3D Printer Auto Bed Leveling Mesh Visualizer | Hacker News Reader

3D Printer Auto Bed Leveling Mesh Visualizer

i.chillrain.com

93 pointsmacinjosh98 comments

Klipper's frontend (Mainsail) has this built in. https://docs.mainsail.xyz/

There's also Fluidd

https://docs.fluidd.xyz/

ABL has been recently improved with the Beacon 3D probe (https://beacon3d.com/), I haven't had a chance to try it myself yet, as my cheap induction probe has been working reliably ever since I mounted it.

But, the Beacon 3D along with mesh leveling and selective leveling area are my favorite recent innovations in 3D printing.

Does it work well? All I want is a reliable first layer, as that's what causes 99% of my failed prints.

Look in to the Prusa Mk4 which uses the nozzle itself with a built in load cell to probe the bed, providing perfect offsets every time!

Ah, damn, that's great! Not CoreXY, but I'm sure someone will put this on a CoreXY printer, and then no more bed leveling!

Oh someone has put this on a CoreXY: ;)

https://www.prusa3d.com/product/original-prusa-xl-3/

Ahaha XLent.

The Bambu Lab CoreXY printers have force sensors in the bed for leveling using the nozzle: https://wiki.bambulab.com/en/x1/maintenance/replace-the-heat...

Vorons also have a nozzle probing design: https://github.com/VoronDesign/Voron-Tap

Can they live-adjust Z to keep constant pressure? Otherwise they aren't much different from a BLTouch.

They measure a series of points to get a bed mesh, the same as other leveling sensors do. You shouldn't need to live-adjust z, because you've measured the bed surface with the nozzle tip.

With a BLTouch, you need to adjust the z-offset, because the nozzle tip and the BLTouch trigger height are not the same. Once that is calibrated, it should continue to work fine as long as the sensor is working properly.

I think it mainly comes down to the accuracy of the sensor. If your sensor gives you a +-0.03mm reading when probing the same spot, then you will have high and low spots in your bed mesh that you can't correct with z-offset.

This could explain your issues with the BLTouch you mentioned in another thread here. I have various brands of knockoff BLTouch sensors, and the worst one had a +-0.02mm range when measuring points, which made it impossible to get a good bed mesh. The one I replaced it with had about a +-0.005mm accuracy, which fixed my issues with the low and high spots.

I've measured the repeatability on mine, and it was pretty good, though I don't remember the number right now.

The pressure sensor on the hottend is superior to any other method, because it has much better resolution (though maybe not better than the Beacon), and it can adjust to print conditions in real time, which no other start-of-print sensor can do.

I'm not aware of any hotend sensors doing as you describe. The Prusa MK4 only uses it to generate a bed mesh the same as any other bed-leveling sensors. It does not do any live calibration while printing.

Really? That's very disappointing, if that's the case.

EDIT: I'm not sure, this article says it can detect spaghetti just from the pressure difference, for example:

https://all3dp.com/1/nextruder-prusa-xl-deep-dive/

It also hints at detection of the height while printing.

Without watching the video referenced in the article, I'm guessing this was based on what they theoretically could do, rather than something that they had actually implemented.

They currently only use the load cells for bed leveling: https://www.youtube.com/watch?v=F7ETSVu04ao&t=940s

I could be wrong, but I would expect that the load cell readings would be far too noisy in order to determine the height of the nozzle while printing. Even if they were accurate, different materials would also behave differently due to different viscosities.

To be fair they really shoved the Mk4 out the door (read up on the release schedule for Input Shaping if curious) so there is probably room for development in the firmware. Perhaps they will have live pressure response in a year or two if such a functionality is deemed beneficial.

You mean in response to pressure changes during print? I’m not sure but that would be nice!

I do have two BLTouch printers and lots of Prusas with the PINDA inductive sensor. What excites me about direct nozzle force sensing is that it eliminates the need to calibrate z height offset between the sensor and the nozzle. I don’t know if Bambu’s bed sensor is as good as Prusa’s nozzle sensor but anything that eliminates the mechanical stack up between nozzle and sensor is an improvement IMO.

Though I do realize much of my gripes come from the fact that the PINDA mount on the Prusa Mk3s is kinda weak and I’ve had several fail, requiring an annoying service job.

Yes, exactly! I'm very excited by it, if it works well, it'll be a revolution. Any other sensor that measures the bed is strictly worse, because it only does so once, rather than continuously. The Beacon does look pretty great, though.

Yes, I have one and it works exceptionally well. The only caveat is that it’s measuring the metal surface directly and not the actual physical surface (such as PEI), so you need to load different offsets if you swap between different material surfaces such as textured PEI, flat PEI or bare steel.

Oh, how does it do it? I thought it used light/laser to measure, if it's measuring the metal it might not be very accurate with my two layers of magnet and PEI sheet on top...

It generates a field using the coil measures the eddy currents in the material directly below it. As long as you have a spring steel sheet and don’t have those large discrete magnets (either lots of tiny round magnets or the sheet magnets) it works just fine.

It’s accurate to single digit micrometers (yes, actually).

Wow, that's impressive. Aren't most buildplates made of aluminium, though? Also I always worry that the PEI sheet over the plate will introduce inconsistencies, let alone that mine has an extra layer of magnet sheet on it (and, as you said, the magnet even makes the detection not work).

However, it seems very doable to have a Beacon for the bed scanning, and a BLTouch for getting the physical distance over the sheet plate.

Most PEI sheets are quite uniform so you don’t really have issues with that. As long as they’re applied to the build plate without bubbles they’re generally as flat as the plate surface.

Also, no, almost no build plates are aluminum. Modern printers usually use a silicone heater bonded to an aluminum frame, which either has magnets embedded in it or a sheet magnet applied on top of it, then a spring steel plate with some sort of PEI on top is what you print on. That’s what Prusa and Bambu use, for instance.

I don’t know why you’d want to use BLtouch to get the physical distance, since you calibrate the physical distance when calibrating the beacon sensor itself. Not a problem unless you frequently swap between materials of different thicknesses.

> Also, no, almost no build plates are aluminum.

Ah, I must be misremembering, thank you.

> Not a problem unless you frequently swap between materials of different thicknesses.

Yep, exactly, I have a PEI sheet (with a magnet) and it has a smooth and a rough surface, and I end up swapping between them sometimes. I guess it doesn't really make any difference, since they're both the same height (as they're two sides of the same sheet), but, as you say, my magnet sheet ruins the Beacon homing anyway.

Explicitly a magnet sheet does NOT ruin Beacon and is the preferred magnet type for sticking down build surfaces. The only problematic magnets are point magnets that are very strong, as the magnet field can penetrate the surface of the material being scanned and impact the readings (but only at that point).

It sounds like what you're describing is sheet magnet that is coated in PEI, but I highly highly doubt that is what you have. The magnet is attached to the frame, not the material that you print directly on.

> As long as you have a spring steel sheet and don’t have those large discrete magnets (either lots of tiny round magnets or the sheet magnets) it works just fine

Hmm, from this I understood that a sheet magnet will ruin Beacon.

I don't have a magnet that's coated in PEI, I have a magnet sheet on the bottom (stuck to the big bed plate), and, on top of that, a thin steel plate (the one you bend to pop the print off) coated with PEI on each side (smooth/rough PEI). If the magnet sheet under the steel plate doesn't mess with Beacon, I'm going to buy a Beacon right now.

See their FAQ documentation:

https://docs.beacon3d.com/faq/

What you’re describing is the ideal material stack up for beacon. Only thing you may need to do is load a different nozzle offset for the smooth vs rough side, since the thickness of the PEI is probably not the same.

I did have a look, all it says is:

> Magnetized rubber sheets have a high number of poles, and result in no detectable artifacts.

Which kind of sounds like what I want, but doesn't actually say "you'll be fine if you have a magnetized rubber sheet", "no detectable artifacts" might very well mean "the sensor doesn't detect your bed".

Thanks for the clarification, this sounds great.

Tons of voron and Rat Rig use beacon. They mostly run thick metal plate with magnet sheet on top. So you should be fine.

You can also join one of the many discord servers (Rat Rig probably has one) and ask there for details

I think I'll get one regardless, all the reviews seem amazing, thank you.

Title needs [2019]

It's something that is build in on most UIs now days.

I've also made something similar. It was originally for the Prusa Mini, but I've updated it with a beta that accepts an arbitrary array of probe points: https://bbbenji.github.io/PMSBLM/beta/

Octoprint has a plugin (Bed Mesh Visualizer) that does this.

So that is two layers of filler to create a perfect bed to print on?

~0.6mm, two layers of you use 0.3mm, people often use 0.2mm layer height though.

But I think mesh bed can handle this.

How does the printer compensate for precision loss do to movement? In industrial robotics, there is the concept of measuring the inprecissions in the machinery at any given point (direct measurement) and compensate that via variables in the robotic matrice. Is there something similar for printers?

I'm not sure. But in the end you are laying molten plastic on top of each other. Even with a perfect flat print surface you will have slightly uneven extrusion that will be noticeable on top down lighting.

Silver cubes with direct lighting https://github.com/lord-carlos/mk3-upgrade/blob/master/asset... Same cube with top down light: https://github.com/lord-carlos/mk3-upgrade/blob/master/asset...

What we have is Pressure Advanced. Where we stop extruding (as much) right before a corner. And we do that by doing some test prints with different values and eye balling it.

Also some firmwares like klipper have some math in it to compensate for heat expansion over time of the printers frame. Or something like that.

Most hobby printers do not compensate at all. There are input shaping to reduce vibration, and hacked together closed-loop upgrades, personally I haven't heard of any more sophisticated controls.

Prusa Link/Connect also has this.

Are you sure PrusaLink has it? I've an MK4 with latest firmware and I can only see the usual temperature graph and file browser, and that's it. Not that the bed visualizer would be so useful with the MK4: this printer (like the XL) really does not have leveling issues.

I think you're confusing the control screen with leveling mesh visualizer. Prusa Connect does not have this, Prusa Link is even more bare bones.

This is really cool!

Context?

A 3d printer bed is never actually flat, especially since 3d printers are cheaper and the tolerances are usually pretty loose compared to a super flat, precisely ground CNC bed for example.

One of the biggest issue/pain point in 3d printing for a while was to adjust the bed to make jt flat, or at least as flat as possible so that you can actually print without massive issues on the first layer. The process used to be manual, tedious and hit or miss (due to drift in the screws, or thermal expansion) etc.

Nowadays most new printers come with a sensor that probes the bed at multiple points (say in a grid of 4x4), then using for example a lagrange interpolation you can get a mesh that represents how flat your bed is. Then, using the visualisation you can either manually adjust the bed (with screws that rise or lower a certain corner), or even just let your printer compensate for the deviations.

That is, the printer moves the Z axis even when it is still printing on the same layer to make the print surface "virtually" flat (not always ideal if the Z axis stepper is trash, or if the deviation is too wide, or if there's too much vibration). That means that usually, the issue of bed levelling is mostly solved for newer printers!

Some printers also just use the nozzle itself as a probe, combined with a load cell. But those are usually on the higher end. The most common way is just a small probe of known size that sticks out and then calculates how much it extended before reaching the bed. (Some even use a pretty high res lidar to make sure the first layer is fine, even in some sub 900$ printers).

There are also issues like even if extremely flat, the way most printers mount the bed, it's going to bend measurably simply from expanding upon heating up and the distance between the mounting points not expanding the same amount.

As long as you always print with a hot bed, that effect doesn't matter - just calibrate it when hot.

You mean "As long as you always print with the same bed temperature"?

Because there is a difference how your bed looks from 60 C for PLA to 110C for ABS.

Just do the first layer at 110 for all materials. PLA actually adheres far better at 110C, but obviously you need to cool it down before building too many extra layers on top.

Slicers have a setting for "Bed temperature initial layer" for exactly this reason.

I find it easier and faster to just print PLA at 50 - 60 C bed and use probes for mesh and z level.

I never heard of anyone printing PLA with 110C first layer.

I usually start my PLA prints at ~80C for the first layer, dropping to 65C for the rest. I first tried it because I was playing with PEO plates that had an iridescent pattern I wanted to strongly embed in the layer, but it turns out to be a good general pattern overall. Really reduced the amount of support material I need on certain prints, which is a big plus.

I've never tried that. I've usually tried to not have the build plate temperature vary too much for fear that the plate might contract a bit if I cool it down and the print might pop off?

Temperature changes above the plastic transition temperature won't cause it to pop off. Only when you drop a reasonable amount below the plastic transition temperature will it pop off.

Or an inductive sensor: https://help.prusa3d.com/article/p-i-n-d-a-superpinda-sensor...

Yep, having that by default since I think 2018, is one of the reasons prusas have a reputation of being super easy to use. It's way better than even a bltouch. Nowadays it's a lot more common thankfully though I don't think other big players use inductive sensors.

Even prusa is now using a load cell that apparently will be used for much more than just bed levelling (the load cell could be able to tell if filament is stuck, because the motor will push without extrusion leading to a variation in load for example). Very exciting times in 3d printing honestly (and I say that as someone that is pretty "bearish" on 3d printing usually)

It's often not the bed that isn't flat, but that the rest of the printer itself moves and expands.

Yeah or the bed itself is fine, but it's on a bad bed support. That's why even a perfectly flat glass bed plate can be completely uneven since the structure itself is crooked! Considering that quite a few printers are self assembled too (with basically no torque specs, and users that have little means to build it accurately), it's just compounded even more.

3D printer made of Invar/Kovar next? :)

For a 3d printer the distance between the print bed and the nozzle when printing the first layer is very important. If the distance is too big the print doesn't stick to the bed and you end up with spaghetti. If the distance is too small either not enough plastic comes out so your print doesn't stick; no plastic at all comes out so your print doesn't stick; or the nozzle digs into the printbed, damaging it.

Since the thickness of layers in a 3d print is around 0.1-0.3mm the margin for error is rather tight. Since no one is going to invest in precision ground printbeds you need to get that precision by some other method. One of those methods is to probe the bed height at various points before printing, and driving the Z-axis to correct for that during the print.

The linked tool can be used to visualize how nonflat your printbed is given a set of such probes. The default example seems to have a ±0.25mm error, which isn't amazing but should be usable with mesh leveling and not at all without.

This is neat, today I learnt. Can an Ender 3 do mesh leveling?

Yes look for the cr touch. But make sure that you have a 16 bit main board. All v1.x boards don't support it

3D printers can do a variety of "bed mesh" leveling, which is where they probe the bed in a grid pattern and then can compensate for any deformities in the bed via software.

This ensures that the first layer (arguably the most important) for any print is perfectly flat and level. This visualizer takes the output of the bed mesh calibration sequence (where each point is a z-offset, usually from what the printer has defined as 0 based on the end stop) and graphs it so you can see your bed deformation.

Most Klipper frontend UIs already have support for this though

Some 3d printers have automatic bed leveling. Which is a process where the printer will touch the print bed in many places with a probe. This allows the printer to detect peaks and valleys, which it can then compensate for. The website above allows you to visualize that data.

Kinda neat to see a scaled up visualization of the vertical height differences of your printer bed.

https://marlinfw.org/docs/gcode/G029-ubl.html

IME, ABL is not that important. Personally I spray bed with a hairspray and tweak leveling screws while first layer is going on[1], that's just fine.

Old Marlin ABL also has bug/tendency to report erroneous tilt towards the last corner it measures, I suspect due to Z axis missing step from microstepping. Ultimately it was too finicky for me without resorting to closed-loop drivers, which also almost completely solved the leveling problem.

1: I know it is a big flat NO to touch a robotic equipment in operation, unless said robot is specifically designed to be almost completely incapable of applying any force to any parts of a human body, but I'm smart and I consider my printer a co-bot so it's okay(it's not okay. A better solution is needed).

ABL is not that important if you don't mind babysitting the printer. But it's an amazing quality of life improvement otherwise. Especially if you're printing stuff that needs precision.

Manually leveling an ender 3 was tolerable to me for about a month while learning 3d printing and figuring out if it was something I would be into long term, but over the course of a year I went from that, to trying various ABL sensors and bed replacements, to now a custom built Mercury One.1 with 3 point kinematic auto bed leveling, and it's such a night and day difference in terms of usability.

It physically levels the bed and automatically determines the z offset, so I don't have to do anything when using different print surfaces and when operating at different temperatures/in an enclosure. Heat expansion is pretty noticeable (I've destroyed pei plates with the nozzle by calculating z offset when cold but attempting to print when hot), and while sticking your hand under the bed of a printer printing PLA is not that dangerous, with ABS the bed is at 110C, not at all worth the risk, so autoleveling is very helpful.

It makes it so much easier to just submit a print job, forget about it and find it finished perfectly, without involving any finicky adhesives.

3D printing has a bifurcation in the hobby; some people enjoy it for the objects they can make, some people enjoy it for the printers themselves. I pass no judgement either way, just an observation. I saw the same thing in the photography world; some were gearheads, some couldn't care less, and were focused on the image.

As for me personally, I spent many entertaining, delightful, maddening and frustrating years playing with a frankendelta of bodged parts, broken firmware and hacked wiring. Learned a ton, don't regret any of it. But the moment I got an X1C, the Rostock has been relegated to Spiral Vase mode only. And I can't help but note my actual printing volume has increased dramatically when the process became less of a babysit.

Yeah, that I can agree with. However I'm very much in the "gearhead" camp :)

That's why I went with a DIY design. Similarly learned a ton, lately I'm finishing up a multi-material unit, then I'll start again by rebuilding and modding the disassembled Ender 3 to gift to a relative who has a slight interest in DIY stuff.

I'm with you, adding an auto bed leveler was the single biggest upgrade on my 3D printer, for the type and printing I do and the frequency I use, it's really just made it fire and forget (I'm sure I could improve it /even/ further) where as previously I felt I had to re-level after every print.

I haven't levelled my bed in months or maybe years... Sure, it was a pain to level it the first time, but once it's level it should stay level unless you disassemble the printer.

Thermal expansion (especially in long cycles) tends to warp beds.

Glass and aluminium do expand and might bend when heated, but they don't bend more when heated hundreds of times. Therefore as long as you calibrate it when hot, it should stay calibrated (unless you let the calibration screws vibrate+turn or something)..

I have a 10-year glass bed on my Prusa clone that would like a word with you :)

(I replaced it with a garolite sheet, am very happy with that)

Interesting. I don't think I've ever knowingly encountered heat expansion of the bed itself affecting the prints.

I print on glass and only occasionally have to do anything about bed leveling. No need to pry or pull and mess up the leveling when all you have to do is just wait a few minutes for the print to cool and it pops right off. Never printed ABS on glass though, so I can't say if it works as well for that. Autoleveling would be cool, but I have very little interest in it because I haven't needed it and it seems like an added point of failure.

I print working firearms and firearm parts. I follow the bed levelling method detailed in GP, and I have never had a gun failure due to lack of precision, which is critical in firearms. I also do this with Nylon and TPU, zero issues. Sure, ABL is nice, but I'd rather use the budget on a proper hotend and direct drive.

> ABL is nice, but I'd rather use the budget on a proper hotend and direct drive.

I agree with you, 5 years ago. But now days even cheaper china designed hotends can be decent. And direct drive is de facto standart.

ABL is cheap and convenient.

This is all true, but FOSSCAD world is a little different, and a good hotend makes much more difference than ABL. Microswiss is the minimum. I need perfect extrusion, not perfect alignment along a non-critical axis. For the hobbyist, you are correct.

Interesting. I recently saw a video where they used normal Nylon CF and PLA+ for the orca gun system. What makes FOSSCAD so different when it comes to hotend demand?

I disagree about the hobbyist comment. You can be a hobbyist and try to print PEEK and ULTEM, or you can be commercial and print nothing but PLA. I would say someone who does it commercially wants to spend less time on bed leveling.

the probe I use for ABL is Klicky PCB which cost 20 bucks. You can print it yourself and use a normal switch and magnets and probably get it for 5 bucks. I don't see why this is a either / or discussion. It's so cheap.

FOSSCAD is different because we're generally printing parts that function near the end of the filament's capabilities. When I'm printing a lower receiver, I need perfect temperature consistency, because uneven cooling will introduce internal tension to the part, reducing its overall strength slightly. That's not an issue for a coat hook, but it is an issue for a Glock.

I also need absolutely perfect extrusion, because a layer shift or missed infill section is a huge deal, not just aesthetically like for a normal print.

I'm not saying ABL is bad in any way, simply that I have higher priorities. I have an Ender with it, and it's nice to have for sure. It's just not what I'd prioritize.

I'm using "hobbyist" as shorthand for "normal 3d printer user", i.e not printing things that can explode.

I have twelve 3D printers and I bought my first machine, an Ultimaker 1, in 2011.

I managed fine without it but every advancement that means less fiddling is a huge quality of life improvement that I always strongly prefer. I got ABL for the first time by modding it in to a printer I had. I then modded it in to my other printers, and now would never consider or recommend a printer without ABL unless someone is very broke and can only afford the cheapest Ender 3.

The Prusa Mk4 improves upon older automated bed leveling techniques by using a load cell in the nozzle itself to measure exactly where the nozzle touches the bed, rather than using a sensor near the nozzle which then requires manual height calibration. I have five Prusa Mk3s but I really want the Mk4 in large part for that.

I am so bleeping tired of fiddling with my 3D printers, I just want them to work.

With proper bed leveling and the right bed temperature you don't need to bother with hairspray.

Having to resort to glue is often a sign that something else in your setup isn't tuned correctly. ABL is one thing that helps.

It also helps to clean the build plate from time to time too.

Over time grease from your fingers will make it onto the surface (no matter how careful you think you are). Some filaments also seem to leave behind an oily residue that builds up after a few prints.

I haven't done much research on what I should be using so I have a bottle of iso and a sponge that I use to wipe down the surface. (A microfiber cloth would probably be better).

Mine is a 300x300x600mm frankenstein gantry build loosely based on Tronxy X1, so that's indeed on myself. But Marlin ABL didn't seem worth it even before I'd done upgrades.

The idea of having to put anything on your bed and fiddling with adjustment knobs seems so wrong when there are plenty of printers out there that are good enough now where I can send a print to them and maybe check on them in 10 minutes to make sure the first layer didn't fail and be confident that they will complete. From the very first time I used it I realized that ABL is a must on any printer now.

Maybe if you don't care much about the budget.

I don't need a printer and only use it sparingly, so I chose an Ender v2 for like 250€. It has bed-leveling problems, but it's not enough of an inconvenience to be worth it to me to cough up 3x of the money for Prusa.

If all else fails I use a glue stick, like the GP. There's nothing wrong with that, it works.

> Maybe if you don't care much about the budget. > I don't need a printer and only use it sparingly, so I chose an Ender v2 for like 250€.

Now days you can get an Bambulab a1 mini for ~335 that has auto leveling, auto z offset, nozzle pressure sensor and can do a 14 minutes benchy.

You can also get a Prusa mini is ~2 times the money of your Ender v2.

They both have smaller printable areas and +40%/+100% is still a pretty big price difference.

Sure sure, but you can care about budget and get a printer with decent sensors for auto z and bed level.

I started with bed screws, and never want back to printer without probes or flexible bed sheet.

I struggled with the cheaply built printer that requires a lot of upkeep the first time like many (Tevo Tarantula in my case), and in the end I realized that between the upgrades to just make it work safely, and the time wasted just fiddling around with getting it to print made the Prusa well worth the money.

Buy the Ender if you want working on 3D Printers to be your hobby, if you want 3D printing to be your hobby, then buy something actually reliable. I would take a small build volume and high reliability over something larger every single time.

> if you want working on 3D Printers to be your hobby, if you want 3D printing to be your hobby

Those two go hand in hand for all hobbyists I know, including the Prusa owners. I don't struggle with Ender, actually I was surprised how good it is, since my reference point was much older (and much more expensive) printers.

Maybe I have lucked out on the build quality, I don't know.

ABLs are nice to have and printers should have them. But every so often, a manual bed level with a piece of paper goes a long way. Especially if you have a budget friendly printer.

I have a long history of woes with this stuff. I initially had a Wanhao that didn't have ABL, that worked fine as long as I didn't change anything. Then, I got a BLTouch, which made it worse because the Z distance would always be inconsistent somehow.

I realized that Marlin must have some sort of bug, because it claimed to be able to print on tilted beds with ABL, but mine just didn't. I upgraded to Marlin 2.x and that was fixed.

However, now I have this problem where the print lifts on certain regions of the bed. I have no idea what's up with that, I use one of those golden, textured PEI sheets and it works fine, except for these holes in areas.

I wish someone would put a pressure sensor on the nozzle so it would always know the exact distance in real time and could compensate for anything, but we seem to be far from that.

> However, now I have this problem where the print lifts on certain regions of the bed

If it's the edges of wide, flat prints, search for problems with "warping". Draughts and uneven heating can cause it, especially combined with poor adhesion to the bed.

It's not, it's a very odd issue in the middle of laying down a layer line, where a tiny mm raises/blisters, then in the next layer the raised part gets a bit longer, then a bit longer, until it becomes a big hole within twenty layers of this. It's very odd.

I had a similar problem with my old Rostock Max, and it turned out to be integer math errors in the delta>cartesian conversion step. FWIW.

I never got ABL to work with marlin, despite trying all possible configurations and versions, so I just did it the same way you do. With Klipper, it just worked when I first installed it. Never looked back.

> unless said robot is specifically designed to be almost completely incapable of applying any force to any parts of a human body

Be careful, but don’t beat yourself up over it. Many CNC milling shops regularly run jobs that cannot succeed without a human gripping the workpiece to control vibration.

The folks I really worry about are the waterjet cowboys. The garnet particles that lend a waterjet the power to cut titanium get embedded inches from the site of the wound, if that stream touches skin. Amputation one or two joints up from the wound is standard treatment.

I click print in my slicer and move on to do other things, casually looking at the camera feed a couple of times during the 1st layer

I can't agree with you on this. With bigger printer beds (I've got a 500mm x 500mm x 500mm machine) it's crucial to have ABL working, as the height of certain points of the print bed can be within a range of over 0.2mm, which, as I'm sure you know, is the thickness of a layer a lot of times.

With simpler/smaller printers, sure, you can get away without an ABL probe.

Reply on news.ycombinator.com