Self-replicating machine

Hi all,
I'm new to RepRap and after some reading through the wiki I have to say this is a pretty fascinating project!
Hopefully sometime I will manage building a Mendel or a Prusa Mendel, let's see.
I like the idea of a self-replicating machine in particular, however, I'm wondering how it is (theoretically) supposed to work.
Every machine has only a limited precision and therefore every generation of self-replicated machines must become worse in my opinion.
How can this limitation be avoided?

Regards,
Michael

... the next generation of a machine with limited precision doesn't mean the 'child' will have an even worse accuracy!

You define the accuracy by selection of the critical components (rods, bearings, motors) and you can even enhabce the accuracy of a breeding with e.g. a stiffer frame, better gliding system, better belts ... and so on.

Actually the reprap needs some 'vitamins' that can't be printed, so by selecting better (and mostly more expensive) vitamines you can make your machine much better than the 'parent' that printed the parts ...

---

Viktor
--------
Aufruf zum Projekt "Müll-freie Meere" - [reprap.org] -- Deutsche Facebook-Gruppe - [www.facebook.com]

Call for the project "garbage-free seas" - [reprap.org]

The accuracy of the printed parts does not affect the accuracy of the child machine, apart from the pulleys. People have commented that with poor pulleys you can actually print a better one as the belt averages out the errors over several teeth.

In general machine tools can be more precise than the tools that make them, otherwise how did we get to the precise engineering tools we have today?

---

[www.hydraraptor.blogspot.com]

Also you have the human quality control factor. If a part comes out really bad it wont be used for a new machine. On the flip side some parts will randomly come out really good and will definatley be used for a child machine.
Statistically an average machine will print parts that vary in quality/accuracy. If this is output is plotted on a graph it will be a guassian distribution. with some really bad, some really good and the majority of parts being ok. If the less than OK parts are discarded by the human quality check then the trend is always positive or leaning slightly toward better quality parts. So as time goes by the child machines should actually get better.
This is also effected by improvements in design and performance tweaks that come along over time.

nophead Wrote:
-------------------------------------------------------
...
> In general machine tools can be more precise than
> the tools that make them, otherwise how did we get
> to the precise engineering tools we have today?

Well, a lot more than the machine tools themselves goes into it. Mostly instrumentation and hand-working the metal, if you're going to be better than the parent machine tool. Mostly it's the measuring tools. Which is why we have to calibrate a newly built printer by measuring the output and making adjustments. Every machine is calibrated back to a geometric standard. Otherwise geometric errors would accumulate in each new generation. Most RepRappers use digital calipers. Machine tool builders have expensive tools for performing geometry checks on new and refurbished machines.

Making better measuring instruments than what you have is a much more interesting subject. You have to devise methods that enhance precision in a way analogous to how a lever amplifies force.

Basically evrything is adjustable. So as long as the parts go together you can mate it work accurately.

So the accuracy depends more on you than the parts. Of course good parts make fitting it together less time consuming and easier.

---

Make your Mendel twice as accurate.
[www.thingiverse.com]

Geometry, There are various mathematical ways of obtaining precise movements from imprecise tools, perhaps the most famous and first was "Watt's linkage"[1]. We have the advantage of being able to order a precision ground straight rod, which we can then use as a guide for straight line motion. However, this was not always the case, In the past, they used clever linkages that mathematically guarenteed motion of a certain type independent of how precise the components were. The same thing is certainly possible with future reprap designs (in fact, some experimental ones do use sarrus linkages), I imagine any "vitamin-less" design will incorporate this sort of thing for precision, rather than rely on the precision of its own parts.

[1] [en.wikipedia.org]