Calibration Part III - the quest for dimensional accuracy/precision models

Continuing the discussion from More Calibration:

Firstly, many thanks to @ChristopherBarr for the great ideas and parts (and also thanks to @Formlabs1 customer support for pointing out the first thread to me…); @JoshK deserves a mention again, no doubt, although it’s no longer very clear to me, I’m afraid, exactly what part he played… but he seems to be active and helpful, so I guess that’s as good a reason as any :smile: !

Anyway, I’m an engineer and looking for precision of at least +/-0,2mm, preferably +/-0,1mm or better - up until now I’ve been getting widely varying results from pretty much spot on to nearly 2mm off, depending upon orientation and part, with a definite tendency for x and y to be too small and the z to be too big. It’s early days, of course, having only just performed Christopher’s calibration procedure, but everything he’s written really makes sense, so here’s hoping.

Here’s my print of his ‘calibration steps’ - I found I only needed to cut away small sections of the supports on 2 sides to be able to access with the calipers, which was not much effort at all, so I think he hit a sweet spot with that; additionally, there’s no real need for marking x and y, since one side has no supports:

(guys, I trust I got X and Y the right way round…?)

Here are the values I got from measuring the steps:

One thing I had already noticed, however, is that even when a part is close to the designed model, external measurements are a touch too big, and internal measurements are a touch too small (i.e. walls are too thick)… I’m assuming this is simply because the laser dot has a finite size, and so cures a touch more material at edges than it ideally would. I therefore decided to make a slightly different cube to allow me to check external AND internal measurements; additionally, I rounded all measurements of the calibration steps down a fraction… this gave me, on average, a 20,1 mm external dimension for my cube, and a 15,9 for internal measurements, i.e. I get 2,2mm walls all round (give or take a few hundredths) instead of the 2mm from the original design - here’s a comparison of both cubes:

So, if I want 2 parts with tight tolerances to fit together, I’m figuring I’ll have to choose the ‘main’ part, print this with the 0,991 factor the excel file gives me, and then adjust that factor down to about 0,98 for the ‘sub’ parts that have to fit to that.

All in all, I think a great printer just got a bit better… I guess the real test will be with some larger mating parts, which I should be printing in the next few days - I’ll post an update when I get to that point.

In case anyone else would like to use my cube, here’s the .STL file:

2cm_confirmation_cube-2015_07_11.STL (20.0 KB)

I’ll include the .FORM file too, although obviously this only helps if you’re using the same 0,991 factor as I am:

hollow-2cm-cube_plus_Sols.form (346.9 KB)

Look forward to hearing some feedback,
Sol

P.S. Has anyone tried a comparison between prints done at 50µ and 100µ with the same material, or at the same resolution with 2 different types of material? I would be very interested to see if there is a significant difference between materials and/or different resolutions… I would tend to expect greater errors in the z axis at finer resolutions, since the stepper motors are performing twice or four times as many movements, but I’m not 100% sure if this is even a valid expectation…?

4 Likes

Hi @Seagull,
Glad to see @ChristopherBarr’s calibration stair step getting more use.
For what it’s worth, @JoshK’s original Form1+ Scale Test.form is a good, fast, simple “sanity check” for calibration testing on the XY and for object placement on the build platform. (Along with @Steve_Johnstone’s spreadsheet I think it forms the basis of Christopher’s tools).

Just like you, I need the Form1+ to have as much dimensional accuracy as possible. This test is a good way of dialing in those numbers, Especially helpful is Christopher’s inclusion of the Z axis.

I haven’t made a second printout of the stair step, which really is necessary for verification of improvement. (It’s on my to-do list. I’m also avoiding a higher resolution printout. When I start my next round of testing, I’ll add your improved 20mm cube to my test scenario.

If it’s reassuring, your numbers were very similar to mine and to folks posting in the earlier thread.

-JD Davison
The Lab Partners

P.S. Yes, you’ve got the X and Y correct. :slight_smile:

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.