What's new
What's new

Open Source CMM & Metrology - Where are the complications

snowman

Diamond
Joined
Jul 31, 2004
Location
Southeast Michigan
So does it exist? A cursory examination says no. There are a few companies out there that will retrofit a CMM, but there doesn't seem to be much of a drive out there to share the stuff...in fact, a very clear motivation to keep it as non-descript and proprietary as possible.

So what am I missing from a operations standpoint? It seems that the electronics really aren't doing anything difficult, just generating a location in real space and sending it to the PC. Even for automatic machines operating in 3 axes, it doesn't seem like rocket science.

The software is just software. It can be stupid simple or stupid complex...depends on the end use.

So where am I going ary? What am I missing?
 
I've wondered this myself. The CMM companies seem to work hard to maintain a monopoly on their market. If you can inspect parts on a VMC with a probe and some macros, how hard can writing CMM software be?
 
A real problem is that in terms of volume (number of units) it's a very very small market, AND many of the customers want, or are required to have, certified solutions. I've been in a shop that does aerospace work, and they say every part that leaves the shop is measured on the CMM. We didn't discuss it, but I'm pretty sure they want/need to be able to say "and we have a {mitutoyo, brown&sharp,hexagon,nikon,zeiss} CMM"

What does that have to do with open source? Most open source projects are about very widely used software, generally unencumbered by the cert issue - operating systems (linux), things like .pdf readers, music players, etc. Huge volume, lots of people interested, generally running on very common and well understood hardware. Note that linuxcnc is made to run on very common off the shelf automation parts.

I doubt CMM software is all that difficult, but that might mean something really good is say a 25 person year effort. Figure $2.5 million....

Also, if new hardware only comes with proprietary software, and they don't reveal the hardware interfaces, that makes it very hard for open source software to get a foothold.

It also suggests that the software may be covering up for the hardware - just like roms/bios/acpi/efi/drivers are used to cover up a lot of broken behavoir in PC hardware.
 
A real problem is that in terms of volume (number of units) it's a very very small market, AND many of the customers want, or are required to have, certified solutions. I've been in a shop that does aerospace work, and they say every part that leaves the shop is measured on the CMM. We didn't discuss it, but I'm pretty sure they want/need to be able to say "and we have a {mitutoyo, brown&sharp,hexagon,nikon,zeiss} CMM"

What does that have to do with open source? Most open source projects are about very widely used software, generally unencumbered by the cert issue - operating systems (linux), things like .pdf readers, music players, etc. Huge volume, lots of people interested, generally running on very common and well understood hardware. Note that linuxcnc is made to run on very common off the shelf automation parts.

I doubt CMM software is all that difficult, but that might mean something really good is say a 25 person year effort. Figure $2.5 million....

Also, if new hardware only comes with proprietary software, and they don't reveal the hardware interfaces, that makes it very hard for open source software to get a foothold.

It also suggests that the software may be covering up for the hardware - just like roms/bios/acpi/efi/drivers are used to cover up a lot of broken behavoir in PC hardware.

A couple of things ....a 25 man year software effort is more like 6 to 7 million USD. If you can find me a bunch of brilliant software dudes who will work for 70 K a year, I hire them. Once you add all the overhead a well qualified software guy cost you well over 200K/year. That is at least in the US. You want to farm it out to india...well that is a totally different elephant that has 3 trunks. it requires a lot more management than you think.

I do not believe that CMM is a 25 man year effort. It is more likely a half or less than that. of course it is dependent on what you want the end result to be. Just the measurement or cam files to make a copy :).

There might be patent and IP issues preventing the open source community to get in the space. It would be worth looking into what NIST has available on FOIA.

dee
;-D
 
I can get sw development done very well, by expert programmers in europe, at 59/hr, short term, and in fact we employ 19 programmers doing so, in one area I work with.
That´s burdened contract rate, no emplyment taxes. 15%-20% less if multi-man-year effort is involved.

That said..
Doing a CMM is not a great big effort, if done properly with modern tools, from scratch.

I would estimate 2 man years, about.
Say, 300k, if I sold such a solution, delivery in 9 months, alfa in 4, beta in 6.

I would write it as a packaged solution running on a microserver, linux box, all integrated, hw costs about 100-200€ / box.
Much cheaper to do/keep/maintain than anything else.
Then either operate it from the machine, or remotely.
Its much cheaper to have the hw+sw packaged together, than install instances of the sw on multiple machines.
Ie it costs more than the 200€ to get the sw running on another machine, properly.

Web front end, some packaged ajax, postgresql, std windowed GUI from an existing framework (this is often 90% of the effort).
Run the probing routines from the database, means adding anything new is 1 hour or less of work.
Forget anything hard-coded, no need for it.

And yes, I have done similar stuff in the past, successfully.

Local-lan web-interfaces are fast enough for real work, these days, just like both remote desktop and teamviewer et al.
 
Yes yes, you can make a very pretty looking angry birds. software interface doesn't mean anything without a machine foundation. Air bearings and super accurately manufactured parts. I think the software side of the CMM game is peanuts compared to the engineer and hardware building side.

How are you going to handle multiply file formats, some of which your buying licenses just to install there interrupter. Do you think mastercam or gibbscam would want to charge for an optional plugin for Catia, if they could just give it away.

Maybe I miss read this, you wanna do a cmm solution on a machine tool? Well the rule of thumb in gauging, just as in measuring. Your standard is 10 times as accurate as your measurement. I don't know about you but on occasion I am working to tenths and if I had a CMM Solution(I use that term loosely), It better be able to do that because that is the reason to have it.

It is not software that makes this trade work, it is the hardware.
 
Yes yes, you can make a very pretty looking angry birds. software interface doesn't mean anything without a machine foundation. Air bearings and super accurately manufactured parts. I think the software side of the CMM game is peanuts compared to the engineer and hardware building side.

How are you going to handle multiply file formats, some of which your buying licenses just to install there interrupter. Do you think mastercam or gibbscam would want to charge for an optional plugin for Catia, if they could just give it away.

Maybe I miss read this, you wanna do a cmm solution on a machine tool? Well the rule of thumb in gauging, just as in measuring. Your standard is 10 times as accurate as your measurement. I don't know about you but on occasion I am working to tenths and if I had a CMM Solution(I use that term loosely), It better be able to do that because that is the reason to have it.

It is not software that makes this trade work, it is the hardware.

That's what I'm asking. Is it the software side of it, or the hardware?

In all of the complaints, it's about 10 year old machine is now obsolete because the software is no longer supported. Or the current software package costs xxx annual subscription fee.

The hardware side of it is done, it's sitting in a lot of shops out there, and it quit running two years ago when the machines battery died and the computer would no longer boot. They lost the disks, and can't get it started now....without dropping $25k on a new software / hardware package. Well, the bits and bytes that communicate with the machine controller are proprietary....no standard transfer protocols. But the skeleton is there, the encoders are still there, the motors are still there...they just need an electronics package that can read the encoders and make the motors turn.

That's what I'm asking, from a motion control / electronics point of view, is it more difficult than it seems, or is it just a closed loop CNC machine that gathers data?

Open source means lots of different things to different people. So what I'm envisioning is "you've got the skeleton"....this is the code to interface with the renishaw probe (openly available), this is the code and pinout of the b&s encoders, or heidenhain, etc. Not here's how you mix epoxy granite resin. That would just be asinine.
 
I can get sw development done very well, by expert programmers in europe, at 59/hr, short term, and in fact we employ 19 programmers doing so, in one area I work with.
That´s burdened contract rate, no emplyment taxes. 15%-20% less if multi-man-year effort is involved.

That said..
Doing a CMM is not a great big effort, if done properly with modern tools, from scratch.

I would estimate 2 man years, about.
Say, 300k, if I sold such a solution, delivery in 9 months, alfa in 4, beta in 6.

I would write it as a packaged solution running on a microserver, linux box, all integrated, hw costs about 100-200€ / box.
Much cheaper to do/keep/maintain than anything else.
Then either operate it from the machine, or remotely.
Its much cheaper to have the hw+sw packaged together, than install instances of the sw on multiple machines.
Ie it costs more than the 200€ to get the sw running on another machine, properly.

Web front end, some packaged ajax, postgresql, std windowed GUI from an existing framework (this is often 90% of the effort).
Run the probing routines from the database, means adding anything new is 1 hour or less of work.
Forget anything hard-coded, no need for it.

And yes, I have done similar stuff in the past, successfully.

Local-lan web-interfaces are fast enough for real work, these days, just like both remote desktop and teamviewer et al.

I think you grossly underestimate it :)...But i am game, lets talk this over.
  1. What are the requirements? I would say that would take a better part of 3 manmonths of analysis
  2. platform selection and breaking it down to manageable chinks another month or so
  3. then you would probably need a team of 5 to 6 for about 6 to 9 months towrirte the code test it document etc
  4. take about 1 to 2 months to filed test it on at least 2 or 3 common cmm hardware with the whole team supporting it
  5. at this point you invested abbout 5 mam years. at you calculation contract 59 buck/h you are already at 600K minimum...and there was no other overhead calculated. so say you need to market it, manage it form above, promote it...ok now you are at a cool mill.

soo, how does that sound to you? Not that i have a million to invest for this, but if i wanted to promise it to someone, i would at least ask for that much, and that is just cost. need profit as well, if you do it as work for hire 20% is reasonable, if it is a product you need to get many times the investment.

dee
;-D
 
If you're asking for an open source software package for a stripped table and machine, no such animal exists to my knowledge. I'm sure Hexagon or someone will be glad to refit one with their interfaces and software, they've done it for us.

There's no cheap way to get from stripped machine to good data if that's what you're asking.
 
I think there are two distinctly different aspects to this: the application for CMM in which a company needs certifiable standardized metrology package, and the application where someone/some company needs an engineering capability for their own internal development purposes.

For first application, open source is likely not viable, due to the mfr needing to be able to control the product parameters. The metrology mfr is selling the RESULTS to a certifiable standard, and the hardware and software is a means to that end. As such, they need to control the ingredients and how they are integrated. This is a major reason that the CMM mfrs have a fair amount of proprietary hardware/firmware/software (along with the ongoing need for support revenue).

For second application, anyone with the skills can "do it", and make a point-cloud generation device, a geometry scanner, a whiz-bang whatsit for their own purposes, and get the results required to move ahead with their own development/engineering needs, as long as the results meet internal standards. It's not an easy project by any means, but it could be done, and you could use open-source software as the basis, I suspect. The limitation would simply be the depth of your engineering bench, in terms of the level of complexity of system integration, and how fast you want to get it done. If you want to take your device and then make it perform for certifiable metrology that you can provide to customers, that might become a significant task on its own, if independent 3rd-party cert is needed.

The major mfrs of CMMs essentially provide a single solution for both applications, and charge accordingly.
 
Those coming from a purely software background have no idea what a truly massive undertaking this is. There is a reason that there are only a few options in this space, and that they are all priced very high. It is similar to asking "Why are there no free, great 5 axis CAM packages" or "Who has software to cut complex airfoil shapes for free". It is one thing to write software to drive a machine from point to point in space, and quite another to be able to manipulate a CMM. Probes on CNC machines cover the most basic 5% of the abilities of a CMM, and they are not flexible at all. I would say that their ease of use isn't pertinent to this conversation. On top of the basic technical challenges, you have increasingly 'proprietary' hardware to deal with, too. Newer renishaw heads are not supported by Hexagon- Hexagon pushes their own Tesa brand instead...etc etc. I don't know how much of this is locked at a controller level versus an interface level. Look too at all of the units it takes to control a CMM outside of just the computer with the software on it. On a typical machine with a Renishaw probe setup, you have a controller box. You have a probe interface (PI200/IS1-2/etc). You have an articulating head attached to the quill of the machine. (PH10/PH10MQ/PHS1/etc). You have a sensor attached to the head. (TP2/TP20/TP200/SP25/SP600). You have a sensor module, with your desired probe, attached to the sensor. You have error maps/tables/dozens of parameters for each one of these components. This is an order of magnitude more complex than driving a CNC around in space.

I hate the options that exist. PC-DMIS claims the vast middle ground, and it is woefully dated and buggy in almost all objective areas. For the extent of the software and the annual volume, it basically equates to the price of a high end CAD package, and in that light it seems like a fair value- IF it worked as well as high end CAD software.
 
follow-up:

1. I used numbers from when I left the industry 10 years ago, and they were too low even then. I'll conceded $200K burdened cost. (That's NOT the wage, that's the total cost, people forget that programmers need office space, a machine to use, and so on.)

2. The real point here withstands the range of numbers - developing a solution costs between $500K to to $several million. (ignoring impenetrable trade secrets, patents, and general FUD.) Using a pretty high level class of programmer.

The cost of buying 1 supported certified seat of the software? Surely way less than $500K.

Oh, and if you have the kind of programmers who could really write and support such an app, then either:
a. You are already selling the app to the metrology market, or
b. You are making money having them doing something else with better returns.

Now, the general trend in anything to do with information processing, over the last oh 115 years, is that very general solutions replace special solutions over time. Various market players have always resisted this. But the trend seems very strong indeed.

And so, don't be surprized if a "more open" model of CMMs appears "someday" - but don't like hold your breath either.
 
I won't say that truly open source CMM is coming, but from what I have read, it seems that the market push is for a more standardized control system.

Either way, this has been an awesome thread. While I had considered the statistical analysis on the software side, I hadn't considered the intensity of it, as well as the difficulty of going from a system that produces really precise numbers, to a system that produces really precise numbers that are accepted by a certification agency. The number of people that can do the math involved really isn't that great, the number of people that can do the math AND apply it, does drop quite a bit.

As was said, it's one thing to be able to produce a system that can be used for reverse engineering. A completely different thing to be able to produce a system that can be used to for inspection, to the extent that the market counterparts are capable.

Thank you all for your participation.
 
How hard can it be?
These things use sine wave encoders so just need to figure out voltage or current requirements and how to subdivide the waves.
Time to latch the probe signal is sort of short so you need a little hardware here as software lag will puck you.
Stable motor drives are just drives. easy peasy. Likely that there are two loops to control here the motor and the load but people do that on machine tools all the time.
How hard can it be to solve a 20 point least squares circle when the points are in 3D and not in the plane of the circle or in the plane of any machine axis?
Align part by 3/2/1 and translate everything around in the world is simple 3D graphics, Cad or VR stuff. Simple and basic matrix math stuff here
Certainly comping out the sag in a granite bridge or horizontal arm as they move across can't that hard. Same with bend/twist of the plate as the bridge moves front to back.
Could approaching from different direction on the same point have pitch.yaw and roll you need to worry about?
DMIS and file import support, translation from machine to part coordinates in 3D space, some measuring routines, it's just all software.
Looks easy enough to me.
Bob
 








 
Back
Top