I still think this should be Delta only.
Too many steps per mm to be accurate and enough torque to bend the bed or damage the nozzle.

That's why I put the following caution in the description of the SENSORLESS_PROBING feature.

CAUTION: This could cause damage to machines that use a lead screw or threaded rod
         to move the Z axis. Extreme care must be taken when attempting to enable
         this feature.

I think I'd rather see #if ENABLED(DELTA) and then remove it if someone can show it working safely on a lead screw driven system.

With the addition of these sensitive features, maybe it would be good to have a separate sketch that does testing on the stepper drivers to figure out the proper value limits and ranges —especially for stall sensitivity— so users can be pretty confident about them before plugging them into Marlin and leaping straight to G29.

Or… have a marlin_calibrated flag that is false by default, and only gets set true after the machine has been put through an initial setup and testing procedure. Until that flag is set (requiring an EEPROM) the firmware will always boot into the setup procedure.

For this, I just used G30 and M914 to dial in the stall sensitivities prior to moving on to G33 and G29.

Also, I believe having a separate sketch isn't currently possible with the LPC1768 HAL.

A calibration procedure would be nice, but isn't always necessary. As I've dialed in the parameters like Z offset and stall sensitivity, I've updated my configuration files, which ensures that the system is safe to operate when updating to newer versions of Marlin.

Holding this until after the 1.1.9 release, which could be pretty soon. Still a few bugs to look into.

I'm referring to bugs generally that I need to look into before releasing 1.1.9. Not bugs in this PR.

I think I'd rather see #if ENABLED(DELTA) and then remove it if someone can show it working safely on a lead screw driven system.

@teemuatlut : We have printers we are working on which use a belted Z axis with direct (non-geared) drive, so the assumption that if it is a Cartesian it has gearing and/or a lead-screw is soon-to-be incorrect. I may give this option a try and see if it works. My concern isn't with damage, but with false positives due to Z binding.

Just a heads up that due to #10294, the polarity of DIAG1 will change. Not sure whether this necessitate any corrections to this PR, but since it was related, I thought I would mention it.

@tcm0116 I saw your video. It's nice to use stallguard to get rid of endstops. It's great to make your printer quieter. But using stallguard as a bed probe is like next level.
If it's accurate it would be awesome.
I'd like to try it out and help get it working on another Delta?
Do I simply bridge xyz diag pins together as one output then connect that to zmin?
Not totally sure what changes need to be made on firmware side, but I'd think the firmware would see signal from any diag pin as a probe hit? So I don't think any firmware changes necessary as far stallguard as probe are necessary? Just the bridging of daig pins? Is that correct?

Hope this feature gets into 1.1.x as well, please, please.

Current configurations claim that StallGuard is only usable with TMC2130 drivers. But this PR claims it can be used also with TMC2208 drivers. Is the current configuration wrong?

Is the current configuration wrong?

No. I was mistaken. I've updated the configuration files and re-based.

I've also done a bit of tweaking, rebased, and pushed changes. Did you change anything other than the comments reading // TMC2130/TMC2208 only…

…and the sanity checks?

Ok, should be good to merge once it passes Travis CI. Previously none of the TMC2130 or 2208 tests were being run, so I want to make sure they're all passing now.

…and the sanity checks?

The sanity checks were all correct. It was just the comments that were wrong.

This error message is correct? How do we use it with TMC2208?

#error "SENSORLESS_PROBING requires TMC2130 or TCM2208 drivers for X, Y, and Z."

Also, oops, a little dyslexia fell in with "TCM2208."

hello, @tcm0116, can you give a quick explanation on how you wired the pins?
i would like to give this a try but can't figure out the wiring.

@Palatis - in my setup, I'm using conventional endstops for homing the three towers, which are wired to the X/Y/Z Max inputs on my board. I then have the diag1 output of the TMC2130s wired to the X/Y/Z Min inputs on the board. Marlin is configured with all six of the endstops enabled and SENSORLESS_PROBING enabled. G30 and M914 can be used to determine the proper sensitivity settings for the TMC2130s. I think I'm running the slow probe speed at 40 mm/s and a stall sensitivity of 4 or 5.