Need to discard the first probe - help me find the root cause

I have developed (and been running for some time) a variation of Klipper that discards the first probe in a sequence. This addresses an issue of unknown cause where the first probe is significantly different to all subsequent probes.

Generally I run this as discard the first probe and use the value of the second probe. Meaning only two probes per point.

The original PR is Richardjm/probe discard first by richardjm · Pull Request #5969 · Klipper3d/klipper ( - with an amount of dicussion and investigation.

As that PR was closed as inactive I re-raised it as Richardjm/probe discard first by richardjm · Pull Request #6095 · Klipper3d/klipper (

Now the underlying cause is still unknown but only manifests after an x/y move - i.e. if you probe once in a location and probe again only the very first probe is different. Wobbling the print head makes no difference to this result, neither does time. No matter how long the wait is the first probe is always off compared to the others.

What suggestions would people have for me to investigate? Or if you want to verify my findings that would also be marvellous.

I see this was first raised in 2018 - “Dismissable probing data points for meshes” so it’s far from a new phenomenon.

1 Like

Interesting phenomenon! I wasn’t aware the problem exists. Especially when you read Add samples_result option to [probe] last by Stephan3 · Pull Request #3828 · Klipper3d/klipper · GitHub. Different probes show the same impact, amazing! But I can understand koconnor Add samples_result option to [probe] last by Stephan3 · Pull Request #3828 · Klipper3d/klipper · GitHub.
Here might be the right place to get any further with the rood cause. I wished the community here would give some feedback.
Where can I find your file?

Kind regards and good luck, hcet14

Idea for testing: The default sample_retract_dist is 2mm. The default horizontal_move_z is 5mm. Has anyone confirmed whether the first sample deviation still occurs (or still occurs to the same extent) if these two values are the same? In other words, is there something about the distance of the initial dive that causes the probe measurement to differ from subsequent dives that originate from a different Z height?

If the phenomenon still occurs even then, I’d be curious to see if exaggerating the dive starting point has an effect on the deviation. For example, what if sample_retract_dist and horizontal_move_z were both set to 25mm? Aside from the fact that probing would take forever, does that have any effect on the values obtained during the probing sequence?

1 Like

I found my way here cause I am having a similar issue.

I think first thing to narrow it down is, lets look for commonalities in our setups to see if there is overlap that might point to the cause.

For example, My setup is…
Voron 2.4 350mm
Voron Tap probe, running at 24V connected to an SB2040 can board
Stealthburner setup with a Clockwork 2 extruder
Phaetus Rapido UHF with CHT .4 mm nozzle
I have a drag chain on my X axis, Off brand (I think it was Fystec but they just rebranded someone elses)
Using a spring steel PEI build plate over a magnetic sheet glued to a MIC-6 bed.

I’ll spend some time this weekend seeing if I can narrow down the cause on my setup.

I have noticed this also. I have a delta printer. It is most noticeable after an X,Y move. In fact I get a different (and stable) Z probe, depending on the direction that he printer approaches the probe position. I had to write a different way to probe the bed where I take 4 measurements (one from each approach direction) and average them (done with a gcode file and macros). I only use the first probe, because the subsequent probes are different and it would take a long time to get two probes that match. I attribute that to the pressure my probe takes to register. That pressure may take out some mechanical play that would not exist when printing. I dont know if it is truely related to your issue, but take it as another data point.

I have a possible theory but I’d need to check against other people.

Are you using a magnetic build plate? Spring steel PEI over a magnetic sheet?

I’m excited, but remember Add samples_result option to [probe] last by Stephan3 · Pull Request #3828 · Klipper3d/klipper · GitHub!

My first line of inquiry is… Checking if this is common amongst people using magnetic beds and spring steel build plates.

The magnetic sheet is not rigid and have some “give” to it. It’s possible the first probe is compressing the sheet under the plate and subsequent probes can’t compress it further (or minimally).

I noticed this yesterday when I was probing with my build plate off and not thinking about it. I was getting weird readings and then looked at my printer and noticed the dimples in the magnetic sheet and it clicked in my head.

This is speculation of course, I’ll have to find a way to test it.

Maybe it’s different with certain types of probes, but there’s no way a BLtouch would be capable of compressing or making a dimple in a magnetic sheet.

My delta printer has a rigid glass non-moving bed with a magnetic “rubber” sheet glued to it and a removable spring steel on top with a PEI 0.1mm sheet glued to that. My probe attaches to the nozzle. It is pressure sensitive, so requires the nozzle to press into a 0.16mm thick pressure sensitive layer and touch the bed to register. It does not dent anything as far as I can tell. Certainly nothing could be dented beyond the surface of the PEI sheet. I would not probe my bed without the spring steel surface covering the magnet. The magnet itself is too soft for that.

I had the same issue when running an inductive probe (I’ve sinced moved on to a tap probe), so it doesn’t have anything to do with compressing the magnetic sheet, at least on a voron 2.4

As best as I can tell, it’s largely an issue with belted Z printers, particularly when the reduction gears are also run with belts. Increasing tension there improved, but did not fix the behavior. I hear switching to worm Z drives fixes the issue, which I will be doing shortly

1 Like

That answers that then, was only a first hypothesis.

I’ll keep digging.

One of my printers has a lead-screw Z-axis and I’m seeing this as well, despite anti-backlash nut and carefully rigging it.

1 Like

Based on the variety of machines and probe types that have this problem, I would guess that it can come from a variety of causes. Even things like probing on a heated vs cold bed. A probe getting close to a heated bed might get warmer the longer it is close to the bed and introduce a bias in the second probe. Something with a pressure contact probe might cause a slight mechanical shift after the first probe. Belts could shift on their sprockets. Magnetic ball arms on a delta might have some stiction shifting from the first probe. It might be more telling to find a machine that does not shift after the first probe and figure out how it differs from other machines that do.

I have a voron trident that doesn’t have the issue. Which is what made me think it was a belt vs screw thing.

1 Like

I have a voron trident that doesn’t have the issue. Which is what made me think it was a belt vs screw thing.

How is it configured?

to put in an update from our discord convo earlier. It appears that this happens only after a XY move on printers with a Belted Z. I have it fairly consistently on a Voron 2.4 and Jamie kindly did some tests on his K3 earlier and got the below results.

I can only add one attachment per post so here is the second example.

that appears to be what is happening. Belted Z’s have the issue where as leadscrews dont

Can’t confirm this. See my comment here Need to discard the first probe - help me find the root cause - #14 by Sineos