Motion Controlers

[davidperry3]

There is an interesting article on CNC Cookbook about motion control that got me thinking about our computers... Do these motion controlers allow us to use more powerful computers? I was playing around with rotary machining and found that I could easily generate a NC file so big that it would bog the machine computer and make the machine go ca chunk ca chunk anytime the A axis was moving. I sent Bob Warfield the following e-mail. His reply leads me to stay in the dont mess with it catagory for now but what do you guys think.

here is the e-mail

Ok so I am following your motion control series. And I am wondering if I need one for my Tormach? I am very much an amateur and most likely don’t know what I don’t know...
It would be nice to have g-wizard at the machine instead of having to run inside every time something changes (which is often).



David, I'm definitely a fan of the motion controllers. With that said, there are some things to consider before adding one to your Tormach:

- Does it void your warranty or limit your ability to get support? I had understood that if you monkey with their Mach3 configuration you can have problems.

- Given that concern, does it limit your ability to use their accessories, such as the ATC if you modify their Mach configuration?

- The Smoothstepper is not completely free of issues. They only recently have added backlash compensation and it has problems when you move less than the backlash, for example.

I really enjoy having Internet access on my shop machine. I refer to it constantly, and also use it to transfer files from my up 2 flights of stairs CAM system. I also really like my Smoothstepper, despite the issues with it. I would like to try a KFLOP too at some point, and see if I like it better.

With all this said, you should also consider the "if it ain't broke, don't fix it" angle. The Tormach is a well-balanced package. I don't know if your performance would improve or not. I'd love to see more articles where people "hot rod" Tormachs:

- How about adding servos and bumping the motions up to circa 200 IPM?
- More HP on the spindle?
- Smoothstepper or other motion control?

Just makes me wonder how much further the machine could be taken.

Best,

BW

[BobWarfield]

Heh, just need to get some popcorn to see what comes of this!



Cheers,

BW

[Scott_M]

Hey David

I recently ran an engraving program that used a drag engraver. I was running at 40ipm. the code I was running was generated by someone else and had all the arc moves as straight lines. This means about 40 lines of code for a 1/4" move. Running at that speed made my machine clunk and chunk and sputter. I have unlocked my control software, so I increased the " Look Ahead" value from the stock Tormach value of 20 lines to 100 lines and the program ran smooth as butter.

Not sure if this is what you are seeing but I thought I would mention it.


Bob

I really wish you would reconsider a stand alone version of G-wizard. Even if it is a "light" version. I would buy it for sure.
But I will not put internet access on my machine controller. ( Or anything Adobe )

Scott

[BobWarfield]

Bob

I really wish you would reconsider a stand alone version of G-wizard. Even if it is a "light" version. I would buy it for sure.
But I will not put internet access on my machine controller. ( Or anything Adobe )

Scott

LOL, Scott that's a different bag of popcorn discussion that I indulge in on other forums.



Cheers,

BW

[davidperry3]

Is there a downside to increasing the look ahead? Mayby a better question is why did they choose 20 lines when 100 is possible?
I belive what I was seeing is the computer itself bogging down. Whenever the A axis was moved by the program (or MDI) the toolpath image was jerky in tune with the noises coming from the machine.
I don't remember the manual suggesting a limit to the size of the text file but there definently is one.

[BobWarfield]

Quite a few suggestions to increase lookahead out there if your machine is up to it. Syil for example:

What Type of Computer / Importance of Look-Ahead Lines

What's happening is lookahead allows Mach3 to forage ahead and fill the buffer for the parallel (or Smoothstepper) by looking further ahead in your g-code program. If the program has tons of simple short line moves as described, it's easy for the driver to suck the buffer dry which leads to some jerking around.

Too much lookahead can lead to bad results too, but it's easy to try an experiment or two to see what your system can handle.

Lookahead is also tied to the CV capabilities in Mach3. See, for example:

ArtCAM Forum • View topic - Artcam and Mach3

And lastly, here is a discussion of how things can go "jerky":

http://www.machsupport.com/forum/ind...topic=13784.10

Sounds like it has to do with Mach reverting to some "bailout plan" if the buffer to the driver gets too starved.

Aren't you glad you unlocked your Mach3? Lots of threads to pull on. Make sure the sweater doesn't unravel!

Seriously, that's the thing about unlocking vs relying on Tormach's tried and true config. All these parameters work together and can take some fiddling to keep them working in perfect harmony. It's all doable, and you can no doubt gain some performance. Just be ready for the learning curve.

Cheers,

BW

[ger21]

I was playing around with rotary machining and found that I could easily generate a NC file so big that it would bog the machine computer and make the machine go ca chunk ca chunk anytime the A axis was moving

You might want to try turning off the toolpath window, as it uses a tremendous amount of resources when large files are loaded, and can definitely affect performance.

[bobeson]

In my humble opinion, Mach3 is by far the weakest aspect of Tormach's product line right now, and I would love to see a replacement for it. I'm reluctant to try a roll-your-own approach with Smoothstepper because I already have my hands full just trying to get my own products out the door. I value the support I receive from Tormach, and don't really care to venture off into the woods of CNC machine building to get my parts built. So, what I'm really hoping for is a new motion control solution that is fully integrated with the PCNC by Tormach.

Hardware timing controllers like the Smoothstepper are not difficult to design, and are absolutely the right approach to use for motion control. A modern FPGA (programmable logic chip) can hold an entire machine control system in a single chip that can be reprogrammed at will by the end-user (for upgrades, bug fixes, etc.) I think this chip should reside on the Tormach machine control board itself, with a low-latency connection to the control PC. Skip the parallel port interface entirely; it's too limited. I'd recommend a point-to-point Ethernet physical link controlled by custom drivers instead of TCP/IP, in the style of EtherCAT. Many other possibilities exist for a good hardware link if you are willing to require a PCI-style plug-in card in the control computer; modern FPGAs feature lots of high-speed (>1 Gb/s) serial blocks, so these are easy-peasy nowadays.

However, what drives all of this hardware? Mach3? This would be an improvement over parallel-port bit-banging, for sure, but the many problems of Mach3 will not go away entirely with better signal timing. Although I hope that whomever is developing Mach right now succeeds in modernizing it, I suspect that it is time to look for a more suitable replacement.

Given all of the work that Bob Warfield has done with g-code interpreting with the g-wizard editor, I think he would be a good candidate for Tormach to approach about starting a company to develop a new control system from scratch. This would be a big undertaking, to be sure, but I think the time has come for a new contender to step into the control software ring, to pull CNC into the 21st century. I don't mean simply replacing the parallel port with something better, but moving to a better approach to machine control, hopefully one that incorporates feedback control at a fundamental level.

For example, since the Duality lathe has an indexing sensor, and thus knows the spindle RPM, why do I frequently stall the spindle while searching for an appropriate feed rate to program? If the machine can sense that the spindle is slowing down, why doesn't it just apply an automatic feed rate over-ride until the spindle speed stabilizes? This would be a very simple task if Mach were designed for such feedback control, but given the open-loop architecture of Mach it becomes difficult, and thus my Duality lathe is no longer used as a CNC machine.

I would also love to see a control system that applies signal processing to a high-resolution spindle load monitoring signal, in order to detect things like chatter and collisions. This would enable automatic feed and speed optimization for every cut, which would allow lightweight machines like the Tormach to be used up to their full potential. I'd also like to see probing incorporated at a fundamental level, so that automated in-process workpiece gaging can form a closed-loop process to implement automatic tool radius compensation, etc. In other words, SERVO EVERYTHING.

[bobeson]

BTW, while we're at it, junk the steppers and move to integrated-driver servos. I haven't done the research myself, but I'd be really surprised if you couldn't build a proper brushless servo with an integrated driver and encoder for very nearly the same price as the current steppers and discrete general-purpose stepper driver modules. The old-school thinking that steppers are inherently much cheaper than servos just doesn't hold much water any longer, given the effects of Moore's Law over the last few decades. And yes, I've read Tormach's white-paper about steppers vs. servos, but it basically just says "steppers are cheaper and good enough." However, servos are better, and need not be much more expensive, if at all.

If I were designing the next Tormach machine upgrade, it would include these features:
- axis motors are servos with integrated controllers, and 220VAC inputs and EtherCAT or similar for the control link
- motion control in hardware on the machine control board for 6 axis or more.
- mini-ITX motherboard control PC installed directly (no case) in the machine control cabinet, with solid-state HDD and 12V DC input instead of a standard ATX power supply.
- get rid of the big transformer and rectifier in favor of a small switching supply for the control electronics only. The main spindle uses 220VAC for VFD, and the integrated servo drivers on the axis motors could also use direct AC input.
- instead of a single probe input signal on the front panel, have multiple input and output ports available for accessories like probes, toolsetters, 4th axis indexers, automation control interfaces, etc.
- position sensor on the main spindle to allow vertical-lathe applications and better RPM control.

OK, I'd better stop now. This is how I procrastinate my real work

[davidperry3]

Everything you are saying sounds good to me.
I suspect though that way back when Tormach had to choose a course to steer the company down and is now limited to minor course deviations only.

The improvements you suggest would have a nice side effect of much fewer user interface problems (closed loop alone would have caught many of my issues).

I have heard it said here that excepting production work a $75K machine is not much better than a Tormach. With some electronics modernization it might be hard to say one is any better than a Tormach.

PCNC IV ...

[davidperry3]

You might want to try turning off the toolpath window, as it uses a tremendous amount of resources when large files are loaded, and can definitely affect performance.

I know this is a stupid question... How do you turn off the toolpath window?

[ger21]

Diagnostics page, below the toolpath window.