Trying To Anticipate Printing/dimensional Errors In Fud (i Give Up)

Just received my third batch of FUD prints in trying to nail down/predict the error (mm of error per mm of distance) for FUD and I'm pretty much ready to call it quits...

I build/sell custom HO scale narrow gauge locomotive mechanisms. For the longest time, I was laser cutting the mechanism (frame/gearbox) parts out of brass and steel (I work in the R&D lab for a major industrial laser mfg.). Discovered 3D printing a few years ago and have been using it to model/produce parts since as it is much easier than fabricating the parts from metal.

Started with 3D printing the actual gearboxes and frames a year or so ago. Started off OK, but soon realized that the finished dimensions of my prints in FUD were a little on the small side, which resulted in tight gear mesh in the gearboxes. To combat this, I just added an additional 0.05mm or so between the gear shafts and this seemed to take care of the problem. These early gearbox/frame prints used laser cut plates to space the axles accurately, to match the the side rods that drive the non-geared axles.

Wanting to try to cut a step out of my production, I started printing the mechanism blocks without the laser cut frame plate provision, thinking that the error was not significant enough to be a problem. Soon realized that this was wrong, but figured I could counter the issue by printing up a chassis block, measuring it and comparing that to the solid model to see what was going on.

Making a longer story a little shorter... I found that in the axis that most of the mechanisms were printed in, that corresponds to the wheelbase, the error was about .004mm-0.0042mm per mm. So, trying to be clever, I scaled up the latest print by this factor, thinking that the final print would "shrink" down to target dimension.

I couldn't have been more wrong... Even though I measured about a half dozen different 3D prints (all FUD) and arriving at that 0.0041mm/mm of "shrinkage", the latest round of parts that arrived today (a day earlier than UPS's anticipated delivery date no less, go figure!), are actually LARGER than the scaled up (error corrected) solid model that I submitted!!! What the hell??!! Every print I've measured to estimate/anticipate the print error has been SMALLER than the target dimension, but when adjusting for this error, the part ends up BIGGER than the scaled up model??

I understand the tolerances given for FUD on the Materials page for that material... I understand that there's going to be a deviation from target dimensions. I was just sure that I could estimate/calculate the error and adjust the model accordingly. So wrong...

Frustrated and ready to give up and go back to my old (tried and true) but labor intensive method...

Ultimately, I guess precision under a certain tolerance can't be forced with this process. Anybody have any luck with this? If so, what's your secret?

JB

Probably a silly question, but have you defined a print orientation for your models or do you still allow shapeways to orient them ? Tolerances will probably vary per axis.

Bear in mind the tolerance is based on a 100mm sample, while the print-head motion does not scale down for smaller distances. Thus while it is tempting to say a tolerance of 0.4mm/100mm is only 0.04mm over 10mm, it doesn't really work that way because the print-head, the pixels, and the gear teeth all remain the same 1:1 size. Then there's shrinkage as the liquid resin is first heated by laser and then cooled down, with a second heat episode while the support wax is melted out. Assuming your axle spacing is closer to 10mm than 100mm, you'll struggle to get repeatable results accurate to the 0.05mm you are hoping for. Best to keep on with your hand-drilled inserts...