New Router: Mach3 / Smoother stepper - Suck

[dkemppai]

Well,

Starting to get the new router table running. Was using an optical centering scope to square the gantry and learned of a Major flaw in Mach3/Smooth stepper.

It turns out that if you Jog an axis with the shift key down, and let up on the shift key before the jog key, the axis keeps going. Nothing short of an E-stop will stop it.  Found this out while moving the Z axis down with my several hundred dollar centering scope installed. Yep, smashed my scope. The only thing it's good for now is the trash can! After a little research, this turns out to be a well known bug. It appears that many motion control boards suffer with this problem. However in about a year of knowledge of this bug, no one has fixed it yet. 

Let me not forget that also today, mach/smooth stepper at random decided to move only one of my axis's without moving the slave axis (when everything has been working fine for hours), thus torquing the gantry. Also a separate time when jogging few jog steps, it decided to slowly start moving several axes at once. There are times at random won't load G code files correctly, with a reboot required to fix the issue. I'm not impressed.

Overall, I'm kicking myself for not spending the few extra dollars for a Flashcut board. Yeah, Flashcut has had a few bugs too, but as soon as I report one I get a response with fixed code back. It's a lot more stable, a lot more responsive, and doesn't look like my 5 year old designed the screen with a box of crayola crayons.

I'm real close to shooting a couple of .45 holes in that damned smooth stepper board, smashing it with a hammer and calling flashcut. I've never been so disappointed with anything I've spent money on before. In my opinion Mach3 is bug ridden garbage software. Now I know why they've had to spend years writing mach 4 from scratch, they have nothing but junk to start with. We've had some issues with USB smooth stepper. It's hardware shows promise, and I believe it's crippled by Mach3. Maybe mach 4 will make it a worth while board. 

Anyone else have similar feelings about Mach3? Or is it just that an EE with 15 years of PC programming and electronics hardware development experience isn't smart enough to figure out a simple software package! 

Dan

[lloydsp]

I've heard other such horror stories about Mach3.  It is, after all, just a commercialized 'ripoff' of EMC^2, which (being open-source and crowd-sourced) may have had fundamental problems.

I'm sorry that happened to you. 

I had considered doing my own conversion (being also an EE and firmware programmer), but finally decided I couldn't afford the time.  I'm glad I bought the Ajax control and Centroid, even if they do charge for each add-on feature.  Except for a couple of minor 'irritations', it's been rock-solid since the day I turned it on.

LLoyd

[dkemppai]

I've heard other such horror stories about Mach3.  It is, after all, just a commercialized 'ripoff' of EMC^2, which (being open-source and crowd-sourced) may have had fundamental problems...

...I'm glad I bought the Ajax control and Centroid, even if they do charge for each add-on feature.  Except for a couple of minor 'irritations', it's been rock-solid since the day I turned it on.

Lloyd,

I'm sure you made the right decision. I've seen some bugs in other 'higher' end CNC controls also. In my limited CNC experience, flashcut seems to be the most stable. Even SW industries stuff has shown some pretty serous glitches. Again, I've never been on a good machining center or high any end stuff.

If you don't mind me asking, about how much did the Ajax and Centroid cost? (Just a ballpark number, if you're willing to share would be great!)  I'm sure it's out of my price range for this router, but may be worth it for other things in the future.

I'm real close to ordering a flashcut board. At about $1200 it wasn't that much more than Mach3/Smooth Stepper/PMDX-126 with speed control. Add a $200 centering scope and we're almost there! With so little time with mach3, I can see the 'writing on the wall'. I'm not going to put anything valuable on that router, as long as mach 3 is running it. Wood and cheap bits are it! 

It's time for a large glass of strong alcoholic beverage now!

Dan

[Bob La Londe]

I've got two SS/M3 machines running, and while I have had the SHIFT/JOG issue they have overall been pretty decent controls.  I did take the time to optimize each of them for Mach 3. 

I don't have anything setup with slave axis, but I do plan to setup a 5x10 CNC Mogul or similar for sheet goods sometime in the next year.  I'll have to learn to deal with that issue.  On a machine that big I doubt I could get away with a cross table belt drive. 

I doubt that there will be any new bug fixes for Mach 3.  They have been spending all their time working on Mach 4 for the last two years.  Mach 4 has already been released to OEMs, but the uber configurable version for DIYs is still in Beta.  It will supposedly cure the display issue when running large files if you have a multi core machine, by running the code on one worker thread, and driving the tool path display with another. 

[Dragonfly]

I wouldn't blame Mach3. It's been used all over the world and despite some known bugs/limitations will not cause sporadic moves by itself. Those things are usually related to user setup, hardware and noise in the electronics.
For example I had a nightmare moment last year when all of a sudden a random stepper would seize and stop moving. On a working for many ours setup which has not been changed. In short finally when I opened the PC I found that the CPU cooler plastic base had broken at one side and the radiator detached from the CPU chip. Thus causing the P4 to enter a built-in protective mode by inserting empty cycles to cool down.
On such a machine I usually turn off all power saving modes, no anti virus, no network connection (especially when controlling the router), no funny wallpapers and gadgets which consume resources and processor time. Power supply from the same power rail.
(I am using the parallel port.)

[lloydsp]

If you don't mind me asking, about how much did the Ajax and Centroid cost?
---------------
The Ajax All-In-One-DC board with the Centroid control software is $2399.  It's for servo-based machines.

The board comprises a 3-axis servo system with encoder feedback, a spindle controller, all the aux function controls (mist, flood, remote pendant control), and all of the machine-local cpu and memory.

I spent way more than that, but I also bought three brand new servos with high-resolution encoders,  a new power supply, a new Windows 8 computer, and a remote jog pendant.

Those were the bulk of the cost, bringing the total up to about $8000.

Lloyd

[Jeff_Birt]

Other than the shift/jog problem the rest of the problems sound like wiring and/or configuration. I have built hundreds of systems using Mach 3 with the SmoothStepper and they work well. Randomly moving or not moving axis is most likely loose wiring or electrical noise interfering with the system.

[Jeff_Birt]

I've heard other such horror stories about Mach3.  It is, after all, just a commercialized 'ripoff' of EMC^2, which (being open-source and crowd-sourced) may have had fundamental problems.

I have to throw the BS flag on this one Lloyd. Mach3 started out many years ago with a little program called 'Master', i.e. Master 1, Master 2, etc. This was back when Art (the guy that wrote it) did not even know GCode existed, he just wanted to control a machine with his PC. After getting feedback from other people using it he went on to create the Mach series. None of these programs were or are a 'commercial rip off of EMC2 a.k.a. LinuxCNC'. The parallel port drier Art created for use in a Windows environment is an entirely different beast than the real time kernel extension that LinuxCNC uses. Mach has also always used an entirely different type of planner than LinuxCNC, Mach plans ahead so it can do CV and bland moves whereas LinuxCNC has been until very recently a one line look ahead planner. I know Art did get the idea for plug-ins from some LinuxCNC guys at a show, they were talking about how cool plug-in would be and Art said to himself, "Hey I bet I could add that too." I bet a lot of folks have looked at the LinuxCNC code, and the original EMC code (which was a US government (NIST) project originally) and gleaned some ideas, but to say they are 'rip offs' is a slap in the face.

Just some background info...

[lloydsp]

Ok, Jeff.  I'll stand corrected.  Some folks who ought to know told me this, but it could just be urban myth.  If it is, I'd say it's a wide-spread one.

Lloyd

[dkemppai]

I wouldn't blame Mach3. It's been used all over the world and despite some known bugs/limitations will not cause sporadic moves by itself. Those things are usually related to user setup, hardware and noise in the electronics.

Hi, I'll look foreword to your comments following.

1. In the case of my smashed optical center finder, I have to blame Mach 3. It turns out this shift/jog issues is a KNOWN problem. It's been known for a while, and has not been fixed. It's also not well documented, which is horrible given the nature of the problem.

2. Can you elaborate on what the know bugs/limitations are? It would be nice to hear from other users what issues have already been discovered, so I can avoid expensive problems in the future.

3. As for electrical noise, I'm using the PMDX-126 board with smooth stepper attached. All I/O to the boards are optically isolated. Step/Dir pulses are optically isolated on the drive side. Spindle VFD speed analog and Run/Rev lines are optically isolated on the PMDX-107 board. Limit switches come into isolated inputs on the PMDX-126 board. Even the ERR/RST line to the Gecko drives is optically isolated. Electrical connection are as short as possible. Anything with high voltage and PWM is routed well away from anything that is an input or sensitive to RFI. Ferrite chokes were applied to the VFD for the spindle, and a well shielded cable is also used for everything.

I have not yet thrown my scope on every input or output to look for noise, however my gut feeling is that with everything optically isolated common mode noise can not be an issue. Differential noise should not be a problem as there are no big loops to pick up noise going anywhere

This being said, I'll continue to look for electrical weaknesses. I'm not discounting the possibility it is noise, but there's so much galvanic isolation and shielding it is hard to imagine.

Anyway, I'm going to continue to play with it. There are still lots of thing that need final configurations, so maybe some settings are not right. Any input would be greatly appreciated.

Thanks,
Dan

[dkemppai]

Other than the shift/jog problem the rest of the problems sound like wiring and/or configuration. I have built hundreds of systems using Mach 3 with the SmoothStepper and they work well. Randomly moving or not moving axis is most likely loose wiring or electrical noise interfering with the system.

Hi Jeff,

FYI, most of the weirdness seems to happen when we're using keyboard commands to jog or move the machine. As long as the G code loads correctly, it seems to run G code fine. Maybe there's something in the setup I don't know about that will affect keyboard response???

As for your builds, do you happen to have some info on your setup you could share. Any info on Sevo/Steppers, and what brand drives? Which break out board with SS, and what sort of step/dir pulse rates are you using? Are you using any pulse stretching in SS to ensure pulses are long enough? Any info you could provide would be great!

As for this build, the system runs one 1200 in-oz stepper for the Z axis, set for 2000 steps micro stepping. There are 2 1125 in-oz servos for the gantry and one 1125 in-oz servo for the other axis. Encoders are set for 1000ppr (or 4000 quadrature states per rev), and the drives are Gecko servo drives. 2:1 belt reduction from the servos to the screws. Smooth stepper is set for 250Khz step dir rates.

I'm pretty sure there is no loose wiring on anything similar. That has been double checked.

Thanks,
Dan

[atwooddon]

I will throw something in here just as a possibility.....  I use Mach3, have for years.  My CNC jobs are more hobby related than any type production environment so they are typically one off projects.  I have experienced the 'keep moving' or 'start moving' error which was very puzzling until I began to look at what did I change in the system.

I had added a USB connected Xbox controller so I could move and adjust the machine easily.  At first it worked very well.  One day, I had left the room with the CNC machine idle and when I returned, it was jammed at one end of the Y axis.  What the *****?   Turned everything off, reset and no more problems that day.   Several days later, I was mounting a piece on the table and the Y axis started creeping, all on it's own.  I grabbed the Xbox controller, touched the Y axis control and the movement stopped.  At first I suspected some type of electrical noise on the USB connection, rerouted the USB cable but the problem returned.  I had a spare Xbox controller so I swapped them and the problem went away for a while only to return weeks later.  I took the controller apart, cleaned it but the problem still happened.  Calibrating the controller in Windows makes the problem go away for a while but will return.  The problem has happened on all axis and will only get worse over time.  The potentiometers in the Xbox controller are basically 'junk' at best, not designed for any type long term accuracy on centering.

Just thought I would throw out the idea of checking any external inputs you might have for Mach 3.  In my case, Mach3 was doing exactly what it was being told to do. 

Don

[lloydsp]

This being said, I'll continue to look for electrical weaknesses. I'm not discounting the possibility it is noise, but there's so much galvanic isolation and shielding it is hard to imagine.

Anyway, I'm going to continue to play with it. There are still lots of thing that need final configurations, so maybe some settings are not right. Any input would be greatly appreciated.
-=-

Dan, obviously you have a good grip on potential sources of noise.  The only other things I would suggest is using twisted-pair for any runs over about 4' long, and making sure signals are terminated at both ends.  I've seen a lot of opto-isolators go wonky when they were driven by an open-collector output terminated only at the driving end, and using no pullup (for active-low) termination at the opto end.

Don's experience expresses the folly of using ANY pots at all for position control, unless you get some of those nice mil-spec 'centering' pots that have a large conductive land at the 'zero' position, so they cannot creep.

Even then, if the pot has another resistor in series with it to adjust the voltage swing, and the resistor and pot don't have exactly the same temperature coefficient, you can see drift.

Lloyd

[Garyhlucas]

I had the shift jog runaway when my 10 year old grandson was at the keyboard. Full enclosure, door interlocks, not any danger to him.  Since I reported it others also were able to make it happen.  I disabled the jog keys and use the popup jog pendant only.  No more problems with that. Mach 3 ran away again after a tool change a few days ago. Something in the code it doesn't like. It is a repeatable problem, if I load that file it will do it again.

However, Mach 3 is built on top of Windows, running on many different versions, on lots of different hardware and we all saved thousands over a dedicated CNC.  If you are a business you should never use Windows for a mission critical task unless you have lots of backups including all the hardware.

So I am amazed at how well Mach 3 works, most of the time!

[Jeff_Birt]

Dan,

What sort of encoders are you using? Gecko servo drives tend to be picky about encoders as not all encoders produce nice clean signals or have a nice strong drive signal. As I recall there was a particular issue with some US Digital encoders but I might be remembering the brand wrong. Gecko has some info about this on their website. When you saw an axis drifting was the corresponding DRO in Mach3 changing count too? If not then it is most likely an encoder issue, i.e. the drive is thinking it is seeing the encoder move so it is trying to compensate and you get drift.

So you see any strange things happen when the VFD is not in operation? VFDs are great for controlling a three phase motors but they also make great electrical noise generators and can interfere with even the most high dollar controls. The noise generated can be transmitted through the air or it can be transmitted through the AC line feeding the VFD. At a minimum I use a line filter on the AC feeding the VFD and use proper VFD cable between VFD and spindle motor. A line filter on the AC feeding the PC/control box is also a good idea.

Also, the power supply on the PMDX-126 is barely adequate for the USB SmoothStepper it will not properly power an Ethernet SmoothStepper. I recommend a separate 5V power supply for either type.