[RCaffin]

all ABB needs to do at this point is buy a CAD company and they're complete from beginning to end
True, true ... with about 0.1% of the market. Meanwhile, the generics are coming in at a fraction of the cost with industry-standard interfaces. Remember all the incompatible variations on the 'personal computer' ?

Ah, but can the generics match the quality of the trads? It may be worth remembering the engineering comment made some time ago, that a modern Toyota Corolla is of a higher quality with fewer defects than a WW II era Rolls Royce. Why? In a word - volume.

Gecko offer their 203V stepper drive with a really fascinating warranty: if it dies (short of deliberate sabotage) they will repair or replace for free. The 'V' in the product name stands for 'Vampire' - unkillable. Volume, volume, volume. I gather they have had a few returns - mostly due to metal swarf getting inside the case and shorting out IC pins. (HOW????) But they still fixed it for free.

Cheers
Roger

[samco]

Couple things.
Yes. You can run the 7i76 and the 7i77. You need the correct bit file (think firmware) in the 7i92. Looks like there is one.

7I77_7I76D is a configuration intended to work with a7I77 six axis analog servo daughtercard on P2 a 7I76 daughtercard on P1

You are going to need rt_preempt realtime kernel to run mesa Ethernet hardware. (The current livecd is rtai realtime kernel.).

If I where you - I would use the new livecd for linuxcnc that has the latest Linux (Debian stretch) and the rt_preemp kernel.
(We need testers )
Index of /testing-stretch-amd64-rtpreempt

Laptops... Hit or miss. I have actually had good luck with rt_preempt on laptops (vs rtai). But you will need to test. Remember to research isolcpus (for isolating cpu cores for real-time).

Linuxcnc is not a buffered system. The computer is communicating with the mesa card in real-time. 1000 times a second every second.

Sam

[samco]

Just to clarify... Mesa hardware doesn't work with mach

"

So, from my Linux research today, it seems like I'm likely looking at:
-Mesa 7i92 motion controller board (ethernet connected to a machine running EMC/Mach/etc)

TCB

[barnbwt]

Few more questions about the controller options out there;

-As far as LinuxCNC's requirement that all the timing/synchronization be performed internally to the PC, is communication via Ethernet to the Mesa cards reliable as far as maintaining real-time sync between the machine/controller/PC? It seems like latency would be the Achilles' Heel of a system that performs these functions so far removed from the cutting tool, and Ethernet, while fast, isn't the quickest thing out there. The fact I'd be running from a laptop is especially concerning (makes me think a PC-based PCI card is almost required here, whereas for other solutions like Galil it merely opens up more options for using eBay equipment). I'm also troubled by my research which indicates that LinuxCNC may still lack the formal servo-spindle capability (the most recent threads I can find mention rigid-tapping as about the most complex spindle operation supported, and more complex moves as likely possible but still theoretical). I know it's possible since Tormach does it, but there seems to be no alternate or open source method at this time.

-Looking into the Galil/Mach option some more; it appears that several years ago there was a major evolutionary diversion point as far as machining technology when controllers went from linear-interpolation to high-order contouring guidance. In fact, it appears the massive increases in data/speed required for this technology was one of the reasons Mach4 came into being (or that it's development required the new software to depart greatly from Mach3). The result is that the Mach drivers are intended only for the newer controllers, with no apparent plans to expand their capability to older drives (only a handful of the last Galil linear systems seem to be somewhat compatible at this time), thus making the Mach3/Mach4 split even more stark for a guy like me trying to put a system together. Are the benefits of contour-motion control and formal driver support by Newfangled Solutions really worth the ~400$ difference that the newer drives & software command? Or is the smooth control of high-order acceleration shaping wasted on a small/light machine that isn't moving hundreds of inches per minute? I'll also add that while I don't have the nicest parts & equipment to start out with, like everyone I'd like the capacity to upgrade to improved equivalents as the machine is proved out (is Mach3 & linear machining locking me into dead-end technology that will soon be irreplaceable?)

-A third option that *seems* like it should have more interest is the FW-7XXX controllers by National Instruments, simply because of Labview (it's not open-source or anything, but it is still fairly common in universities IIRC). I'm having a little trouble figuring out exactly how their systems are put together, but it appears they have multiple software interfaces that can set up or operate the controllers, and not all of them require payment of a month's wages like Labview. It looks like NI Motion is the free servo/driver configuration utility, LabWindows/CVI a free "Labview-like" environment to run a theoretical G-Code interpreter in, and NI Motion as the user interface (I think, not sure on this one). Anyone have any experience with these NI products?

TCB

[PCW_MESA]

Few more questions about the controller options out there;

-As far as LinuxCNC's requirement that all the timing/synchronization be performed internally to the PC, is communication via Ethernet to the Mesa cards reliable as far as maintaining real-time sync between the machine/controller/PC? It seems like latency would be the Achilles' Heel of a system that performs these functions so far removed from the cutting tool, and Ethernet, while fast, isn't the quickest thing out there. The fact I'd be running from a laptop is especially concerning (makes me think a PC-based PCI card is almost required here, whereas for other solutions like Galil it merely opens up more options for using eBay equipment). I'm also troubled by my research which indicates that LinuxCNC may still lack the formal servo-spindle capability (the most recent threads I can find mention rigid-tapping as about the most complex spindle operation supported, and more complex moves as likely possible but still theoretical). I know it's possible since Tormach does it, but there seems to be no alternate or open source method at this time.

-
TCB

There is no real issue with LinuxCNC/ Ethernet for real time, though Laptops are generally not suggested as real time hosts since they often have power
management firmware that cannot be disabled and introduces large latencies. A Mini PC like a Intel NUC or Gigabyte Brix is usually a much better real time host
( standard laptop keyboards are also not good around metal chips...)

When running external stepgeneration/encoder hardware, LinuxCNC only requires a reliable 1 KHz communication thread rate which is easily done over Ethernet,
in fact with the right hardware, 4 KHz is possible with Ethernet.

Latencies up to 500 usec (on a 1 KHz thread) will cause no major errors with Mesa Ethernet hardware since the stepgen and encoder sampling is done in hardware
via a host synced DPLL which reduces the position sampling latency to about 100 ns

I think that a servo/standard spindle mode change will require a custom hal component, though it should not be terribly complex

[ger21]

it appears that several years ago there was a major evolutionary diversion point as far as machining technology when controllers went from linear-interpolation to high-order contouring guidance. In fact, it appears the massive increases in data/speed required for this technology was one of the reasons Mach4 came into being

There's no major technology change from Mach3 to Mach4. They both actually use a very similar trajectory planner.
Mach4 was born when Mach3 became to complex to make bug fixes or modifications, without introducing more bugs. Mach4 is a totally new program from the ground up, so Mach3 plugins do not work with Mach4.

Are the benefits of contour-motion control and formal driver support by Newfangled Solutions really worth the ~400$ difference that the newer drives & software command? Or is the smooth control of high-order acceleration shaping wasted on a small/light machine that isn't moving hundreds of inches per minute? I'll also add that while I don't have the nicest parts & equipment to start out with, like everyone I'd like the capacity to upgrade to improved equivalents as the machine is proved out (is Mach3 & linear machining locking me into dead-end technology that will soon be irreplaceable?)

Again, Mach4 is not doing any magic that Mach3 is doing. The difference is that Mach3 development stopped 5 years ago, when Mach4 development started.I would expect it to be at least a few years for Mach4 to get a new trajectory planner, if it gets one at all. And it won't be any new ground breaking technology.

[Jim Dawson]

-Looking into the Galil/Mach option some more; it appears that several years ago there was a major evolutionary diversion point as far as machining technology when controllers went from linear-interpolation to high-order contouring guidance. In fact, it appears the massive increases in data/speed required for this technology was one of the reasons Mach4 came into being (or that it's development required the new software to depart greatly from Mach3). The result is that the Mach drivers are intended only for the newer controllers, with no apparent plans to expand their capability to older drives (only a handful of the last Galil linear systems seem to be somewhat compatible at this time), thus making the Mach3/Mach4 split even more stark for a guy like me trying to put a system together. Are the benefits of contour-motion control and formal driver support by Newfangled Solutions really worth the ~400$ difference that the newer drives & software command? Or is the smooth control of high-order acceleration shaping wasted on a small/light machine that isn't moving hundreds of inches per minute? I'll also add that while I don't have the nicest parts & equipment to start out with, like everyone I'd like the capacity to upgrade to improved equivalents as the machine is proved out (is Mach3 & linear machining locking me into dead-end technology that will soon be irreplaceable?)

TCB

While the current evolution of contouring mode is OK, it puts all of the heavy lifting on the controlling PC. It seems a bit ridiculous to take a very intelligent motion controller and turn it into a zombie slave. Contouring mode is still a series of linear interpolated moves. The difference is that you are commanding a new position and a time to arrive at that position as opposed to letting the controller figure out how to do it. The net result is the same. You get around any latency in the ''LI'' moves by pre-loading some number of lines of code and the controller blends it into one continuous motion. I normally keep around 250 lines in the buffer (out of a possible 511 lines). This works well with both Ethernet and PCI controllers. Contouring mode works best with PCI cards because you are communicating at bus speeds. It would seem to me that contouring mode would work very well for an aircraft autopilot system, robots, or maybe driverless cars where you have a relatively long period between new commanded positions.

Last time I looked, Galil did have Linux drivers, but I don't know if it will work with LinuxCNC. No experience at all with that.

[barnbwt]

I figured it must not be a problem since it obviously works for so many people, but I did see the occasional thread by some poor guy who couldn't get his latency down to usable levels by virtue of his hardware setup. I'd have thought a fairly late-model laptop running a low-load OS with all the distractions turned off should be able to accomplish this, but it's not like I'll know for sure unless I try.

My only issue with adding a dedicated machine for Linux or cards to the mix, is that at that point I'm in it a total of three Mesa cards and a PC; gonna be pushing the better part of 1000$ minimum. For that price, I could get a nice Galil controller plus Mach4 hobby & the ready-made driver software/translator. I do like that the little thing only draws 65W; since I'm on a 15A outlet, power supply is but one of the many design constraints...constraining my options.

"I think that a servo/standard spindle mode change will require a custom hal component, though it should not be terribly complex"
That's pretty much what I keep seeing; that the ingredients to do what I seek are certainly there (the "ORIENT" command, specifically) and yet that's as far as anyone seems to have gotten, apart from basic mechanical gearing. I can program if forced to, but have little natural aptitude for it, so having to construct a custom machine control is a distinct disadvantage (if not impossibility) for me. The goal, so to speak, is to be able to profile a funky candlestick shape at speed, then engrave text into it across either axis at a constant depth. Whatever method gets me there most efficiently 'wins.'

As I said, Tormach obviously figured it out, using a Linux-based system, so it's certainly doable.

TCB

[barnbwt]

"While the current evolution of contouring mode is OK, it puts all of the heavy lifting on the controlling PC"
Interesting, so the real advances in these controllers was more in their capacity to take in & parse huge amounts of data very quickly, as opposed to controller function (vs. the little piddly stream of simpler G-code directing it along jointed paths, which would be monstrously large/slow tape files if made into optimized spline routes). While I want to be capable of machining aluminum potatoes & engraving them, the typical job will be of the more common cylinder/taper variety, and certainly not at what is considered a "high speed machining" pace. It sounds like my little machine really won't benefit from contour capability all that much, anyway.

So, the next question, is am I dooming myself to non-functional obsolescence in the near future by sticking with an older Galil (let's say 21xx series) and now-unsupported Mach3? I just think that Mach4 will only grow in popularity in the next few years, particularly among the heaviest users/developers that the rest of us sponge off of like leeches, lol. Unless there's a spindle-servo capability available right now, I'd be very concerned it will never be developed at all (for Mach3, at least). And as with LinuxCNC, it's not really clear if anyone has pulled it off yet in a way that can be duplicated. If I already had a Mach3 system, it'd be no contest, but for a guy just starting out, I'd hate to be backing the lame horse.

A quick search suggests that LinuxCNC isn't really meant to run a Galil controller, since their functions are basically redundant (i.e. you'd have to lobotomize the Galil like you mention). I guess those drivers are so you can at least get the Galil configuration utilities running on a Linux machine?

TCB

[RCaffin]

There are many myths in the CNC world. It would seem that 'contour mode' is one of them. As Jim wrorte:
Contouring mode is still a series of linear interpolated moves.
There are no two ways around this.

Now, the real difference is probably historical. Before the advent of PC control, all the numerical grunt had to be done in the controller - and that meant the controller had to include the equivalent of a CPU chip, or the ability to do mathematics. Low-end controllers could do linear moves OK, but you had to spend a whole lot more $$ to get the ability to program curves into the controller. So some machines could do curves and other machines could not.

But today we have PCs with multi-core gigaHertz CPU chips, and they are CHEAP. You want a circle? The PC can issue the all commands needed. To be sure, there may be a thousand very short moves generated by the PC, but so what? Well, it turns out that another 'feature' is needed for this to work well. In the Mach world it is called 'Constant Velocity Mode', and that means that the program does some look-ahead to handle the dynamics of the machine. Corners can be taken smoothly, without stopping. So when the program issues a thousand very short moves to cut a circle, that is done in CV mode, and there is no stopping between the moves. You get some very nice curves.

So where does 'Contour mode' come into this? Basically, it is now redundant. Instead of having the controller try to plot out the path for a shape (using any of several interpolations techniques), the PC can do that under the user's control, and using any mathematics you want. The NIST g-code language has enough embedded mathematics that you can program almost any shape you want. G61/G64 is now obsolete.

Attachment 368432
Attachment 368434
These shapes were created using parametric programming handled by Mach3 and an ESS. The contour mode instructions were NOT used. They are too limited and are now redundant.
(Yes, some of the shapes show machining marks: they are only prototypes created out of softwood while testing the programs. Harder materials will follow.)

Cheers
Roger
PS: paper tape ...yeah, sigh. These days we use 100 MHz and GHz Ethernet instead.

[Jim Dawson]

I would say that going with a 21xx and Mach3 would allow you to run virtually forever. Once it's installed and working, it will work until your computer blows up. Just keep backup copies of all the files so you can load them on your ''new'' computer.

As soon as I can find a 2-3 KW servo drive/motor (at my price), I'll be CNCing my lathe with my CNC software, and it will work with a Galil controller. Maybe by this fall you will have an alternative to Mach3/4. I have a little more design work to do also because I will keep full manual capability as well as 3 or 4 axis CNC that I can switch back & forth with a mouse click. I have most of the hardware on the shelf, just need to find the time to do the finish designs. I just bought another breakout & cable today for $100.

[PCW_MESA]

"While the current evolution of contouring mode is OK, it puts all of the heavy lifting on the controlling PC"
Interesting, so the real advances in these controllers was more in their capacity to take in & parse huge amounts of data very quickly, as opposed to controller function (vs. the little piddly stream of simpler G-code directing it along jointed paths, which would be monstrously large/slow tape files if made into optimized spline routes). While I want to be capable of machining aluminum potatoes & engraving them, the typical job will be of the more common cylinder/taper variety, and certainly not at what is considered a "high speed machining" pace. It sounds like my little machine really won't benefit from contour capability all that much, anyway.

So, the next question, is am I dooming myself to non-functional obsolescence in the near future by sticking with an older Galil (let's say 21xx series) and now-unsupported Mach3? I just think that Mach4 will only grow in popularity in the next few years, particularly among the heaviest users/developers that the rest of us sponge off of like leeches, lol. Unless there's a spindle-servo capability available right now, I'd be very concerned it will never be developed at all (for Mach3, at least). And as with LinuxCNC, it's not really clear if anyone has pulled it off yet in a way that can be duplicated. If I already had a Mach3 system, it'd be no contest, but for a guy just starting out, I'd hate to be backing the lame horse.

A quick search suggests that LinuxCNC isn't really meant to run a Galil controller, since their functions are basically redundant (i.e. you'd have to lobotomize the Galil like you mention). I guess those drivers are so you can at least get the Galil configuration utilities running on a Linux machine?

TCB

C axis LinuxCNC lathe setup:

https://www.youtube.com/watch?v=hDwQDdJOqnA

Pretty sure Eu-surplus would supply the config files if you asked
also pretty sure this is using Mesa hardware

[barnbwt]

Hmm, that's more like what I'm after! Now, I'll just take it on faith that the spindle can still ramp up to speed & do all the profiling/facing moves prior to the gear cutting, even though it wasn't actually shown (exactly how the transition between the two modes works programmatically is what I am struggling with, as have many others doing these retrofits it seems)

More servo spindle porn;
https://www.youtube.com/watch?v=RfnoAFW2L2c
Much like a guy who drives a 3/4 ton pickup to work, "I can't say I'd always use the extra capability, but I like having the option available to me." With a Foredom Tool-sized live spindle, I wouldn't be doing much more than engraving, but being able to engrave along uneven surfaces seems like it requires the same control functions as the video above.

RCaffin;
So it sounds like our modern glut of processor speed, data storage, and transfer rates allow the G-code side of things to simply "brute force" their way through complex tool control, eliminating the need for high order interpolation. I know the bandwidth/bus speed limits of the older parallel/etc protocols put a limit on how much data could be sent to the controller at once, but now that we're using terabyte drives and multi-core processors over Ethernet, we can essentially have the CAD/G-code processor make our toolpaths as complex as we care to? In a sense, I suppose that means that the only *real* functionality I am obtaining via these off-board controllers is the +/-10V analog control signal the computer can't make by itself; basically the same thing as the 7i77/7i76 Mesa cards rolled into one.

I think I'm starting to make sense of how these legacy boards are being utilized; one of the Linux threads mentioned that the allowable acceleration limits are increased to crazy levels in the controller, so it won't try to round off or slow down around curves, essentially defeating the entire purpose of those (then) sophisticated control enhancements. Now, they'll just have the software send the controller every little data point it needs to make the sharp turn at the slower feed. I have to assume the Galil/Mach machines are doing something similar in the controller settings, or are at least capable of it.

Jim Dawson;
"I would say that going with a 21xx and Mach3 would allow you to run virtually forever. Once it's installed and working, it will work until your computer blows up. Just keep backup copies of all the files so you can load them on your ''new'' computer."
Well, until the capacitors blow up, anyway, lol. You make a fair point, that a working configuration available now is a working one going forward. Can Mach3 run on emulators of the older Windows OS's, or does it have to be a full-blown native installation operating it? Having to run XP would be considered a drawback (mostly because I have not looked into how hard it is to build an XP-compatible system with today's components). Now, supposedly the Mach4 plugin does have some kind of functionality with the 21x3 models, it's just that the developer isn't devoting any time to them since he feels they're antiquated (apparently his cutoff point is pre/post "contouring" capability since the forward-buffering aspect of that technology really changed how the controller & computer talk to each other). It sounds like Mach4 works with them, just not with the newer/more intelligent capabilities that justify the greater expense (and it sounds like we shouldn't expect future additions via development that are backwards compatible, either)

TCB

[Jim Dawson]

Can Mach3 run on emulators of the older Windows OS's, or does it have to be a full-blown native installation operating it? Having to run XP would be considered a drawback (mostly because I have not looked into how hard it is to build an XP-compatible system with today's components).

TCB

Not sure if Mach3 will run on an emulator, I suspect not because of the way the kernel works. But XP runs just fine on my Core i7 computer, I'm set up for dual boot. I work on a lot of legacy stuff so I need both XP and later versions.

[RCaffin]

It is not hard to build a system to run XP. ANY supplier of industrial controls will enthusiastically send you a whole catalog of options. You see, a huge fraction of industrial control systems are still running XP, often because XP does not have the straitjackets on drivers which W8 & W10 impose. Local PC shops often don't carry that sort of stuff though.

That said, I know Mach3 runs just fine on W7 with an ESS - because that is what I am using. If you use an ethernet interface like the ESS then it may be that you can run Mach3 up to W10, because once you take the PP driver out the rest of Mach3 is reasonably well-behaved. I THINK. Confirmation anyone?

This does lead to the question of whether you can run Mach3 under WINE. I wish, oh I wish. Has anyone succeeded?

now that we're using terabyte drives and multi-core processors over Ethernet, we can essentially have the CAD/G-code processor make our toolpaths as complex as we care to?
YUP.
Carving a wooden plaque with a 2.5D router is really very, very simple. You just have a huge number of XYZ data points which make a raster. The file can be megabytes though!
Carving more complex shapes, like the wood ones I showed in #50, actually requires genuine 4D machining: X, Y, Z and A are all changing smoothly and simultaneously. There are plenty of 'classical' CNC machines which cannot do that: they can only use their rotary table as an indexer. Mach3, Mach4, LinuxCNC, UCCNC all can do this, as they rely on the CPU chips in a PC. (The program for the wooden shapes is about 10 kilobytes - it's parametric.)

Cheers
Roger

[ger21]

Imo, if you are going with a Galil controller, I would recommend Mach4, rather than Mach3, as Mach4 is being actively developed.

[RCaffin]

Hi Gerry

Good point about Mach4.
What features remain to be fully implemented in Mach4 now? Or are they listed somewhere in a simple form?

Cheers
Roger

[ger21]

1) It depends on which controller you are using, as a lot of Mach4's "features" are handles by the motion controller.

Since the Galil is the only hardware that's actually supported by Artsoft (they write the plugin), it shouldn't be lagging behind any other controllers feature wise.

2) I 'm not sure, as I gave up on it a while ago, and don't follow too closely. I don't that it doesn't have path tolerance settings like all of it's competitors do.

[RCaffin]

Hi Gerry

Ah well ... thanks.

Cheers
Roger

[barnbwt]

Imo, if you are going with a Galil controller, I would recommend Mach4, rather than Mach3, as Mach4 is being actively developed.

But only the newer controllers; that's why 4 locks me in to a (somewhat) higher cost system, not counting the software. The real question is whether the added functionality of the newer hardware/software and future support/development are worth it.

I think, since I would like to be capable of upgrading my replaceable components once the lathe is running, that Mach4 is the better path to take, even if it slows down the rate at which I can acquire the remaining parts a bit (but they aren't going anywhere)

So how does Mach4/Galil compare with Linux as far as capability, or are they mostly equivalent? The video linked earlier also has the Linux rig doing tapered threads, which I think demonstrates full multi-axis coordination capability (I think a square axial-profile or uneven-ended surfacing move could be approximated with a bunch of 3D segments to follow the nominal surface properly)

TCB