Do 3D printers really reach 50 micron (0.05 mm) accuracy?

  • I've always been wondering about the actual accuracy of 3D printing devices. When looking for the perfect machine to buy, I looked at the speed, price, filaments supported etc, but also accuracy. I once asked somebody who could give me some advice on what to look at.

    One of the things I was told about was that many printers don't necessarily have that crazy precision of 0.05 mm (50 micron). Another person told me something different - he said most of those printers actually were capable of putting out 50 micron layer height. How is it really?

    Another thing is that the official slicers for those machines also claim that this precision is real, for instance the PrusaSlicer v2.0.

    There are many high-end, very expensive machines and even they sometimes claim their resolution is worse than 50 microns.

    Ultimaker has a 20-60 micron resolution for the Z layer height depending on the used nozzle size, I've used these small layer sizes a couple of times, products come out amazing, almost looks like an injected part but takes really long to print.

    It's worth noting the difference between precision and accuracy. Being able to move in ~50 micron increments doesn't necessarily mean that it will always end up within 50 microns of where it's supposed to be. If it advertises 50 micron precision but the steps are actually 48 microns on average, that means that trying to move 10cm will result in moving only 9.6cm—definitely not accurate.

  • There's not a simple answer to this question, or if there is, it's "no". However the situation is a lot more complicated. When printer specs cite accuracy like this, they're usually basing the claim on the nominal size of the smallest movements on each axis by one "microstep" of the stepper motors. There's a great article on Hackaday explaining the how this affects accuracy: How Accurate is Microstepping Really.

    At the mechanical positioning level - putting the print head where it needs to be to extrude the material with the desired accuracy - you have at least these factors limiting your accuracy:

    • Microsteps are generally spaced roughly monotonically between whole steps, but do not necessarily divide the whole step into even portions. How well they do is a matter of the stepper drivers your printer's controller board uses. Generally, microsteps are 1/16 of a step (although there are drivers with 1/8, 1/32, or even 1/256, maybe others too), so if you see a rated accuracy of 0.05 mm, a whole step, which might be the minimum you can get reliable accuracy from, is likely 0.8 mm.

    • Stepper motors are deflected slightly - up to 2 whole steps but less than one step is more likely if they're not overloaded - under load. So are belts. How much this affects you depends on the design of the printer and how much mass each axis is moving. Direct drive extruders are much worse in this regard. Delta printers are probably best in it.

    These can be mitigated somewhat, with tradeoffs, by using stepper motors with more steps per rotation, better stepper driver chips, reduction with gears, etc.

    On top of that, you also have extrusion and properties of the print material limiting your accuracy:

    • The extruder motor is subject to the same accuracy issues as the positioning ones. If you extrude too much or too little material anywhere, you'll necessarily have accuracy issues. You can compute them based on the cross-sectional area of filament, size of extruder gear, extruder motor step and microstep size, etc.

    • If the filament diameter is not perfectly consistent, you'll also extrude too much or too little material.

    • If material is not cooled or kept warm appropriately as it's extruded (this varies by material), it will sag, warp, or curl, ending up in a different place from where you wanted it.

    • The more you vary the ratio between nozzle/extrusion width and layer height from an ideal ratio, the more the shape of the extruded material paths will differ from the model you're trying to print. With thick layers especially they'll become rounded rather than near-flat along the walls.

    In theory, a lot of these issues probably could be mitigated a lot better than they are now just by better slicing - the logic that happens on a computer to convert the original 3D model into instructions for where to extrude material.

    With all that said, you can get pretty amazing accuracy still, especially with a good or well-tuned so-so printer. On my cheap Ender 3, after dealing with some issues now and then that made glaringly obvious problems, I can get dimensional accuracy within 0.1 mm in the X and Y directions, at least for some models. So I think it's very plausible that a better, or better-tuned, printer could get 0.05 mm accuracy.

    This only really addresses the precision of positioning, not of the printed part.

    @SeanHoulihane The second half of the answer explicitly covers the accuracy of printing, doesn't it? And it's trivially obvious (and axiomatic from control engineering) that it is not possible to print more accurately than you can position. So the accuracy of movement is the limit for printing precision - printing can only ever be worse than that, not better.

    Sorry, yes, you covered it (and the relevant factors), but I think your 'trivially obvious' is not at all obvious to people who want to ask about precision - and with today's technology it seems positioning is one of the smaller sources of end result error - making it a bad metric.

    Then then is the tolerance of the step angle. a 5% tolerance on a motor with 1.8° per step means that the rotor can move anywhere between 1.71° and 1.89°

    Nominal resolution of positioning being a "bad metric" is largely what I was getting at in this answer - it doesn't tell you whether positioning is actually that precise or accurate, nor whether other parts of printing will limit the accuracy in much larger ways.

    I'd love to see a widely agreed-upon metric involving dimensional accuracy of simple and complex shapes. See also: Dmitry's answer and the example of dice, which are a good "simple shape". I actually did this with 20 mm D20s, working out some mechanical problems in my printer, and now have 20.0 mm (fluctuating between 20.0 and 19.9 on exactly one) between all opposite faces as measured by my cheap 0.1 mm caliper.

License under CC-BY-SA with attribution

Content dated before 7/24/2021 11:53 AM