Adjust Sensitivity without Duet Board

kcramp858

I was able to adjust the sensitivity.

I am having a hard time getting my delta level.

During my calibrate routine (it just taps the nozzle against the glass bed), my z values are simply not repeatable. I switched to Klipper and using the probe_accuracy command I get significantly different values every time.

I assumed it was because that the sensor was not sensitive enough.

I increased the sensitivity and I still can't get any repeatable results.

I seem to go from "activates reasonably, but not precise enough (+/- 0.1mm z height discrepancy)" to "skips across the bed and doesn't activate".

Is this a geometry issue?? I can't figure it out -- does the sensor just not work well for non-duet boards?

The best accuracy I was able to achieve was 0.08mm z discrepancy every tap, but that is still a big variance..

T3P3Tony

@kcramp858

What speed are you probing at.
where are you probing (start the center of the bed)
what do you have the sensitivity set to? - generally it should not be so low that it triggers during the probing move but its not an issue if it triggers during normal printing or travel moves.

dc42

@kcramp858 the repeatability will depend on various things:

How fast the firmware you are using responds to the probe triggering. If the latency between the probe triggering and the controller stopping movement is variable, then this will lead to a variation of (probing_speed * latency_variation) in the trigger height. I have no idea what the latency is in Klipper. If the motion is stopped by the controller board without the Pi being involved then it should be low. If the Pi is involved, it may be much higher.
The probing speed. The Smart Effector works best with a high probing speed, e.g. 20mm.sec, however higher probing speed will also increase the effect of latency as already described.
Filament stuck to the nozzle reduces reproducibility. We recommend probing with the nozzle hot.
Stickiness and backlash in the mechanics will of course also reduce reproducibility.
Hard ned surfaces give better reproducibility than soft bed surfaces.

On my delta, I achieve reproducibility of about 0.02mm using the standard sensitivity, as long as the nozzle is clean.

kcramp858

@dc42 Thanks for the reply --

Current plan is to eliminate the latency factor by dumping Octoprint and using a command line based version -- with Klipper, no processing is done by the board, it is all done by the rPi, so this should not be an issue but I will test it anyway.

Mechanics shouldn't be an issue -- I am using the brand new sets of the following:
https://www.seemecnc.com/products/290mm-ball-cup-arm-set-of-6pcs?_pos=4&_sid=198d4b9cb&_ss=r
https://www.seemecnc.com/products/injection-molded-cheapskate-carriage-set?_pos=14&_sid=198d4b9cb&_ss=r

I suppose that I am using 10 year old belts -- I will look into replacing them if they are cheap enough.

Probing speed -- I was testing the probing speeds to find repeatability and I found the best test combo that I got was 30mm/s with a 15mm lift after each tap.
I am doing all of this testing cold, but I have not yet run filament through this nozzle yet, so it is clean (a new e3d v6 hotend).
I am using a glass bed.

I will report back after removing octoprint and trying to use a command line based controller.

@T3P3Tony
For the sensitivity, I switched it to 35 and have been trying to move slowly to avoid triggering -- it does not work well. I tried 45 and had better repeatability than the default 50, but I still wasn't getting good enough reproducibility (supposedly it is recommended to have 0.025mm or below for accurate bed leveling).

I am doing the initial testing in the center of the bed, but the skipping occurs the further away I get from the center. It calibrates in a mesh-like fashion, tapping ~50mm out in every direction. I have significantly more issues the further I get from the center of the bed. Skipping tends to occur as I get closer to the A tower rather than the B or C tower.

I did check the squareness of the towers to the heated bed PCB and it appears to be square. The glass plate is not as square, but I thought that was the point of a sensor like this -- account for the variances in the glass plate height.

I don't know -- maybe I am missing something extremely basic here.

T3P3Tony

@kcramp858 said in Adjust Sensitivity without Duet Board:

I am doing the initial testing in the center of the bed, but the skipping occurs the further away I get from the center.

start by getting a repeatable sensitivity set at the center and then once that is done you can look at why its triggering at the wrong points further way from the center. (then you will know that the probe is working correctly but there is something in the mechanics that causing the trigger).

dc42

@kcramp858 said in Adjust Sensitivity without Duet Board:

Current plan is to eliminate the latency factor by dumping Octoprint and using a command line based version -- with Klipper, no processing is done by the board, it is all done by the rPi, so this should not be an issue but I will test it anyway.

I very much doubt that Octoprint is involved in bed probing, apart from sending the command. I don't know if Klipper on the Pi is responsible for terminating movement when the probe is triggered, or not. If it is, that could be a problem, because the Pi is not running a real-time operating system, and therefore the latency may very depending on what else is running on the Pi. However, it may be that Klipper tells the Duet that it should monitor the probe and stop movement autonomously when it triggers.

kcramp858

@dc42 I switched to the command line version and I did get down to a range of 0.03 ONE time (set of 10 taps), but I am not sure if this is coincidental or based on environmental factors. I repeated the test in the same spot and got much higher.

I attempted to use the same settings (30mm/s and from 15mm height) at X-75 Y0 Z15 and the nozzle just keeps skipping and it ruins the test.

I tried to make it more sensitive (went to 45), but that just causes premature triggering during movement.

It seems like 30mm/s at 15mm height works..passably.. for the center, but not for anywhere else outside the center.

I am not sure how to proceed from here. I can try to increase the sensitivity to ~70? and then increase the speed, but I will really be smacking my glass bed at that point.

I ordered new belts just in case, they arrive tomorrow.

T3P3Tony

@kcramp858 said in Adjust Sensitivity without Duet Board:

causes premature triggering during movement

Is that all movement or just the probing move? potentially slow down the probing move and ignore triggers outside of the probing move?

kcramp858

@t3p3tony Looking into ignoring triggers outside of probing moves..

I lowered the non-probe speed to 10 and it seems OK.

As I said before, probe speed of 30 in the center works great... not so great on the edges

When I up to 40 or 50mm/s in the edges, it looks like the acceleration is causing the probe to trigger.. Feels like im stuck between a rock and a hard place, I need to go faster but it keeps triggering early when I go faster.

I have noticed that there must be some sort of resonance going on (when testing in the center), because at 35mm/s the steppers are super loud and it triggers almost immediately, but at 30mm/s it works fine and at 40mm/s it works fine. Very odd.

When I use too high of a speed, I think the deflection of the actual effector is causing bad z-readings, which means that it needs to be more sensitive and trigger earlier... but when you increase the sensitivity, it false triggers. ugh.

dc42

@kcramp858 said in Adjust Sensitivity without Duet Board:

When I up to 40 or 50mm/s in the edges, it looks like the acceleration is causing the probe to trigger.. Feels like im stuck between a rock and a hard place, I need to go faster but it keeps triggering early when I go faster.

Triggering during non-probing moves is common and shouldn't be an issue. If it's triggering during the acceleration of a probing move, can you reduce the acceleration? RepRapFirmware automatically reduces acceleration for probing moves.