Connecting to a CNC machine, pros and cons of USB vs PP vs Ethernet etc.

[Jon.]

Both Ethernet and usb are digital so noise is likely to effect both if large enough, i have my uc300 running an inch away from my 2200w vfd and two inches from a honking great 55v toroidal transformer without issue so sounds like your noise may have been in your power supply, were you using ac filters? either that or there was something massively wrong happening in the next room.

[Jess]

however nobody uses it as a communication method between a computer and CNC machine

Not on Mach 3; but that part of the conversation hasn't been about Mach 3 for several days now! And, of course, that ethernet motion controller is a computer, but that's not what any of us mean.

On the LinuxCNC side, whilst RS485 is more popular, there's definitely people interfacing their machines to their computers with CANbus.

The hardware's been available for years. There's plug in cards that provide the interfaces plus industrial PC motherboards that already have CANbus and/or RS485 on board.

but you're then into a very glorified parallel port like system.

Rather unfair and misleading. CANbus gives you something much closer to a PCI I/O board in function - you're not having to do software step generation, and you actually have the ability to read encoders.

To seque this into the topic of the thread:

Approaches that allow your computer to more directly control the machine is definitely of huge value if you've got some complicated/non-standard kinematics going on, but even on a standard perpendicular axis mill then there's a potentially a significant financial saving to be made.

The additional costs associated with Mach 3; Windows licenses, Mach 3 licenses and ethernet hardware motion controllers could buy you all the electronics you need to convert something like a Novamill or Triac, even once you've thrown in something like a Mesa Anything I/O.

LinuxCNC on BeagleBone Black? A back of the envelope calculation suggests that I can have my entire dedicated control computer as well as software and breakout for the cost of the Mach 3 license alone; that's a lot of difference.

[m_c]

I wouldn't normally take the bait, especially from someone who has very little if any practical experience building or running CNC machines, but here goes.

And, of course, that ethernet motion controller is a computer.

No it's not. It'll be an embedded micro-controller, DSP or FPGA, which nobody who deals with such things will class as a computer.

On the LinuxCNC side, whilst RS485 is more popular, there's definitely people interfacing their machines to their computers with CANbus.

The hardware's been available for years. There's plug in cards that provide the interfaces plus industrial PC motherboards that already have CANbus and/or RS485 on board.

But not in the form of communication between a computer and a motion controller. In LinuxCNC the computer is the motion controller. External motion control cards are not available for LinuxCNC, as they defeat the whole purpose of LinuxCNC's realtime kernel.
CANbus/RS485 is just another method of communicating between the controller and drives/IO boards, just like lots of other possible options.

Rather unfair and misleading. CANbus gives you something much closer to a PCI I/O board in function - you're not having to do software step generation, and you actually have the ability to read encoders.

And what exactly is a parallel port?
Last I checked, it was simply a bunch of I/O pins conveniently arranged in a standard format, that can be accessed directly by software (I know windows blurs this, but the principle is still there), just like a PCI IO card does.
If you really wanted, you could make a parallel to CANbus to BOB setup, and it would give you a similar setup to using a direct CANbus PCI card, just with more limited IO and speed.

To seque this into the topic of the thread:

Approaches that allow your computer to more directly control the machine is definitely of huge value if you've got some complicated/non-standard kinematics going on, but even on a standard perpendicular axis mill then there's a potentially a significant financial saving to be made.

The additional costs associated with Mach 3; Windows licenses, Mach 3 licenses and ethernet hardware motion controllers could buy you all the electronics you need to convert something like a Novamill or Triac, even once you've thrown in something like a Mesa Anything I/O.

LinuxCNC on BeagleBone Black? A back of the envelope calculation suggests that I can have my entire dedicated control computer as well as software and breakout for the cost of the Mach 3 license alone; that's a lot of difference.

And this argument over cost will rumble on indefinitely.
The fact remains for your typical DIY CNC enthusiast, the familiarity of windows based computer systems, means they'll remain the most popular option for the foreseeable future, despite the additional cost.
You've got to remember, a lot of people who build CNC machines, are not computer or electronic geeks.

For me personally, time is money, so although I may save on licenses, the extra time familiarising myself with a new operating system/hardware, is time that could be spent making money. For me running Dynomotion products, a KFlop costs about £230 delivered. For that I get nearly everything needed for a motion controller including some very reliable software, and only need to add a windows PC, which can be picked up for very little. I can have the controller configured and ready to go in a couple hours. If the computer fails (which lets be honest, is usually the weakest link), I swap the configuration files onto a new computer, and away I go again.

Add something like Mesa cards into the mix, and you've then got to start swapping parts around, and hope they didn't get killed when the computer died.

[m_c]

Both Ethernet and usb are digital so noise is likely to effect both if large enough, i have my uc300 running an inch away from my 2200w vfd and two inches from a honking great 55v toroidal transformer without issue so sounds like your noise may have been in your power supply, were you using ac filters? either that or there was something massively wrong happening in the next room.

I suspect Jazz's source of interference was a HF Plasma Cutter, or HF TIG welder, both of which will produce and emit interference that will far exceed anything a properly wired VFD will.
If you get the chance, you might want to try firing one up next to your machine, and see if it still remains as stable, as there's nothing quite like a few hundred volt high frequency unsheilded arc to highlight the slightest deficiency in your shielding/grounding/power filtering/communications.

[Jon.]

Be interesting to find out. Though a grounded metal enclosure around the controller, which should be general practice, would most likely solve the issue. If it didnt id be suprised and very much suprised if the etherenet controller wasnt effected by such levels of noise.

[m_c]

It probably would be, however Ethernet is far better at handling and recovering from noise problems, as it can handle and resend lost data packets, whereas USB doesn't have that capability. I do suspect USB comms could be improved to handle these situation better, but it'll add cost, complexity, and processing overheads.

I know from various USS discussions, Greg struggled to fit all functionality into the USS, so I suspect it was having to do the bare minimum to handle communication. When it came to the ESS, Greg used a larger FPGA, and I do suspect there may be more going on within it to handle communication problems.

I also suspect other manufacturers have looked at the issues from the USS, and implemented strategies to better handle communication faults, as other good quality USB controllers don't seem to suffer from anywhere near as many communication faults.
You have to remember USS was the first successful USB external motion controller for Mach3, and it did achieve a lot, however I do think it's probably time to retire it, or at least update the hardware.

[JAZZCNC]

Both Ethernet and usb are digital so noise is likely to effect both if large enough, i have my uc300 running an inch away from my 2200w vfd and two inches from a honking great 55v toroidal transformer without issue so sounds like your noise may have been in your power supply, were you using ac filters? either that or there was something massively wrong happening in the next room.

No offense meant to you Jon but I'm doing this all the time and know all about what or not to do. The particular "next room" incident was actually a Compressor with faulty starter but the more the problem was the USB Smooth stepper had a design flaw so didn't take much to upset it. (It was one of the very first models.!! I've since tried newer versions but still had problems.!)

Regards the uc300 then fire up a Tig or Plasma cutter up near by and see how it reacts.! Vfd's and torodial supplys are nothing and it would be very bad state if they couldn't live along side one or both.
It's the things that go off around the machine that seem to affect USB/PC but what exactly I'm yet to determine and believe me I've looked and tried to put my finger on what causes it.!

End of the day I'm into building and using a CNC machine and haven't got time to chase the many things that can bring it down. What I can do is eliminate using those that don't meet the mark by testing them as I go along building machines.
First test is the Tig/Plasma test and then I work backwards from there.!!. . Very few USB devices Pass this test.!

LinuxCNC on BeagleBone Black? A back of the envelope calculation suggests that I can have my entire dedicated control computer as well as software and breakout for the cost of the Mach 3 license alone; that's a lot of difference.

Ah ah now we get to the meet of the Beef.!! . . . Linux Geek. . . Wanting to Pick a fight with Mach user's.!! . . . . Well Not happening I'm afraid been there too many times.! . . . Jog on.!

[JAZZCNC]

Be interesting to find out. Though a grounded metal enclosure around the controller, which should be general practice, would most likely solve the issue. If it didnt id be suprised and very much suprised if the etherenet controller wasnt effected by such levels of noise.

All the machines I build use Metal Enclosures and believe me it doesn't help much. Ethernet isn't affected nearly as much and better Motion devices like CSlabs it doesn't affect at all.
I'm just currently testing the Pokey's 57CNC which uses Ethernet and it survived the Tig test but then funnily enough it threw a little wobbly by E-stoping when I was using the Mig welder.? It's behaved since thou so could have been that and the fact i'm using the 5V signal for E-stop. (plus it's rough wired on the bench)

[Boyan Silyavski]

That thread made me laugh, really. How did i miss it. There is no respect for experience it seems. "Knowledge|" without experience is just a collection of facts or suggestions, from here and there.


I don't know what the hassle is about? Most expensive motion controllers have tried USB and then upgraded to Ethernet controllers. See Galil for example. Even cheap controllers have done the same. So there must be a reason for that.


What happens at low level / cheap/ controllers is that they do the same just like a follow up , after a couple of years, when parts are already cheaply available for the low end market.


Or you say people who make industrial control are stupid and don't know their business.


My blue Makita drill worth 360euro drills holes in steel and my friends green Bosch worth 50 euros drill holes in steel. My Makita can drill 10 years everyday without a glitch and the Bosch will die in month of industrial environment.


As i have pure HF plasma at home let me clear something. Glitches come from not good home grounding, the non shielded mouse and other Pc cables, not from USB or Ethernet cable if they are shielded and all electronics are in a shielded box. So basically both should work if everything to the last detail is well done.


Well Jess, now we know you . You have well presented yourself here. I am half expecting soon to hear that RP is better than ball screw and that HD alu profile gantry is stronger than steel one..

[Jon.]

Really id say its down to personal preference. I think if your using an Ethernet controller over usb because of noise its because you haven't taken the correct steps to shield the controller and that you are also happy to have your controller bombarded with noise, just because it can it doesn't mean it should. You may be correct in that ethernet has better protection against noise but that is providing the noise is entering via its ports. If your talking about noise that is infact capable of entering a grounded enclosure then there's no telling how that noise is going to enter the board, but you say it does fair better but with the correct steps either can work so again personal preference.

Bearing in mind a pc usually has multiple usb ports, but can only handle a single ethernet port so its restrained in that respect.