Quote Originally Posted by Thunderskunk View Post
Hello gents,

I am looking for a CNC router design that is really stinkin accurate, and I definitely need some help on this one.
The cost for accuracy goes up dramatically compared to the average CNC being built here. That is even if you have a well equipped shop already to allow for precision machining of the machines components.
This forum is bloody huge! I drooled over CNC routers as a kid, but I ended up as a CNC machinist and then an engineer for an EDM shop. I have a fanuc robodrill in my basement, which somewhat dampens the desire to build a CNC router, haha. Still cool that folks make em themselves.
CNC routers are one of the few machines that are easily built by DIY types and actually be extremely cost effective over a commercial build. So there is a huge attraction to DIY routers due to getting a lot of capability for a small outlay (relatively). Building a DIY CMM is a different story all together. Frankly you could buy a Datron and still not be good enough for a CMM depending upon the required accuracy. Which brings up the question of how much accuracy is really required.
I am on a different quest. I fully understand that a router with a .0001" resolution would be... well, fairly pointless. What I need is the base to build a Coordinate Measuring Machine (CMM). I haven't seen any cheap already-made machines that fit the description, and there's a lot of things I don't need for a CMM that a router has.
A router with 0.0001" resolution isn't pointless at all, it just isn't the domain most builders on these forums play in. Note that resolution isn't precision so I'm hoping this isn't mixed up. If you are looking to build a CMM I'd first look for a used machine. These come on the market rather regularly as the electronics and software quickly become unusable in modern facilities. The problem with CMM's is software which is well beyond trivial. While this is an off the cuff guess I suspect that the software will require ten times the effort required to build the mechanical systems.
Full-fledged CMMs are not cheap. With good reason, but there's definitely no low-end version. My idea is to take a gantry DIY CNC router minus the router, add a wired probe, use Mach3 with a probing routine, output the measurements in CSV, and use excel to generate a report. Heck, you can even use encoders for position.
Sounds to be extremely hackish. I would suggest a combination of C++ and Python to build the software infrastructure. Even then you would likely have years ahead of you, especially if you want the lights out operation as mentioned elsewhere. At this point I wouldn't even get near Mach 3 for a new project.

As for the costs, low volume and high precision are expensive. You might actually be able to control costs if you can accept less accuracy. If you have production parts sometimes alternative metrology approaches can be less demanding and frankly more reliable.
If I can build it cheap enough and make it WORK (key word), even if the results are garbage, I can build a platform that is more rigid and square using a surface plate and some granite straight edges.
A simple machine should be fairly easy to make work assuming you can find a reliable probing head. Work in this context would be a machine that can take a reading if you manually guide it to the surfaces in question. That is send the machine to a surface of interest, run a probing cycle and capture a positional reading or two. It gets a lot more complicated to collect data from arbitrary surfaces and then generate surface profiles and feature measurements.

In fact you may be able to find some of the simple stuff done already out there on the net.
Anyways, let me know what you guys think. Thanks for the help!

Matt
This is actually a very interesting endeavor you have here. I just have visions of whole teams of programmers trying to work on the software for such a machine. Software is a two edge sword here, there needs to be much effort put into post processing to get valid reports. This especially measuring anything beyond the trivial. Likewise you will need to generate NC code at the front end of the data chain to actually collect the data in the first place. If you don't have code to simplify data collection that will mean manually coding a huge number of data probes.

You are right to start out on a simple machine and frankly you don't always need the high precision that common CMM's offer to get good results. This would allow focusing on software which in my mind is the real challenge.