CNC daydreaming

[rabler]

There are a whole lot of different choices in this as far as CNC control. I'm hoping when I'm eventually done to have a machine with ATC capability, etc, as it has the hardware for that. The existing electrical cabinet is 8'h x 4'w, plus the front console.



I've checked that these servo controllers are 10v analog controllable, including the spindle. The PC-2100 pictured is the old Siemens acramatic controller. I'll need to map out the rest of this before digging into it. I.e, I haven't really started on this yet. I'd like to try replacing the acramatic and break out panels without completely redoing everything but that may be wishful thinking.

Obviously I get started on too many projects
Mesa card availability (FPGA/chip shortage) has kept this one on hold.

The Mesa card can handle up to 6 analog servos, plus 16 discrete inputs and 6 outputs. It also expands via breakout boards for additional I/Os. They have breakout boards specifically for front panel controls. Haven't decide yet, will be interesting to see if I can map the current front panels into one of those, or just go new.

 

[dkemppai]

I've read up on LinuxCNC and the Mesa boards a long time ago. For whatever reason, I decided to go a different path. It looks like very capable hardware.

The big fear with upgrading anything that complex is the process of reverse engineering it well enough to use some of the leftovers. Sometimes you run into something that is so brilliant you ask "How did they come up with that?". Other times you run into something so stupid you ask "How did they come up with that!". The problem is always deciding which of the above it is!

You've got a pretty nice collection of equipment there. I'm sure the VMC will be nice to have running. Being able to keep aluminum flooded with coolant and chips contained would be really nice. On my list is a small home made VMC for the basement. Stefan Gotteswinter has a new video out with a really nice looking frame, about the right size for the basement. I make take some extra notes from that design...

 

[rabler]

I've read up on LinuxCNC and the Mesa boards a long time ago. For whatever reason, I decided to go a different path. It looks like very capable hardware.

The big fear with upgrading anything that complex is the process of reverse engineering it well enough to use some of the leftovers. Sometimes you run into something that is so brilliant you ask "How did they come up with that?". Other times you run into something so stupid you ask "How did they come up with that!". The problem is always deciding which of the above it is!

You've got a pretty nice collection of equipment there. I'm sure the VMC will be nice to have running. Being able to keep aluminum flooded with coolant and chips contained would be really nice. On my list is a small home made VMC for the basement. Stefan Gotteswinter has a new video out with a really nice looking frame, about the right size for the basement. I make take some extra notes from that design...

Well, the Mesa approach is looking more and more like a boondoggle. I have plenty of other things in the queue of projects, but this one has sat for a while.

I've been watching their site since December for the relevant hardware to be available, the 7i97, sold as a "Servo controller" for servos with analog 10V control inputs, which matches the Yaskawa documentation I have for this machine. The 7i97 is "LinuxCNC" compatible, ethernet interface, onboard FPGA. My vague understanding is the FPGAs weren't available during the chip shortage. I have since seen the FPGA, by part number, available in fairly substantial quantity on digikey a copule months ago.

In late May I contacted Mesa and they were in the process of re-designing the boards. Basically it seems they have broken out the front end (ethernet interface and FPGA, or PCIe and FPGA, etc) and the back end (analog, or digital inputs). They contacted me yesterday and have the front-end FPGA board available, but not the back end servo interface part. These were previously sold as one integrated board. They're going to "get back to me" on the servo back end board.

The whole issue here is that general purpose computers (Windows, Linux) are not great at real-time response. They are optimized down to a pretty low level for human reaction times, millisecond or greater. There are real-time variant of operating systems that can help. But at the lowest level, multi-tasking in a virtual memory system doesn't really want to switch contexts at the microsecond level timing needed to accurately maintain sub thousands movement on a CNC motion controller. Offloading to an FPGA can alleviate some of those timing issues. There are other tricks that can be used, but those tricks get progressively more messy from my perspective. The MESA series of cards seems to be the best supported FPGA cards with LinuxCNC.

 

[dkemppai]

My fear with Linux CNC was partly the knowledge needed to get good at it. Not necessarily the CNC, but the Linux part of it.

I've spent a lot of time mulling over just what is the best CNC controller would be. I look a most of the stuff I run across. All of the same concerns you list above. (FYI, I absolutely HATE (with capital letters) Mach 3!) Also not a big fan of GRBL, but for small simple machines it might be good. (And great exposure to CNC for those who can't afford anything else!)

For a long term 'dream' project, I'd like to try the Kflop board. It's a motion controller first, and a CNC controller second. But, like anything it's probably something you have to devote time to become proficient at. The Flashcut systems have worked right out of the box. They have for well over a decade now. When the chance arrived to pick up 3 more controllers (2 with 3 axis 3ph servos.) I jumped at opportunity.
Sort of a case of "The evil that you know" sort of deal. No need to learn a new system!

Something as nice that VMC will be, I'm not sure that the best controller would be. That's certainly a tough call!

 

[rabler]

My fear with Linux CNC was partly the knowledge needed to get good at it. Not necessarily the CNC, but the Linux part of it.

I used started on Unix in the mid 80s and slid into Linux in the 90's so I'm awfully comfortable with the nuts and bolts level, although modern GUI's and package manager/update systems will easily send me to Google.

Haven't quite given up on Mesa although obviously I'm frustrated. If I do, the CNCZone.com forum will probably be my next stop in this venture.

 

[dkemppai]

Haven't quite given up on Mesa although obviously I'm frustrated. If I do, the CNCZone.com forum will probably be my next stop in this venture.

Yeah, that's crummy that they're dragging their feet on getting hardware out. One real concern is if you do get get hardware, get it up and running, are replacements available in the event of a failure? That's a concern with nearly anyone, but more so if you can't even get the hardware the first time around!

 

[Karl_T]

I may have mentioned before that i use Galil cards for the real time motion control. https://www.galil.com/ Jim Dawson does also. These are PERFECT for CNC machines. Used prices on these are very reasonable.

The problem with these is a reasonable cost human and logic interface. Jim built his own, not 1 in a 1000 can do that. I went with Camsoft but the cost is extreme to buy retail, not worth it to most. ( I bought everything used and then became a software consultant and dealer for them) Mach was going to interface these cards but gave up.

Personally, i don't understand why no one in the Linux world has not picked up on this.

 

[rabler]

I may have mentioned before that i use Galil cards for the real time motion control. https://www.galil.com/ Jim Dawson does also. These are PERFECT for CNC machines. Used prices on these are very reasonable.

The problem with these is a reasonable cost human and logic interface. Jim built his own, not 1 in a 1000 can do that. I went with Camsoft but the cost is extreme to buy retail, not worth it to most. ( I bought everything used and then became a software consultant and dealer for them) Mach was going to interface these cards but gave up.

Personally, i don't understand why no one in the Linux world has not picked up on this.

I've done a lot of C coding in my time. And some assembly and PHP, not to mention a few other oddities. Enough that I don't really want to do more. But also enough that I like the idea of an open source controller that I can tinker with if the mood strikes me, thus LinuxCNC. It puts me in an odd position as a hobbyist ...

The problem with Galil is LinuxCNC wants to close the control/feedback loop itself rather than delegate that to a separate controller. FPGA controlled timing seems to be about as far as the LinuxCNC philosophy is willing to accept.

 

[Shotgun]

Once I figured out that the stepper motors I got from StepperOnline do not like to have extra connectors added to their supplied cables, I've been making great progress using LinuxCNC on an OrangePi5 paired with a Mesa 7i96s. That board is targeted at stepper motors with 5V signals for the drivers. So it may not apply to your system with servos.

 

[dkemppai]

The problem with Galil is LinuxCNC wants to close the control/feedback loop itself rather than delegate that to a separate controller. FPGA controlled timing seems to be about as far as the LinuxCNC philosophy is willing to accept.

So, is LinuxCNC the proverbial stick in the mud here? You'd think that adding support for motion control cards would only open up the power of LinuxCNC to more applications. Is this just legacy (code) from the way LinuxCNC was written?
(Having never looked into the code I'm not sure what that would entail. It may be that a complete rewrite would be required.)