[baden0001]

Hello all,

I have been lurking around cnczones forums for a while, trying to gather information about EMC2 and CamSoft.

From what I gather EMC2 closes the servo loop at the CPU processor, while CamSoft can close the loop at a PCI based motion controller. So if the CPU crashes the PCI card will finish its commands but not continue on after that.

My question is this. For those people that have used EMC2, have you ever experienced a CPU failure and what was the effects on the machine that was being ran with it?

Thanks,
Will Baden

[samco]

If the computer/hadware is setup correctly.. The second the watchdog/charge pump circuit detects that the computer isn't there anymore - the machine will go into estop.

That being said. I can't think of one time that emc 'crashed' on me and or caused a bad part (that wasn't caused by me ).

sam

[baden0001]

Sam

How do you have the watchdog timer setup? Is it watching power to the pc, or actually watching a thread in the pc?

You say it is a charge pump, so I am guessing that you have the pc send an output to the charge circuit to drain the cap, but if the cap isn't discharged in time the cap will charge and possibly activate or deactivate a relay that is tied into the E-Stop circuit. Something along those lines?

Thanks for the fast response,
Will Baden

[Seb]

How do you have the watchdog timer setup? Is it watching power to the pc, or actually watching a thread in the pc?

The watchdog setup is dependent on the control hardware you're using.

I'm most familiar with Mesa's Anything I/O boards, which connect to the host PC via PCI or EPP. On these boards (with the hostmot2 firmware), the watchdog runs on the FPGA on the AnyIO board. The emc2 servo loop resets the watchdog each time it runs. The watchdog timeout is configurable from microseconds up to about four seconds. A watchdog timeout of a few times the servo period is normal.

With this setup, if emc2's servo loop stops communicating with the AnyIO board (due to the PC crashing, or emc2 crashing or locking up, etc), the watchdog will bite. When it bites, the watchdog safes the system by switching all the I/O pins to high-impedance mode, weakly pulled high.

The motor amplifiers "downstream" should use an active-low enable, so that the pulled-high "NotEnable" pin causes the motor to shut off.

Hope this makes sense...

[cadmonkey]

You say it is a charge pump, so I am guessing that you have the pc send an output to the charge circuit to drain the cap, but if the cap isn't discharged in time the cap will charge and possibly activate or deactivate a relay that is tied into the E-Stop circuit. Something along those lines?

A charge pump is really simple - it's a relay or solid state switch that is stand alone from the computer and all the other controls (ie it's its own device) and monitors a single pin on the interface for a certain pulse rate. If EMC is running properly that pin will continue to output the pulse rate you've setup in the configuration file (documentation is a bit sketchy on exactly what you get but typically the commercially available CP's aren't that tight on the frequency) and keep the relay (or SS switch) energized powering the drives. If EMC "gets lost" (which I have never had happen - my system specs P3 650MHz, 384MB, on board LPT for motion control and the CP, add on PCI LPT card for limit & homes, a sound card [hey, sometimes you want some tunes when you're doing maintenance...], and a PCI 10/100 card, driving 3 Gecko G203V drives through a DIY high speed opto isolating BOB) the CP signal will stop and the charge pump will in turn shut down the contactor which supplies power to the drives and the spindle, effectively stopping any motion. I've tested it by doing things like pulling a stick of memory, shorting the pin, disconnecting the pin, etc. and it worked each time.

It's nice to be able to boot the computer and do maintenance on the computer (updates, backups, playing some tunes, etc) and not having to worry about the mill being powered up. As soon as EMC starts, the mill is live and ready. Some people choose to have the CP also part of the E-Stop chain - such that if E-stop is pressed, then the CP shuts down also. If I were running big iron - I would too. But on my X3 I feel my sequence of operation is safe.

HTH - YMMV

Greg

edit - forgot I have onboard video too (typically a no-no but the bios on this machine won't let me turn it off so even if I put in a different card it still runs the onboard video.

And in case you got overloaded with my windy explanation I'll answer the thread question - EMC has NEVER died on me during a run or at idle, as long as the computer boots, it runs. Never had to hard reset or pull the plug - this is steppers not servos, but EMC is still counting pulses like an encoder, just instead of an encoder being used, the stepper configurations tie the step pulse generator to the counter so as far as the internal code goes, it's not a big difference.
/edit

[baden0001]

Thanks seb and cadmonkey, you guys answered my question about the watchdog timer. Both ways seem logical to me.

Also, cadmonkey, knowing that you havent had the operating system crash is what I was looking for in this thread. Thanks.

Right now the company I am working for is looking to retrofit an old Okuma LS-N possibly. We are looking for any and all options and trying to weigh them against safety, ease of use and how fixable in case of cpu failure.

Thanks,
Will Baden

[baden0001]

I'm most familiar with Mesa's Anything I/O boards, which connect to the host PC via PCI or EPP. On these boards (with the hostmot2 firmware), the watchdog runs on the FPGA on the AnyIO board. The emc2 servo loop resets the watchdog each time it runs. The watchdog timeout is configurable from microseconds up to about four seconds. A watchdog timeout of a few times the servo period is normal.

Seb,
I have been looking into this setup a little more and wonder what all drivers are needed for these cards (particularly 5i20) to run with EMC2?

Thanks,
Will Baden

[Seb]

There are two things called "drivers". One is software that runs on the PC and talks to the motion control hardware, and the other is electronics that provides power to the motors.

There are currently two software drivers in emc2 for talking to the 5i20. The older, more stable one is called m5i20. This one has been widely used for a long time, but it's somewhat limited in its capabilities. The other one is called hostmot2. This one is pretty new, but working well for lots of people.

These two drivers let emc2 talk to the 5i20 (hostmot2 supports all the other AnyIO boards as well). But the AnyIO boards by themselves can't drive motors, you need the other kind of driver for that: a power amplifier.

There are tons of options here, and your needs will dictate which you choose. Mesa sells amps for both steppers (7i32) and servos (7i33, 7i29, etc). Lots of other hardware manufacturers sell power amps, Gecko being perhaps the most well-known.

[baden0001]

There are two things called "drivers". One is software that runs on the PC and talks to the motion control hardware, and the other is electronics that provides power to the motors.

Sorry for being loose with my terms. I was looking for software drivers.

But since you did bring up Amplifier drivers for the servos, which did you end up using?

We are looking to use possibly the OKUMA drivers but that is looking like it wont work out. May end up using Advanced Motion Control Servo Drives.

Thanks,
Will Baden

[Seb]

But since you did bring up Amplifier drivers for the servos, which did you end up using?

My machine is still a work in progress... I haven't selected a servo amp yet.

I'm leaning towards the Mesa 7i29.

[samco]

the mesa hardware can do pwm+dir or +/-10v depending on the interface. That really covers the bulk of drives that are out there.

The new mesa driver (hostmot2) will also do hardware step/direction. (I would not use this though for servos as encoder feedback to emc is a much better solution)

[baden0001]

My machine is still a work in progress... I haven't selected a servo amp yet.

I'm leaning towards the Mesa 7i29.

Good luck on the build! The 7i29 looks like a good way to go.

Will Baden

[BMG]

Will,

If your looking for additional info on using EMC in an industrial setting, you might sign on to the EMC users list (https://lists.sourceforge.net/lists/listinfo/emc-users). There are a many active participants on that list that are in shops with small to big iron running EMC. Kirk Wallace and Stuart Stevenson are two of the most recently active with machine upgrades from older control systems to EMC.

I am a hobby user myself, but it is interesting to read the work they are doing presently with 5 axis Cincinnati.

[baden0001]

If your looking for additional info on using EMC in an industrial setting, you might sign on to the EMC users list (https://lists.sourceforge.net/lists/listinfo/emc-users). There are a many active participants on that list that are in shops with small to big iron running EMC. Kirk Wallace and Stuart Stevenson are two of the most recently active with machine upgrades from older control systems to EMC.

I recently joined that group and have been just reading the threads, and getting used to reading it.

I will look for more info from Kirk and Stuart, thanks!

Will Baden

[BMG]

here is the link to kirks stuff. http://www.wallacecompany.com/machine_shop/

I cannot recall Stuart's site. Its an aerospace jobshop near Witchita, KS. I'll post it if I see it again

[samco]

here are a few videos of stuarts 5 axis machine. He is using pico systems boards with existing amps. IIRC

[ame="http://www.youtube.com/watch?v=35tHYaDUmZQ&feature=channel_page"]YouTube - EMC2 cinci test cut[/ame]

[ame="http://www.youtube.com/watch?v=Nn1bJ3YAQdI&feature=channel_page"]YouTube - EMC2 5 axis cinci ball test[/ame]

supported hardware
http://wiki.linuxcnc.org/cgi-bin/emc...orted_Hardware

[skullworks]

Will

1st} I'll say I've benn running Okuma Mills and lathes since 1990 and just love them.

2nd} I think EMC2 is the best software interpreter option to retrofit old Iron.

BUT - Okuma (Not Okuma Howa) uses some very proprietary Absolute Position encoders as part of ther OSP controls systems. There is no "Home" just machine zero which can be a normal or very strange number.

Machine zero for our MC4VAE mill (OSP5020M) is the center of the table with the head all the way down at the Z- limit switch... Thats almost normal. Now an LB-15 (also a 5020) the Z zero point is somewhere 5 meters to the left of the machine and an equilly bizarre number for the X zero location.

To be honest your best bet might be the rebuild program by the company just down the street for Okuma NC facility.

I think that of all the iron out there Okuma will be the toughest nut to crack if your trying to retain the existing servos and amps. Finding the info you need will be difficult - make friends with your local Okuma service tech as he will be surely needed to obtain interface info.

[chester88]

I have an okuma flatbed lathe circa 1980s that i'm slowly converting.
After not being able to find info on or figure out how the absolute encoders work
(though it seems sorta like bubble memory), I removed the encoder and built an adapter to install Hiedenhain encoders. The existing Okuma amp use the tack to close it's control loop and EMC uses the encoder to close it's loop. Had to use the limit switches as home switches. For testing I had installed a linear scale on the x axis. I have a big problem as the (many) VFD's I installed to get around needing 3 phase power create interference with the tack feedback. I replaced the z axis with an ac servo motor (from ebay) because of a bad board on one of the amps. Having fun building it....
Anyone need a DC motor and drive from one?

[jmelson]

Hello all,

I have been lurking around cnczones forums for a while, trying to gather information about EMC2 and CamSoft.

From what I gather EMC2 closes the servo loop at the CPU processor, while CamSoft can close the loop at a PCI based motion controller. So if the CPU crashes the PCI card will finish its commands but not continue on after that.

My question is this. For those people that have used EMC2, have you ever experienced a CPU failure and what was the effects on the machine that was being ran with it?

Thanks,
Will Baden

Back in 1999 there was a bug in EMC that caused sporadic crashes of the linux OS and Xwindows manager. I had one of these while running the Bridgeport, and the screen froze, but the CNC system completed the part and stopped. I then had to reboot the computer.

Well, that bug was fixed in mid 1999, and I have never had a crash of the system or failure of EMC since then. We are now coming up on TEN years of use on that machine tool with EMC. I finally replaced the computer I was using on the Bridgeport as the newer EMC2 needs more resources, such as CPU speed and memory. I wanted to use the Axis GUI with the built-in preview.

I have had a few anomalies over the years, like the time I pushed the wrong button and brought a boring head down on my hand. And the time the lights went out for at least a quarter second, although it felt like a full second, while I was in the middle of machining something. The whole system rode through the outage without anything resetting or faulting (servo amps, VFD, computer, CNC control interface).

I know of few computer programs of any sort that have a reliability level like EMC.
I only use servo systems, and have a hardware E-stop system built into that, with a watchdog timer to cause E-stop if the computer locks up. You can test it by killing EMC while the servo system is out-of-estop, and observe the watchdog trip the E-stop. But, I've never had this happen except when I'm testing it.

Jon

Jon

[baden0001]

After not being able to find info on or figure out how the absolute encoders work
(though it seems sorta like bubble memory), I removed the encoder and built an adapter to install Hiedenhain encoders. The existing Okuma amp use the tack to close it's control loop and EMC uses the encoder to close it's loop.

Chester,

What lathe are you retrofitting? We are looking at the servo drives but are unsure about a couple connections to the drives. They look to be enables but somebody who has an LS-N said there wasn't any voltage across them. Dunno.

I thought there were absolute encoders on them and was planning on pulling them.

Thanks,
Will Baden