Linux CNC going Commercial...

I'am using LinuxCNC on a big milling machine (4-axes + horizontal an vertical head) with 12 ton weight. After two years of more or less daily industrial use I can give my statement at that discussion: One big problem is the look ahead at cam generated programs with a high accuracy or high speed when the increments are very small, you can't go fast enough. Also for example I have programmed a cone helix for a NPT 1/8 thread with polar coordinates. It works but not very fast! An other problem at my version 2.4.6 is that in the Y-Z plane with G19 no radius compensation (G41, G42) is implemented. Maybe that is fixed in the newer releases. Another "not very nice situation" is that you have to define a new tool length definition for each plane G17 with Z, G18 with Y and G19 with X. I will say that LinuxCNC works but you can't compare it with a 20000$ control.
I'am sure that sercos3, ethercat, ssi and/or drive-CLiq is also in LinuxCNC the future but it is a long way! When I started my retrofit I thought also in that way to offer a LinuxCNC system comercially with training courses, but now I guess it need a couple of years to be a real competitor on the german market.
This is only my view and I'am a fan of this software!

Ingo (Germany)

For sure there is a lot of Work to do...
But I dont want to start making Money with that thing in a few weeks or months... I`m calculating this Project to take about 3 to 5 YEAR`s minimum!

Maybe its going to change from LinuxCNC to another Open Source-Software or something else. But I have to Start somewhere, and actually I do not see an alternative to LinuxCNC. Anyway it is a Great (and Cheap) way of Learning how to solve all that Problems that come with the building of a controller.
And I think, then more People work with LinuxCNC then better are the chances, that somebody will integrate a working Look-Ahead and all that Stuff, that is still missing in LinuxCNC...

Latest Info for the Project itself:
Because i`ve been travelling a lot arround for Customer-Service the last weeks, I only was able to set up my testing System:

I slaughtered my old 19"-Rack-Fileserver, and changed it`s internals (Hardwaredetails will follow). I discarted my first Idea to use Arch-Linux and changed to Debian because of the already working RT-Preempt-Kernel of Pengutronix (www.pengutronix.de/software/linux-rt/debian_en.html).

Thanks to the Arch-Linux Community for that Advice!

So my first Testsystem is a Debian with RT-Preempt. I will try to continue to install Linux CNC over Christmas Vacation, so i can start experimenting in January.
Maybe later I will Migrate to Arch-Linux or ART-Linux. But for the Start Debian does a good Job I think, and its much smaller than Ubuntu.

Last but not least all of this depends on my peronal ability to learn how to deal with Linux. As I wrote, I`m still a newbe in the Linux Universe and there are so many mistakes for me to make and so many things to learn

For sure there is a lot of Work to do...
But I dont want to start making Money with that thing in a few weeks or months... I`m calculating this Project to take about 3 to 5 YEAR`s minimum!

As a short term goal, may be you want to start doing some professional front panels plenty of hardware buttons, there is no such a vendor in the market and i need one

nursum wrote:

I'am using LinuxCNC on a big milling machine (4-axes + horizontal an vertical head) with 12 ton weight. After two years of more or less daily industrial use I can give my statement at that discussion: One big problem is the look ahead at cam generated programs with a high accuracy or high speed when the increments are very small, you can't go fast enough. Also for example I have programmed a cone helix for a NPT 1/8 thread with polar coordinates. It works but not very fast! An other problem at my version 2.4.6 is that in the Y-Z plane with G19 no radius compensation (G41, G42) is implemented. Maybe that is fixed in the newer releases. Another "not very nice situation" is that you have to define a new tool length definition for each plane G17 with Z, G18 with Y and G19 with X. I will say that LinuxCNC works but you can't compare it with a 20000$ control.
I'am sure that sercos3, ethercat, ssi and/or drive-CLiq is also in LinuxCNC the future but it is a long way! When I started my retrofit I thought also in that way to offer a LinuxCNC system comercially with training courses, but now I guess it need a couple of years to be a real competitor on the german market.
This is only my view and I'am a fan of this software!

Ingo (Germany)

your miles behind in terms of code , we are now at 2.5.1
and a lot of bug fixes too

Yes I know, and I bought a new PC last week to run 2.5.1 separately. I 'am working in that way: Never touch a running system and I don't want to update my working system. So I want to have a complete set of hardware for spare. That reduces time when something happen....

Ingo

@ VNR:

What do you think a Professional Frontpanel has to look like? Touchscreen, separate Rapid-Feed, and Feed Overrides, Buttons for Coolant and other Peripherials, Waterproof?

That wont be too Hard to do... Buy the single Components, Find a Case you like and assemble them. The Difficult thing will be make it work with the Software and the connection to the Controller.

Or build an Embedded System which uses only a USB or LPT port and needs a seperate Driver?

But its a good Idea to build something like this... if there is no vendor for that.

cncbasher wrote:

your miles behind in terms of code , we are now at 2.5.1
and a lot of bug fixes too

But, the lookahead is not any different, yet! I hope this gets worked on soon.
There are some ways using G61 Pnn that can help with some programs
that have many short segments.

Jon

As a short term goal, may be you want to start doing some professional front panels plenty of hardware buttons, there is no such a vendor in the market and i need one

There are actually dozens of such shops in the US alone, probably dozens more in the EU. There are three or four less than an hour away. We gave up building custom HMI's because the market is flooded. And there is no such thing as a standard HMI, even for a 3-axis mill.

The problem is that every customer wants something better (euphemism for different). The prices are high because each job ends up being a one-off. By the time you pay for reinventing the wheel you might as well go with a commercial package. Centroid will sell a complete system for what a panel shop gets just for the HMI.

Doug