Load Cell Probing Future Work

[garethky]

This is an issue to track ideas for future load cell probing work and take community feedback on what to do next.

Features

• Build sheet detection: probe in a small area and then look at the Z variation. Over short distances textured sheets show about a 20 micron variance, but smooth sheets don't. Z offset/extra squish could be set by sheet type or by the variance measurement.

Sensors

• Support SPI I/O to the HX71x chips. Similar to this idea. It would require hardware SPI capability on the MCU and correct wiring but it could lower MCU overhead and IRQ contention. Doing this before anyone designs a commercial toolhead board around the HX71x might be a good idea.
• Support TI ADS131M02 docs. This was shipped on the Mellow ALPS-V6 link. This could be a good way to get hardware into existing printers without a complete re-design.
• Support Bondtech INDX - pending availability of hardware and an assessment of the board.

Calibration Tools

• drift_filter_cutoff_frequency - This could be tuned by doing a bunch of full height Z travel moves all over the bed (maybe bed mesh locations). The force data would be recorded and the drift caused by bowden tubed would be calculated. Then a filter would be designed to null out the drift.
• pullback_speed - Right now its 1 micron per sample, but I expect there will be diminishing returns above 1000 SPS. It should at least get capped by default.
• pullback_distance - this is hard to tune because oozy taps need longer pullback distances that is seen on clean taps.
• pullback_angle - This is the only required parameter for enabling the tap quality classifier. It would be pretty easy to do a bed mesh and then compute the average.
• minimum_tap_quality - Data analysis showed an optimal point at 4.5 sigma from the mean of the clean taps. But this may not be transferable between machines. This would need a pretty dedicated group of testers to validate.
• temp_coeff or Nozzle Expansion - I wrote a macro to do this: measure_nozzle.cfg, but it would be nice to have this as a gcode command. The messy bit is configuring which z_thermal_adjust section to apply the config to and saving it.
• "Uber Calibrator" - all of the calibration tools are combined into a single command that allows the user to do them in the correct order to bring up a printer from scratch.

Documentation

• How to build a Tap Classifier - explain what they do and what capabilities they have.
• Improve documentation for Tap Quality

[rcloran]

• Support SPI I/O to the HX71x chips. Similar to this idea.

Since there is no real hardware support (SPI data is sent one byte at a time) on any platform except for Linux I ended up changing that PR to only support the Linux MCU. I don't know of any good reason why not to add proper DMA SPI support on other platforms, other than that it's not historically Klipper's default way of doing things. Not sure how many bytes you'd need to represent the data for these ADCs.