Kossel Delta FSR G33 P2/P3 Autocalibration probe failures

Hey folks.

I have a hand-me-down delta style printer. I believe it to be a Kossel 250 from Ultibots.

RAMPs 1.4 on Arduino Mega 2560 running Marlin 2.0.6

Bed heating control broke, apparently, so the heating leads have been removed.

I had a terrible time trying to manually calibrate this thing, but got pretty close. So many guides out there refer to using DELTA_SMOOTH_ROD_OFFSET for fine tuning, but best I can tell this is deprecated in recent builds of Marlin?

Anyhow, I was able to finish a half-decent Benchy, then put in an Ultibots FSR kit for autocalibration.

It would not finish a full `G33` pass without probe failure. Sometimes the effector would seem to drift too high when going from the X to the Y tower. I started down this path when I tried to print items that were within the supposed 200mm print diameter, but I'd end up printing in the air half the time.

Now, `G33 P0` and `G33 P1` pass every time. Sometimes I can get `G33 P2` to pass, which I believe is just probing center, and the 3 towers. I have only been able to get it to pass P2 at all with my printable radius set to 80mm rather than 100m as desired. `G33 P3` has never passed, this is when it does a triad of probes at the center, then starts probing in a circular manner counterclockwise around the bed. It usually starts floating upwards and failing shortly after the X tower.

Here's my fork in GitHub with a diff against the default Configuration.h (which is cartesian, but still highlights my changes) [github.com]

Here's the output from the terminal, sans temperature updates: [gist.github.com]
In short, I connected, ran M503 to show delta settings etc, then went through the G33 commands to failure.

Come to think of it, occasionally when attempting to home after a successful print, it will also fail and require reset. None of my prints ever get anywhere near the top limits. I don't run EEPROM, and I leave software endstops enabled at all times.

I could also upload some video to Youtube later, shots of Benchy or other recent failed prints if that would be helpful.

Thanks!

For what it's worth, I plan to swap in the Azteeg x5 mini board and try running Smoothie on it, since the previous owner sent that gear along with all of this.

I figured I should ask while I'm still set up on a (possibly) more familiar configuration, just to hopefully get to the bottom of the issue before changing another variable.

I'm not quite to the point of saying "this isn't worth it" and shelving the whole thing to shop for a more current and common printer, but I'm getting close

I have had nothing but problems with an FSR and ditched it for a piezo sensor. Now I'm convinced that piezo sensors are among the best options for bed leveling. However, I also lean towards inductive probes still. The downside of a piezo sensor is that when the hotend is heating, filament will drip out, and a little bit of filament coming out of the nozzle can mess up a reading because the piezo discs are very sensitive.

An FSR sensor on the other hand needs too much pressure imo and makes it so that the whole carriage on the Delta is pushed into the bed too hard which causes the effector to tilt slightly and give skewed readings.

I now did the G33 sequence only once to get a good reading, after that a G29 to get a complete surface mesh model of the bed and now at every beginning of a print use 'G29 J'. This makes it so that the machine probes three points in a triangle and it levels the mesh model to the measured points, always giving the best flat first layer.

Ever since I got to fix the Delta machine I'm in love with it. It has some surface artifacts (more than my crappy cartesian model) but the speed and ease of use when dialed in are just superb.

---

http://www.marinusdebeer.nl/

Yeah, I'd heard folks complain that FSR was too "fiddly" ... definitely feeling that right now.

Even with my "successful" G33 P2 runs (more like, they didn't fail), the effector was still _on the glass_ and I needed to raise the DELTA_HEIGHT 1mm so there was room for extrusion. Otherwise it would push the bowden tube out of either end.

There's supposedly some jumpers on this board I can set for higher sensitivity, as to the effect above, I think the head was needing to push TOO hard to fully trigger the Z min. I might try that next.

Otherwise I had printed one of these, but didn't want to mess with probe offset measurements if I didn't have to [www.thingiverse.com] ... that and I'm not entirely clear on how I'd wire that up to RAMPs just yet. I need a high power magnifying glass to even see which pins were C and NC etc. [www.amazon.com]

Most of the decent guides for Kossel deltas I can find assume you're adding BLTouch. Maybe I'll try that eventually. If it works, great, if not, maybe I'll use it on a Creality printer if I go the bargain/hacker route rather than Prusa.

I'm not trying to dismiss the value of delta printers here. It's been fun and fascinating to work on, and a happy diversion back into technical problem solving after moving to more "management"-ish work from software dev (which may have been a mistake )

But I _am_ trying to give myself a clear line where I eventually just call this a "learning experience" and see what else I can do with the parts. 1 more 700mm vrail and maybe I could build a decent simple posterboard-sized CNC/foamcutter/drawbot. You know, so I can face come completely new and different problems with similar levels of frustration!?

Thanks for the feedback so far.