Flow Calibration for Vanilla Klipper (Kobra S1 + ACE)

Automatic pressure-advance calibration using the Anycubic Kobra S1's built-in CS1237 nozzle pressure sensor — on a vanilla / fork Klipper instance connected to the toolhead MCU over a socat tunnel. No extra hardware: it uses the load cell that's already in your toolhead.

Integrates with the ACEPRO multi-material plugin (slot picker, auto tool load, filament temperature from the ACE), and works without ACE too.

Important

Requires the original (stock) Anycubic nozzle-MCU firmware. The calibration relies on the CS1237 pressure-measurement commands built into Anycubic's stock toolhead firmware. The open-source nozzle firmware Kobra-S1/klipper-kobra-s1 does not include these commands , so this plugin will not work on it. Stay on the stock nozzle-MCU firmware if you want to use this plugin.

---

✨ Features

• One command to measure and apply pressure advance: FLOW_CALIBRATE
• ACE integration — pick a loaded slot from a Mainsail/Fluidd popup, auto loads/swaps the tool, reads the filament temperature from the ACE, multi-unit aware (T0–T3 / T4–T7 …)
• Works without ACE as well (normal filament rack)
• Per-nozzle calibration, repeatable (median of N runs)
• Saves the result to your config via the SAVE_CONFIG banner
• Installable & updatable through the Mainsail/Fluidd Update Manager

---

⚙️ Requirements

• Klipper connected to the Kobra S1 nozzle_mcu over a socat tunnel (setup guide)
• [respond] enabled (for the slot popup)
• Stock Kobra S1 macros (G28, TO_THROW_POSITION, CUT_TIP, PURGE_AND_POOP, …)
• Optional: the ACEPRO plugin for multi-material

---

📦 Installation

Update Manager (recommended)

cd ~
git clone https://github.com/Jupsi/flow_calibration.git
cd flow_calibration
bash install.sh

install.sh symlinks the modules into klippy/extras/, registers the repo with Moonraker, and restarts. Updates then appear in Mainsail → Update Manager — one click to update.

Manual

sudo cp extras/*.py ~/klipper/klippy/extras/
sudo systemctl restart klipper

---

🔧 Configuration

Add two sections to your printer.cfg. The fully annotated example is in config_example/flow_calibration.cfg.

[cs1237] — the sensor (stock pins):

[cs1237]
level_pin: nozzle_mcu:PA7
dout_pin:  nozzle_mcu:PA6
sclk_pin:  nozzle_mcu:PA5
register: 60
sensitivity: -2500

[flow_calibration] — the algorithm. Important: copy the interp_fit_factor table from your own machine's stock config — it is machine-specific and gives the best results. See the example file.

After any config change: RESTART.

---

🎯 First-time calibration (once per nozzle)

The measurement is relative, so it has to be anchored to one real print once per nozzle size:

1. Run FLOW_CALIBRATE TOOL=0 VERBOSE=1 two or three times and note the K_raw value from the Flow cal fit … line.
2. Print a PA test and pick the best value by eye (call it PA_real):

   TUNING_TOWER COMMAND=SET_PRESSURE_ADVANCE PARAMETER=ADVANCE START=0.005 FACTOR=0.0005

3. Add nozzle, PA_real / K_raw to nozzle_pa_scale, e.g. 0.6, 2.33.
4. RESTART — now FLOW_CALIBRATE outputs the right value for that nozzle.

The built-in factors (0.2 / 0.4 / 0.8) are rough estimates — calibrate each nozzle you actually use. When i can collect enouhg values i will adjust these in the example config file.

---

🚀 Usage

FLOW_CALIBRATE                 # popup: pick a loaded ACE slot
FLOW_CALIBRATE TOOL=0          # calibrate slot 0 directly
FLOW_CALIBRATE TOOL=0 RUNS=3   # median of 3 runs (best after a filament change)

When it finishes, click the SAVE_CONFIG banner to keep the value.

In your slicer's print-start g-code (after the tool is loaded and hot):

FLOW_CALIBRATE TOOL={initial_tool} KEEP_HOT=1 SAVE=0

QUERY_FLOW_SENSOR checks the sensor without any movement or heating.

Parameter	Meaning
TOOL=	ACE slot (omit → popup)
TEMPERATURE=	override nozzle temperature
RUNS=	repeat N times, apply the median
KEEP_HOT=1	skip cooldown (for print-start)
SAVE=0	don't stage to config
VERBOSE=1	print measurement details

---

🩺 Troubleshooting

Problem	Fix
QUERY_FLOW_SENSOR → NO RESPONSE	check the socat tunnel to nozzle_mcu
Result is always too low / too high	not anchored — do the per-nozzle calibration above
First run after a tool change is way off	use RUNS=3 (median)
[cs1237] ... is not a valid config section	copy cs1237.py and restart Klipper
Popup doesn't open	make sure [respond] is in your config
G28 shuts down: CS1237Sensor ... has no attribute _cmd_reset	your Klipper build already ships its own cs1237.py (shared with the bed probe). Update to the latest release and re-run install.sh — it keeps the build's module and adapts to it.

---

ℹ️ How it works

It sweeps pressure advance while extruding a small zig-zag over the poop chute and reads the nozzle back-pressure from the CS1237 load cell at each step, then robust-fits the best pressure advance. (A small XY move is needed for PA to apply, and a short dwell after each move lets the sensor capture the pressure decay.)

---

Support

If you find this project useful and want to help cover the cost of filament spools from more brands for NFC testing and implementation — donations are always welcome, never expected.

---