[Riceburner98]

Wow, good stuff! I thought the part numbers were odd.. Probably like the visions systems the place I work makes; base part is the same but customer can order different firmware versions, which changes the end part number..

It's great to see someone else using these drives. The only reason I can think that more people don't is they just don't know better.. The price on eBay isn't bad at all when they come up for what you get. I'm curious about the current doubling - did you replace any parts, or just tweak something inside? I doubt I'll need more than 4A per stepper with my fairly small machine, but you never know..

My fat fingers seem to have typed "Parker 6KE" instead of "Parker 6K4". Whoops. There's a few of them on eBay for ridiculous $$. I like Mach and have always used it (I did boot up an embedded PC with Unbuntu and EMC for about 10 minutes, but never actually ran a machine with it), I just wondered if the 6K4 drive would produce any smoother or "better" motion considering how much the thing cost new. Maybe it's just expensive because it replaces what a PC would do..? Though I guess you wouldn't use Mach to control an industrial machine anyway, so you'd need something like that.

I'll have to keep an eye out for Zeta drives on eBay.. I don't think the $100-ish ones are too expensive, but if I see any for less than that I'll have to think about picking them up just-in-case. I currently use IM483 drivers which I got on eBay, and now I have a dozen or so of those. They work WAY better than the Xylotex board I started with, for the same or less money. We still use them at work in our simple stepper-driven equipment, because you tell them over serial where to go. (I use the step/direction input of course.. Can't do coordinated movement with no way of syncing them.)

[HawkJET]

I thought the part numbers were odd.

They were originally sold as a motor/driver pair. The part number reflects the motor they were paired with. All the ZETA drives are the same, the difference is how the dip switches are preset from the factory. If you use a different motor that needs different current, you just change the dip switches....

The price on eBay isn't bad at all when they come up for what you get.

Geckos cost just over $100 per axis. But then you also need to buy a power supply. I got my ZETA drives for about $100 per axis, but I don't need a power supply. The ZETA drives are every bit as good as Geckos (and vise versa).

One thing you need to watch though, since the ZETA drives use rectified line voltage (170 volts), it is easier to overheat the motors (since it is easier to get higher speed performance). I am building an industrial machine and pushing the motor pretty hard. So I have a fan, heatsink and temperature sensor/interlock to protect the motor.

I'm curious about the current doubling - did you replace any parts, or just tweak something inside?

I have worked for various motor companies for most of my career (I was one of the original employees of Compumotor). So I have a good sense how things work. The ZETA drives are a 15-20 year old design, so the MOSFETs are that old too. I replaced the MOSFETs with newer technology parts with 1/3rd the on resistance and double the voltage capability. With these new parts, the heat generated by the driver is less than original, so the heatsink can handle the higher current. I also had to change the current sense resistors to "fool" the firmware into increasing the current.

Maybe it's just expensive because it replaces what a PC would do..?

They are expensive because, back when they were designed, Mach wasn't around (or at least is was barely known). Also, in those days, the customers for such devices were generally corporations (not hobbyists), so they charged more, plus the market was much smaller. Since the advent of the likes of Gecko, Compumotor is really struggling. Gecko knows how to make a product that is both, very high quality, AND inexpensive to manufacture. Unfortunately, Compumotor did not. The items on eBay are expensive because Compumotor controllers are so specialized that if you have one that fails, it HAS to be replaced with another Compumotor (so they have you by the short hairs).

Though I guess you wouldn't use Mach to control an industrial machine

There is no reason NOT to use Mach3 on an industrial machine. I am currently building a machine to automate the production of a carbon fiber product. That machine is controlled with Mach3 (I started with EMC2 but found that Mach3 is FAR EASIER to customize for special applications). There are a number of turnkey CNCs that use Mach3 (I believe Tormach does).

I'll have to keep an eye out for Zeta drives on eBay.. I don't think the $100-ish ones are too expensive, but if I see any for less than that I'll have to think about picking them up just-in-case.

The $75 price is very rare, which is why I almost bought it.

[sansbury]

(I started with EMC2 but found that Mach3 is FAR EASIER to customize for special applications).

That depends a lot on the application. If you need to build a few custom screens to control a typical machine with cartesian kinematics, then yeah, definitely easier. It will beat EMC in the quarter mile every time. Where EMC shines is when things get a lot more complicated, whether you're doing closed-loop control, complex kinematics like a robot arm, rigid tapping... that sort of thing.

As an aside, I started switching one of my machines last night from Mach to EMC, and right off the bat, I'm able to sustain slightly higher speeds than I could with Mach. Whether this is the PC or the software is impossible to say for sure, since the hardware is different, but I'm not the first person to find this kind of thing.

Likewise, I've long felt that Mach is just plain buggier than EMC. In EMC, things seem to mostly either work or throw errors. Mach seems to have more scenarios where the machine wanders off in a totally unexpected way.

[HawkJET]

That depends a lot on the application. If you need to build a few custom screens to control a typical machine with cartesian kinematics, then yeah, definitely easier. It will beat EMC in the quarter mile every time.

This machine has one linear axis and two rotational (take-up and pay-out reels). The standard Cartesian screens are of little use.

Where EMC shines is when things get a lot more complicated, whether you're doing closed-loop control, complex kinematics like a robot arm, rigid tapping... that sort of thing.

Well, maybe... here is a thread that shows Mach3 running closed loop and doing rigid tapping...http://www.cnczone.com/forums/bencht...g_1000rpm.html

As an aside, I started switching one of my machines last night from Mach to EMC, and right off the bat, I'm able to sustain slightly higher speeds than I could with Mach.

My experience was the complete opposite. I was able to more than double my speed when going to Mach3.

Mach seems to have more scenarios where the machine wanders off in a totally unexpected way.

It has never happened to me and I have not heard of this. Mach works flawlessly for tens of thousands of users. This seems like superstition to me....

[SCzEngrgGroup]

It has never happened to me and I have not heard of this. Mach works flawlessly for tens of thousands of users. This seems like superstition to me....

Hawk,

Nope, far from superstition. Come by my place sometime, and I'll demonstrate some of the many problems in Mach3. Hang out on the Mach3 forums, and you'll see how many people have weird things happen. I've personally reported, and helped track down and fix, a dozen or so mostly pretty serious bugs in Mach3, and there are several more that have been proven, and acknowledged, but not yet fixed, even after several years.

Want to see a good one? Configure your machine so the Z axis acceleration is much less than X/Y (not at all unusual), then program a linear move in X/Y that then transitions into a helical move in X/Y/Z. The Z acceleration limit is not respected, and, on a real machine, the Z axis will fault, or lose position. That one cost me a $100 endmill the first time I encountered it.

If I had a nickel for every time Mach3 has just flat weirded out, and gone "off script", I coulda bought me a VMC by now. I've got quite a collection of broken tools and probes from random, non-repeatable moves Mach3 decided to make all on its own. If I ever have to pause a program, and re-start from some intermediate point, and see that "preparatory move" dialog come up, I STOP, exit Mach3, and re-start, as those moves are virtually NEVER right, and will do such interesting things as send the machine completely outside it's working envelope, or rapid a tool right thought a vice or clamp, taking it completely outside the bounds of the program being run.

And this is across at least three different complete systems, some servoed, some steppered, some with SmoothStepper, some without, completely different PCs, drivers, BOBs, etc, and spanning well over 5 years. Never once had any of these problems with TurboCNC, BTW. Many people never have a problem, many people have frequent problems. I am, sadly, one of the latter, and am seriously considering going to EMC as a result. Mach3 has great potential, but it really is not all that stable, all things considered. And each new release brings a new raft of problems.

Regards,
Ray L.

[HawkJET]

Nope, far from superstition.

Interesting....

Both Mach3 and EMC2 are complex programs. It seems that the more complex the program is, the more likely there will be a large number of users with absolutely no problems and those with several - as if they were using different programs altogether.

Ray, sorry to hear of your issues. Good luck if you switch to EMC2.

My sense is that both Mach3 and EMC2 are very comparable, although they have different design philosophies. It is really just a matter of which one fits your application best.

[sansbury]

This machine has one linear axis and two rotational (take-up and pay-out reels). The standard Cartesian screens are of little use.

Screens and kinematic models are *totally* different issues. Let's say you have a SCARA arm. In this case, moving the actuator at the end of the arm from X=0 to X=1 is not simply a matter of spinning the X-axis leadscrew 5 rotations, you may have 3 or 4 axis moves needed depending on the degrees of freedom on the end of the arm and whether you want to keep the orientation the same. Serial kinematics are a classic example of this.

With Mach, you can sort of handle this, but you need to pre-calculate all the individual axis moves, and even then, you may not be able to keep them perfectly synched during the move, which may or may not matter. With EMC, the motion controller can have a model of the arm, and you can simply issue G1 X1 F5 and it will do exactly that.

Now, does this matter to 99.8% of people here? Not really

Well, maybe... here is a thread that shows Mach3 running closed loop and doing rigid tapping...http://www.cnczone.com/forums/bencht...g_1000rpm.html

As I posted in that thread, I'm curious to understand a little better what's really going on under the hood. That said, the test of what he built isn't what the book says, it's whether the thread comes out right. If it meets his needs, then it works for him.

My experience was the complete opposite. I was able to more than double my speed when going to Mach3.

Yeah, you definitely had a lot of trouble with EMC. This piece of the stack is still a huge mess. I've always had good luck with getting machines up and running but I know enough really smart guys who've had a hard time to know it's a mess. The fact is that using a PC to generate steps is a hack. Art's printer driver is one hack and the Linux RTAPI module is another. It may be a more formal and in some cases better hack, but it's still kind of a hack.

It has never happened to me and I have not heard of this. Mach works flawlessly for tens of thousands of users. This seems like superstition to me....

Well, I don't know that I'd ever use the words "flawlessly" and "software" anywhere near each other

This is a hard situation because there's enormous variation. You have different PCs, then you have different machine configurations, then you have different G-code. Even if you just saw something weird happen on your machine, it can be hard to figure out what got you there, whether it was something you did or the controller, and replicating the bug can be really hard.

Neither Mach nor EMC are anywhere close to perfect. They have different flaws and strengths. If I was a custom machine builder, I can easily imagine projects where I'd choose one or the other depending on what I need to do.

[HawkJET]

Screens and kinematic models are *totally* different issues.

Yes, I agree. I was on a different page in my reply.

Well, I don't know that I'd ever use the words "flawlessly" and "software" anywhere near each other

BUSTED! You really got me on this one

Neither Mach nor EMC are anywhere close to perfect. They have different flaws and strengths. If I was a custom machine builder, I can easily imagine projects where I'd choose one or the other depending on what I need to do.

Very well put.

[ihavenofish]

mach3 sucks. emc sucks. there. i said it.

mach3 has serious issues on contouring at high speed and high detail. bascally it cant figure out how to decellerate when the command is very short - 0.1mm or less in my case. this shows up when you have a sharp corner in your program. it will overload the stepper and cause it to stall. its a massive part and tool killer.

emc has no such issue, however its constant velocity contouring allows it to deviate from the requested tool path ridiculously causing issues like facets, gouges, and out of square cuts. on the bright side this can be tuned out at the expense of speed, without risking damage to your part, tool or machine.