Friday, February 09, 2007


RepRapped extruder parts

I've finally managed to run off a complete set of parts for the screw section of the extruder - clamp still to come. The screw holder itself took me 6 goes to get right (at least we know the extruder can generate a lot of output), with various hardware and software issues being resolved along the way. Printing the clamp requires the current release of Adrian's software, so I'll be upgrading soon.

So now I need to make a new screw drive, bearings and extruder tip!

Vik :v)

Bloody hell that's so beautiful!

Gawd! This makes it all worthwhile!

Thanks, Vik!
It's extruding... an extruder...

That still does lovely things to my head.

Thank you Vik.

So that makes it, what, about the 8th or 10th part to be directly replicated? And some of the builds only have like 30 major parts, about 1/3 of the way to complete replication? (excluding motors and circuitry for the moment)

Rock on!
There are only 9 other part designs in Zaphod. But I'm not going to replicate Zaphod - it's just a testbed. Darwin is going to be faster, more precise, and have a bigger build volume. When the Darwin AoI/STL parts are available I'll have a go at them.

Meanwhiles, the extruder design is a part of Darwin, and we need to know if it can be made to perform when made from Polymorph. Still not there yet, but it's looking promising.

Vik :v)
What are everyone's thoughts about accumulated error? You must have thought about it.

I mean, the first replicator isn't going to have zero error tolerance - so when it makes parts for the second generation machine, they will be less than perfectly precise. But when that second generation machine makes the third generation, won't the error in the second generation parts cause bigger errors in the third generation parts - bigger still in the fourth generation and so on. Is this like making a copy of a copy of a video tape?
I was designing some earth-moving estimation software for my uncle a couple years back. Trying to get the precision down to like the cubic centimeter level when my uncle pointed something out to me.

"We found out years ago that the round-off errors on the cut mostly cancel out the round-off errors on fill, so there's no need to go below a tenth of a yard precision" he told me.

I think the reprap will probably benefit from the same sort "averaging out" of errors. As long as the control software's precision can be adjusted to match the scale of the extruder head's output, errors should fall below a threshold where they accumulate between generations.
We print and measure callibration parts. Then we tweak the RepRap's software to cancel out the aberations.

Vik :v)
I suspect that the simple solution is to design the parts so that they fit over-tight and then trim or heat-soften to get them to fit. Include adjustment screws where possible. Car manufactures use a technique called "functional build" which basically means measure the fit peroidically and adjust. Honda (who I worked for in college) has a half dozen people with mallets that pound out poor fits for example.
Why not just include the CAD data (in an open-source format) as part of the distribution? That way the there is no "accumulated error" possible, plus the CAD file provides a vehicle for publishing upgrades that's absolutely consistent.
Post a Comment

<< Home

This page is powered by Blogger. Isn't yours?

Subscribe to
Posts [Atom]