What's new
What's new

The DIY CMM: can it be done?

thunderskunk

Cast Iron
Joined
Nov 13, 2018
Location
Middle-of-nowhere
Hey folks,

We have three Hexagon CMMs at the aerospace shop I work at, and I'm familiar enough with PC-DMIS to know it's a pain in the butt. That said, the value of a DCC CMM in any machine shop is unquestionable. Little old me with my single fanuc robodrill would love to have one small enough to fit in the garage. It would save me thousands worth of inspection equipment such as sunnen bore gauges, larger micrometers, specialty tools, fixtures, etc.

CMMs are not cheap though. With good reason, but there's definitely no low-end version. I love the idea of the Renishaw Equator, but it's only a comparitor and goes for 30k new. Add another 15k or so and Verisurf will retrofit it to have a CMM capability, but... why not just buy a CMM?

Here's an idea: You take a design for a gantry DIY CNC router, build it really stinkin square, 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.

Total might be... $1500-ish? You're not going to get .000001" accuracy, but probing programming has come a long way, and if you're just measuring hole diameters and feature locations, it automates feature inspection at a fraction of the cost it would be to attain the real thing.

What do you think? Worth a shot?
 
If you don't mind results with .002" uncertainty, fine. If you want tight and trustable results I think you want better than a CNC router. Then there's the software, which if you want to do anything more complex than making a list of touch points, will take a sizable programming team a few years to get working.
 
If you don't mind results with .002" uncertainty, fine. If you want tight and trustable results I think you want better than a CNC router. Then there's the software, which if you want to do anything more complex than making a list of touch points, will take a sizable programming team a few years to get working.

But that's just it, I think it can do better than .002". Lets say you use linear encoders like you would for a Bridgeport DRO. The DRO would be separate from the CNC controller. When the probe takes a hit, it sends a signal to both the controller to end the movement and also to the DRO to record the hit.

You're still right about the software, but again this has already been done somewhat. I've seen macros for part probing that already do these things, and I've used excel equations to do some pretty complicated math.

It obviously wouldn't be able to tell you how perpendicular one cylinder was to another or anything crazy, but it also doesn't cost 30-100k.
 
Years (decades) ago there was talk of using CNC mills as CMM's. The idea was that you'd map the errors (for software correction). Then put a probe in it and check the part on the machine, saving (the developers hoped) all sorts of time and errors.

Point being that a proper machine (say your Robodrill) should be intrinsically more accurate (and repeatable) than a cheap 3 axis CNC router. And that being said, pretty sure that idea hasn't really caught on. One would assume for good reason (CNC machines being as expensive as CMM's). Still, it might be a way of doing some otherwise inconvenient checks of you don't mind tying up your mill?

I suppose one might also "map" a cheaper structure and use the corrections to maybe do a bit better; in the hope that errors such as out of parallel would repeat.

I'd think the sweet spot for a one Robodrill shop would be a decent sized surface plate and a somewhat recent vintage electronic height gage?? Or maybe a used Faro arm or equivalent if surfaces need to be more or less checked??
 
:popcorn:
As a CMM and machining vision designer I so see the this should would work from a gut level.
I could just stick a scale on it and it would be right,.. maybe if Mr. Abbe would go away.
Pitch,Yaw, Roll tiny errors become the bane of your life.
It looks easier than it is but I would not dissuade any from coming down into the rabbit hole.
If you don't make one you won't know what does not work or what works good enough.

Overall envelope or worst case accuarte or repeat wanted is not mentioned.
$1500 seems like not a lot of money here.
Bob
 
I'd be mighty impressed if you could make a structurally stable gantry device of reasonable size within a $1.5K budget that had anywhere near the stiffness and accuracy needed to even be reliable to .002".

Even good grades of linear rails must be mounted on rigid bodies and adjusted straight and true, they're not perfectly straight/true as received.

This is not to discourage you from trying this, I'd love an option for a measuring structure that was reliable and accurate to useful levels, say at least .001"/foot. If really good laser measuring devices (LMD) could be found, I'd say make your gantry to "best effort" levels, then actively correct with an appropriate array of LMD's giving real-time corrections.

But pretty quickly I'd bet you're in the $10-20K range, without counting the software needing to integrate the laser data.
 
To echo CarbideBob, this is quite a bit harder than just adding scales and “squaring up” a basic gantry frame. Elastic deflections, straightness, Abbe errors, stick-slip, thermal gradients, heatsources, etc. There’s very little room in the error budget when trying for CMM level volumetric accuracies.

Check out some of the design studies in Precision Machine Design by Alexander Slocum. That book covers everything you’d need to design a CMM, but after reading it they sound cheap to me.

If I were to DIY a CMM i would fix up an old one with bad pneumatics and have the table relapped. At least the difficult alignments and design would already be complete.

Edit: A DIY Faro arm seems less intimidating to me, since precision rotary bearings are a bit simpler. But still those encoders aren’t cheap, and I don’t see it replacing a dial bore gage, for example.

Sounds like a fun project, please report back if you do it!
 
Someone mentioned a rabbit hole...

Everyone here was on the mark: harder than it looks. My first step will be to order some junk: a cheap-O Chinese router frame, KMotion and Mach3 software with a KFLOP board, and a $200 wired spindle probe off ebay. The probe claims +/-.001, the machine +/-.007, and who knows what else. All said and done, probably +/-.020 if I'm lucky.


If I get that far, I can source the 1um encoders from DRO PROs for $200 a pop. Add a Kanalog expansion board to the KFLOP, then build a gantry out of a surface plate and some granite parallels, and it might... improve.
 
I frequently ponder buying a old faro arm and and see what I can do. I think I could buy one at my fav getting spot for a grand. Seems like I could reverse e the software but I remind myself to work on making money, not work on something with zero resale value. I love the idea, but I don’t have time to reinvent that particular wheel.
 
Someone mentioned a rabbit hole...

Everyone here was on the mark: harder than it looks. My first step will be to order some junk: a cheap-O Chinese router frame, ....

If I get that far, I can source the 1um encoders from DRO PROs for $200 a pop. Add a Kanalog expansion board to the KFLOP, then build a gantry out of a surface plate and some granite parallels, and it might... improve.

Why not just a old CMM frame, they are cheap. Often just this side of free.
The scales from your source are 1 micron resolution, not accuracy. There is a very big difference. How will you map this out in the software you will use?
A X-Y-Z is oh-so much easier than an articulated arm. The math gets pretty heavy duty with multiple axis rotations.

If your manufacturing tolerances are +/-0.100 and you need a .010 gauge not so hard. Ten thou tolerance and .001 R&R is beyond what you are thinking about.
I hope you do know the hows and whys of the rule of 10 if you are going to build your own inspection equipment.
This rule can be a small pain in one axis, a problem in two axis, a nightmare in three.
Bob
 
To some extent the 'old' mill has merit. I believe SIP used their mill base for CMMs. I think the Swiss National Labs have some converted. SO, instead of a Bridgeport mill, why not a Jig Bore or such.
 
So, I'm going to straight-up conclude: No. This doesn't work. At least not in my budget.

My ultimate goal was to do cheap lights-out inspection. It's not that it can't be done, and that I won't do this at some point in time just to play around, but I'm going to join the many voices I've heard over a few forums and say this isn't the project I want to get in to right now. I'd love to toss a few K at the project and get something that's semi-useful. At the very least, I need a surface plate, and I'm sure the components would sell after the fact if it didn't work, but I'll never get back the hours into programming something I'm not all that knowledgeable about. Hours better spent making money with my mill. A good few months with the mill might even pay for a CMM, and those dollars gone towards tools and holders will multiply faster than my home-grown CMM will.

Maybe somebody like Tormach will see the niche market and make a $10,000 CMM for simple, low tolerance inspections. But maybe not.

In my search, I found this fella who made a CNC router using surface plates. Now, he marked his pilot hole positions with his personal CNC laser, which didn't look like some cheap Chinese marker to me, but still pretty cool. If I were going to do this to win, I'd build that exact setup with the addition of linear encoders.

YouTube
 
Does not hurt to keep an eye out for a good used CMM. Mine is an old B&S Microval that I purchased from a shop that was closing. $500.00 and it was only 15 minutes from my shop. At the same time, one of my customers replaced an old CMM with a new one. So I actually had two to choose from.

Bill
 
That’s exactly what I did. Just bought a really old mitutoyo for cheap. An FJ-403. Comes with software, and I’m fairly familiar with Geomeasure, though I’m not sure why version it comes with. Paid about 1k.

I think going the opposite direction would be easier first: start with a working CMM and develop backwards. Obviously building a less-accurate CMM with less functional software isn’t top on my priority list, but the input from this and other threads makes it seem like a plausible and fun project at some point.
 
Never heard of Horizon. I don't want to rag on the little guy for name brands sake, but $50k doesn't exactly scream "unrivaled reliability." There's offerings from Zeiss and Hexagon that aren't too much more, just depends on your application.
 
CMMs are not cheap though. With good reason, but there's definitely no low-end version.

This is mistaken. There's a ton of cheap cmm's. They just happen to be in China.

They discovered ages ago that there's not much to a cmm now - granite block, some ways, a simple computer - and voila, you've got a basic cmm. They aren't a Zeiss but for utility work, they are cheap, fast, useful. Not a lot of profit that way so Mitty and Zeiss and so on are not interested, which is why you don't get them.

what size do you want ?

In my search, I found this fella who made a CNC router using surface plates. Now, he marked his pilot hole positions with his personal CNC laser, which didn't look like some cheap Chinese marker to me,
When was the last time you looked at a "cheap Chinese marker" ? Those are also pretty damn good for a low price.
 








 
Back
Top