe/flex; soft TPU
measured 0.094, close to the old end. The "no result" rejection uses the
end of the range actually tested.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
_PA_CFG.material_max_volumetric lowers MAX_VOLUMETRIC for materials that
cannot extrude 25 mm3/s (tpu, tpe, flex: 8). The calibration line shows
the flow used.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- LANE_CALIBRATE_PA: when a calibration will run, assign the lane's spool
again (SET_SPOOL_ID) so AFC reads the Spoolman temperature now; AFC
only reads it at spool assignment and Klipper start (TPU calibrated at
210C after Spoolman was set to 230C). The calibration moved to
_LANE_CALIBRATE_PA_RUN, rendered after the refresh.
- bdp_guard: an emergency stop during PA_CALIBRATE leaves the sensor in
PA mode (FIRMWARE_RESTART does not reset it). When arming, read the
mode register and switch the sensor back to probe mode with a fresh
baseline before watching it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- flex_lanes.cfg: a 5 s poller sets SET_LONG_MOVE_SPEED to 50 mm/s (AFC's
minimum) for lanes whose material is TPU/TPE/flex and back to the
configured long_moves_speed otherwise; reapplied after a restart.
- BLOBIFIER_CHECKED: remembers the temperature of the filament in the
nozzle (saved nozzle_last_temp). When it is more than flush_margin
(10C) above the new filament's, the purge runs at that temperature and
the nozzle cools back afterwards, so the leftover melts out (seen: ASA
leftover with TPU loaded at 210C).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- bdpressure_check.py accepted only extension version 1. The sensor now
runs firmware version 2 (bytes 54-65 unchanged, diagnostics added), so
every purge check reported bd_pressure as not answering. Accept 1 and 2.
- _LOAD_FOR_CHAMBER: the Blobifier only purges with the QGL applied, so
the load before the chamber soak skipped its purge (0.0 mm extruded).
Level the gantry and home Z first when the QGL is not applied.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Test print start: guard peaks 1-8 in every phase except the brush scrub
(peak 47, 4 reads above 30 in a row), 0 I2C errors; a light finger push
gives 41. Stop mode for all phases, the brush stays log-only.
The contact homing in HOME restarted 6 times (sample spread too large)
while the nozzle was still heating to 150C; PRINT_WARMUP now waits for a
steady contact temperature first, as PRINT_START does.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
All macros now come from this repo. bd_pressure.cfg ([bdpressure bd_pa]
and the vendor PA_CALIBRATE macros) and Macros/TEST_SPEED.cfg were loose
files in ~/printer_data/config.
PA_CALIBRATE no longer sets Z back to z_current at the end: it was read
before the test ran, so a Z hop in the G28 XY (Beacon home_z_hop) would
leave Klipper believing the nozzle is lower than it is. Z is never switched
off or moved by the test, so it keeps its position without that line.
The ACC_WALL bug (read from ACCEL_TO_DECEL, always 4000) is documented,
not fixed: the cached PA values were measured at 4000.
printer.cfg must drop [include bd_pressure.cfg] and Macros/macros.cfg its
TEST_SPEED include, else the old copies (included later) still win.
Reference values from the load-check-afc branch and from the guard and
Z check tests added to CLAUDE.md.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The guard's timer read the sensor while G28 ran; bdpressure.py writes the
probe mode at every homing move, and Klipper allows only one I2C request
at a time per device: KeyError ('i2c_response', 1), Klipper shut down
during HOME. The guard now wraps the sensor's i2c in a reactor mutex at
connect; bdpressure and bdpressure_check (now looks up sensor.i2c at
every use) go through it too.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- HOME always homes Z with Beacon contact and recalibrates, also when the
printer is already homed (home_method_when_homed = proximity used the
model of an earlier print without touching the bed). PRINT_START homes
Z again after QGL, before the brush takes the nozzle below Z0.
- _Z_CHECK: after the soak, compare Klipper's Z with a BEACON_QUERY
reading at Z3 over the bed centre and stop the print above 1 mm.
Warns and continues when BEACON_QUERY leaves no new
printer.beacon.last_sample.
- Nozzle guard (extras/bdp_guard.py, print/nozzle_guard.cfg): reads the
bd_pressure force over I2C every 20 ms while armed and, in stop mode,
shuts Klipper down on a touch. Armed for HOME X/Y, QGL, Z homing, brush,
mesh and the Z check; starts in log mode to collect peaks per phase.
- Chamber prints: load and purge the first tool before the chamber heats
(_LOAD_FOR_CHAMBER replaces _UNLOAD_FOR_CHAMBER). A cut unload leaves a
piece of PLA in the extruder that softened in the hot chamber and jammed
the ASA load.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
With the chamber heated, a filament left in the toolhead path softens in
the PTFE tube (PLA loaded, ASA print). After the first HOME, PRINT_WARMUP
unloads the loaded lane when it is not mapped to the print's first tool
(PRINT_START would swap it anyway). A failed unload stops the print
before the chamber heats.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
PRINT_WARMUP: bed heats during the warmup, the nozzle only waits while it
is above 170C (Beacon contact limit 180C), one HOME to get over the bed,
then M190 and HEAT_SOAK. No more filament load and 220->150C cycle there.
PRINT_START (on the soaked bed): one QGL, one contact Z calibration, mesh,
heat over the bucket, load the initial tool, PA for all lanes, load the
initial tool again if a calibration changed lanes, G28 X Y reference.
Loading after the contact keeps ooze off the nozzle for the calibration.
PA: calibrate each lane at its own filament temperature (was the
temperature from before CHANGE_TOOL), and apply the loaded lane's PA
after PA_CALIBRATE_ALL_LANES (the last cached lane's PA won before).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The motion timeout switched the nozzle LEDs off during prints, motion
switched them off when idle or paused, the timeout counted from the first
movement, and every _status_* macro stopped the nozzle light.
- _LIGHTS sets case light and nozzle LEDs from presence, print state and
CASELIGHT MODE=AUTO|ON|OFF; the timeout (60 s) starts at the sensor
release
- _STATUS STATE=... replaces the _status_* macros; logo effects only
cover the logo LEDs and use REPLACE=1
- heater gauge while heating, print progress, paused, error flash from
_ERROR, part ready once the bed is cool, dim night mode when idle and
nobody near, cleaning during BLOBIFIER_CLEAN
- removed unused effects and macros, 12 fps for the toolhead effects
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
action_raise_error aborts while the macro renders, so an M117 in the same
macro never ran and errors such as the RESUME lane check only reached the
console. messages.cfg adds _ERROR (M117, then raise in a separate macro)
and _WARN (RESPOND TYPE=error + M117); all errors and warnings use them.
PAUSE shows AFC's error as the pause reason; RESUME, CANCEL_PRINT and
PRINT_WARMUP clear the display.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
PAUSE stores the lane the print needs (AFC's next_lane during a failed
tool change, otherwise the loaded lane). RESUME stops with an error when
another lane or none is loaded; RESUME LANE_CHECK=0 skips the check.
AFC_CLEAR_ALL_MESSAGES clears AFC's whole message queue
(AFC_CLEAR_MESSAGE only removes the first message).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Lanes whose AFC prep and load sensors do not both see filament are
skipped with a message, as are names that are not AFC lanes. The
"complete" message is now a RESPOND, so it appears after the
calibrations instead of before them.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A bd_pressure that does not answer used to be reported as "pressure
skipped" and the purge passed unchecked. It now fails the check, so in
enforce mode the printer pauses (and a resume purge stays paused).
PURGE_CHECK_IGNORE_SENSOR lets purges pass without the sensor until the
print ends; _PRINT_CLEANUP clears it so the next print checks again.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
bdpressure.py leaves the extruder at the last PA_CALIBRATE test value
(0.098) when it finds no result, and _PA_AFTER_CALIBRATE saved that as
the calibrated PA (lane1 after an AFC pause, lane5 after the sensor went
silent). A run is now rejected, with nothing cached or written to
Spoolman, when the PA is still at the end of the scan, the printer
paused during the test, or bd_pressure gave too few or too few different
readings. The previous valid PA (or fallback_pa) is applied instead.
bdpressure_check reports pa_points and pa_distinct from the bdpressure
module's last PA_CALIBRATE data for this.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
bdwidth's motion reading stays at 0, its per-reading width compensation
made the flow jump between 77 and 119%, and a USB dropout shut down
Klipper. PRINT_START no longer enables it. _PRINT_CLEANUP,
BLOBIFIER_CHECKED and PURGE_UNTIL_EMPTY only send bdwidth commands when
[bdwidth fila_width_0] exists, and bdwidth_status reports present: False
with neutral values when it is not configured, so the section can be
removed from printer.cfg.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
bdwidth reads its CH340 from a reactor timer without error handling, so
"write failed: [Errno 5] Input/output error" shut down Klipper mid print.
bdwidth_status now wraps the sensor's Read_bdwidth (bdwidth.py itself
unchanged): on a USB error it closes the port, reports once and returns
False, so bdwidth retries every 10 s; the port is reopened when the
device is back. New status fields usb_ok and usb_errors.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>