ABL and accuracy, which probe to choose?

General Discussion
1

I’m in search of a probe with high accuracy to use with ABL.

What i have tried is:

  • inductive probe (5€), has shown an accuracy of appx. 10 micron
  • CR-Touch (50€), has shown an accuracy of appx. 3.5 micron

On the one hand I like the accuracy of the CR-Touch and on the other hand the ‘no moving parts’ of the inductive probe, since the CR-Touch emits awful rattling noise on movements above 80mm/sec.

So, what have you running? What are your insights and recommendations, to make up a reliable and accurate ABL?

2

I came from an inductive probe (SupCRprobe) as you might remember but now use the BD sensor (~30 €), kind of a clone of the popular Beacon3D but without the metal free zone.

image
image1920×1479 105 KB

The bigger electronics case needs to be mounted nearby.

It is easy to install yet not natively supported by Klipper. This results in a dirty Klipper when being patched but is reversible easily.

If you properly measure your probing points and bed joints ABL works fine!
The sensor supports nonstop bed meshing and now seems to have a strain gauge mode in the latest firmware.

It can be directly connected to the used EBB36 that runs via USB on my current setup.
So no extra cables.

3

Hi Brian,

yes I have seen your topic about its behavior with the PROBE_ACCURACY command.

Does it keep to have range 0.000000? That looks a bit strange to me. How does it work in practice, with Z_TILT for example?

splitting the bigger electronics case from the probe itself seems to be a nice idea, since it makes positioning of the probe much easier. How long is the cable? Would be quite nice if ‘nearby’ means apart from the head.

How long is the triggering distance? If I look at your setup, it might be about 10mm?

4

Check the github page:

On AliExpress there is a photo with measurements and if my calculation is not wrong the cable should be about 82 mm long. Did not measure it before installing the sensor.

After I fixed my wrong offsets yesterday the first layer was perfect. Z_tilt is then done with one or two iterations.

Trigger distance is within 4 mm but you should mount it as close to the nozzle as possible and lower is better.

I do not want to advertise this sensor but I just installed it and it seems to be fine as an all in one solution for Z probing and bed meshing.

However in nonstop mode my bed looks terrible. Will verify this in normal mode…

5

This is actually a very good question and also very hard to answer.

To be a bit provocative:
No probe is good, the best option would be a printer that is mechanically so capable to not needing one.
Unfortunately this would make them so immensely expensive that it is no fun as well.

What would be THE BEST probe:

  • High accuracy / repeatability
  • No offsets to the nozzle
  • Fast
  • Not depending on bed’s surface
  • Actually probing the surface and not the underlying steel sheet
  • Fully temperature stable
  • Able to probe with fully heated nozzle and bed
  • Low weight
  • Easy to retrofit
  • Low complexity / high robustness
  • Low price
  • Fully supported in Klipper

Looking at todays alternatives and

  • Based on my experience or personal judgement
  • Without claiming to have them all scientifically tested
  • Without claiming that it is a comprehensive and final picture

my view is:

image
image1268×779 23.7 KB

I invite everyone to amend / change / re-evaluate this view. Excel attached
ABL_Probes.zip (7.4 KB)

6 Likes
6

…just to underline the poor accuracy/repeatability of inductive probes:

17:53 probe accuracy results: maximum 0.842812, minimum 0.794063, range 0.048750, average 0.817894, median 0.814063, standard deviation 0.015853
17:51 PROBE_ACCURACY at X:335.000 Y:360.000 Z:1.927 (samples=100 retract=1.000 speed=4.0 lift_speed=20.0)

Increased microstepping on Z-axis from 32 to 64 shows:

18:04 probe accuracy results: maximum 0.834062, minimum 0.818281, range 0.015781, average 0.826381, median 0.826562, standard deviation 0.004796

With further increase of microstepping from 64 to 128 no nameable increase of accuracy is visible:

18:16 probe accuracy results: maximum 0.846641, minimum 0.832656, range 0.013984, average 0.841155, median 0.841094, standard deviation 0.002355

So I’ll stick with 64, since that seems to be the best compromise with my setup.

You can find the used probe here. It was cheap and sufficient to learn using them (have mine running at 24V), but don’t expect better accuracy than 15µm. Since I have five of them I’ll try to find the best one later on.

7

Good table. I’d make some comments about it based on my experience.

One comment that would make it more useful; it would be nice to have some numbers in there for the baseline of the various options. For example, I’d like to get not sure what you mean by “speed”. You might also want to add the category, “Ease of Installation”.

I have a fair amount of personal experience with BL Touch (and clones), some experience with Klicky and probably above average experience with inductive probes and I’d like to offer my observations.

First off, I found Klicky to be only suitable for masochists and not all that extrermely accurate. It requires some solid engineering to make sure it is solidly mounted to the toolhead, is pretty fiddly to set up and get the macros tuned right. I did use the Omron D2F-5 (and L, removing the lever) and the (mean) accuracy I had was 8μm. I had something of a difficult time locating the recommended magnets from sources other than the recommended kits. Now, this was done a year plus ago on my Voron 2.4 so maybe things are better now but for me, it was a pretty negative experience. If you’re buying a kit, expect to pay $10 USD (but you have some 3D printing to do).

As for BL Touch and clones, I’ve worked with a number of them over the years with my most accurate one being a metal tipped clone that delivered a consistent 2μm accuracy. I do have a CR Touch that had 10μm accuracy. For a genuine BL Touch (and I’ve owned five of them), I see between 4μm and 8μm accuracy. I’ve never seen the claimed accuracy/standard deviation of sub 1μm. Cost for clones can be as low as $15USD and $50USD for genuine Antclabs or Creality units.

As for inductive probes, you need to do some searching and experimenting but there are some excellent ones out there. I started with the Omron TL-Q5MC2-Z on my Voron 2.4 and it provided 10μm accuracy - it’s crazy expensive ($100USD) if you can find one, somewhat large and you have to provide a protection diode on the sense line as it only works at 12V+. I have used some 12mm ones as they work at 5V and other than size, they’re pretty good with less than 5μm accuracy with a 4mm detection distance. I have tried three Super Pindas and been reasonably happy with them; they do achieve the advertised 3.5μm accuracy - they are somewhat expensive and they’ll run you $25+ USD and only have a 2mm detection distance (I use a jig to make sure they’re 0.75mm above the nozzle to give them an adequate detecting distance).

The best inductive probes I’ve found and am moving all my printers to are:

They’re shorter than the Super Pindas and run on 5V without complaint. They have the same 2mm detection distance as the Super Pindas so you’ll need a jig with them as well. Accuracy is sub 2μm. Nice price on them.

Yes, I’ve probably spent too much time (and money) playing around with Z-Axis sensors but with what I ended up with, I’m getting excellent, consistent results.

3 Likes
8

Thanks for your comments @mykepredko

To be honest, it wasn’t my intention to dig too deep into this special matter. It was just meant as a quick and dirty way to convey my view.
I also do not feel well-equipped to drive it much further, since I definitively lack the experience with quite some of the options and as such it is my “outside view” or personal judgement.

Speed is referring to the probing speed, e.g. a comparison between the quite sluggish BLTouch types and blazing fast Beacon style probes.

For me “Ease of Installation” would have been covered in the heading “Retrofitability”, i.e. how easy / convenient / straight forward is the deployment of such a probe.

As I said, this is certainly a topic that can be argued about and certainly offers a very wide range depending on personal experience.

To wit, I have not made any satisfying experiences with inductive probes but I also gave up quite quickly and never tested too many types (2 cheap ones I think?)

1 Like
9

Interesting. How did you get it running with Klipper?

10

I simply used a [probe] statement:

[probe]
pin: toolhead:TOOLHEAD_ENDSTOP1
x_offset: 0
y_offset: 9
#z_offset: -2
speed: 3

It’s wired to the endstop connector on an EBB42. Wiring is +5V, Gnd and Open Drain Output.

I don’t feel like this is what you’re asking though.

1 Like
11

You don’t have an aluminum bed, do you? I’m just asking because my probe should have 4mm detection distance but already has only 2mm due to the fact that I have an aluminum bed. I’ve somewhere read that this is normal behavior for inductive probes used on aluminum, but don’t remember where. So, if those 2mm are shrinking by half also, it doesn’t seem to be feasible to use that with an aluminum bed.

12

I use stainless steel PEI build surfaces.

13

I’m currently using a BIQU Microprobe. It has a retractable probing pin like the CR-Touch - but when sensing it does not depress. It is allegedly more precise than a CR-Touch. When going at 150mm/s at 10000 mm/s*2 it does not rattle… It uses the same wiring harness as a CR-Touch but is smaller.

1 Like
14

No, perfect :+1:

Did you do some deviation checks?

Is it ok to put 5V to the STM32 MCU’s input pins?

15

This is a rat hole I’m familiar with :sweat_smile:

My Probing Journey

I started printing on a Prusa MK3 which has consistently good first layer performance. Since then I have been on a quest to replicate that in other printers.

I had a terrible time with the switch probe in the E3D Tool Changer. I had no good luck with the stock Voron 2.4 inductive probe. It just wasn’t consistent enough. I tried Clicky and then Euclid. These are better, especially with higher end microswitches.

Then I did a custom mount for the Euclid that attaches to the ADXL mounting point so the switch nub could be horizontally in line with the nozzle. This helped a LOT. It’s no coincidence that this is how Prusa mounts their inductive probes: off to the side, in line with the nozzle. A lot of printers have this design flaw, if the probe is not horizontally aligned with the nozzle its not going to work well, no matter how good the results of PROBE_ACCURACY are. If your printer has this issue I would look at options to that re-locate the probe.

I never did get a Tap mounted on my Voron. I would assume that Tap requires an offset. It doesn’t trigger on contact, the toolhead has to slide up to break the beam or click the switch. That puts the trigger point lower than the contact point. But Tap solves the Voron’s probe offset problem, so it should outperform anything you can mount to the stock location.

The other thing I did to improve probing consistency was build the interruptible heat soak macro. This is by far the largest time in my print start routine now. If I really wanted the Voron 2.4 to start fast I would have to go to a PCB heater bed.

Load Cells

My hope is that Load Cells work out to offer the best performance blend, i.e. the smallest range and the least fuss.

The temperature stability & probing while hot are common issues for all nozzle contact probes. I have a change for z_thermal_adjust that would let you compensate for the thermal expansion in different parts of the printer by configuring multiple z_thermal_adjust sections.

Having independent thermal compensation of the hot side, cold side (The nextruder has a thermistor there) and the frame should result in a printer that can switch temps and configurations while keeping consistent first layer performance. i.e. PLA @ 210C, door open to PC @ 285C and door shut. Most people don’t attempt those kinds of back-to-back prints. Maybe they “have a printer for ABS”. I take that to mean they would have to tweak their offset because there is some uncompensated thermal expansion.

As for speed, I think people are most concerned about bed mesh times. If the repeatability is good enough you don’t need to average multiple probes, this saves a LOT of time.

2 Likes
16

Just did it on one of my printers. Custom CoreXY cold with a TriangleLab probe:

2:49 PM
probe accuracy results: maximum 0.729932, minimum 0.727435, range 0.002497, average 0.728185, median 0.728685, standard deviation 0.001500

This is pretty standard for what I’ve been seeing with the 8mm inductive probes.

Sorry, I don’t have any of the 12mm or Omron TL-Q5MC2-Z probes set up right now.

There are no internal pull ups on the 8mm inductive probes I’m using so there is no danger of overloading a pin.

Regardless, I just looked at the schematics for the EBB42 I’m connecting the inductive probe to and when I look at the Endstop socket, I see:

image
image915×561 17.5 KB

There’s a voltage divider which provides a current limiting resistor as a measure of protection for the STM32 pin in an overvoltage situation.

There is similar protection on the BLtouch “PROBE” pin:

image

Just out of curiosity, I took a look around and:

  1. The Manta M8P has a similar BL Touch circuit as the EBB42 with the current limiting resistor going to 3.3V on PROBE. They do NOT have the 5 pin endstop connector of the EBB42.
  2. The Octopus and Octopus Pro do NOT have the current limiting resistor going to 3.3V on PROBE. They do NOT have the 5 pin endstop connector of the EBB42.

I could look around at other boards, but these are the ones that I have schematics for at hand. This is probably something to consider when selecting a main controller board if you expect to use a BL Touch.

2 Likes
17

I’ve seen the BIQU Microprobe on sale at Ali Express for $15 USD … it’s very tempting.

I recently bought a Geeetech BLTouch knockoff for $10 USD and an open box CR Touch from Creality for $20 USD. The build quality of the CR Touch is much better, IMO.

BTW, I accidentally destroyed the fake BL Touch by using Loctite on the M3 nuts. A small amount of Loctite got on the BL Touch and completely disintegrated the plastic case (many small fractures). Thus I learned that Loctite-type thread lockers can be deadly to many plastics.

18

Good to know about the :smile: Loctite. The quality feel of the Biqu seems very convincing. But to me the operation of it is quite convincing to look at when it operates. The pin extends and as soon as the tip touches it registers, it does not have the extra travel of the pin like the CR/BL-Touch. It’s also about half the height of a CR/BL-Touch so it’s easier to find a spot to mount it.

hotend
hotend1065×973 157 KB

19

Did you design that cooling setup? Thats pretty slick.

20

Although offtopic but yes that is my printhead design and @MFBS duct design: NoName print head with Gaping Maw fan duct and 4028 fan by Life_Of_Brian - Thingiverse

1 Like