[grubscrew]

Now that my machine is up and running, I decided to put it through its paces with a fairly complex 23"x23" 2.5d carving. I've ran the piece twice and have the same results both times. the Z axis appears to be loosing steps in the + . over the course of the cut the z axis gained +.13". the bit is barely engaged in the material near the end. The finish pass uses a 1/4" box core cutter at 10% step over resulting in a g code around 750,000 lines.
I have chucked up the dial indicator and my spoil board is within +/- .004 in the y axis.
I wrote a code to cycle the z axis .5" up and down. I ran it for 30 minutes at 100ipm. at the end there was no variance on the dial indicator.
I reran the code with the spindle running at 22,00 rpm to rule out interference.

My current set up:
fineline automation 48x24"
nema 34 960oz/in 1/2-10 5 start on the z axis
g201x gecko drives
pmdx-126 bob
ESS smooth stepper

All my research on this problem has brought up two definite reasons that as far as I can tell are not related to this issue. The first is a problem with one of the old gecko servo drivers, and second is a parallel port problem.

Has anyone else experienced this and found a answer? I'm left scratching my head as I can not replicate it just testing the axis individually.
Is this something that can be related to mach 3's constant velocity settings?
The other thing I should mention, occasionally the machine shuts down due to the smooth stepper "buffer warning" - could this be related?

for reference here is a picture of the last test, you can see near the end of the cut (gouge side) that the z has risen and not fully cutting the shapes. don't mind the giant gouge either, aspire decided to add this lovely tidbit to the end of the code:
X23.0000Z0.0000
Y22.9934
X22.9770Z-0.5030
X0.0000
G0Z0.2000
G0Z0.5000
G0X0.0000Y0.0000
M30
It really was doomed from the start!!

Any help or direction would be greatly appreciated!
Thanks!

[rich909]

Hi Grubscrew --

Now that my machine is up and running, I decided to put it through its paces with a fairly complex 23"x23" 2.5d carving. I've ran the piece twice and have the same results both times. the Z axis appears to be loosing steps in the + . over the course of the cut the z axis gained +.13". the bit is barely engaged in the material near the end. The finish pass uses a 1/4" box core cutter at 10% step over resulting in a g code around 750,00 lines.
I have chucked up the dial indicator and my spoil board is within +/- .004 in the y axis.
I wrote a code to cycle the z axis .5" up and down. I ran it for 30 minutes at 100ipm. at the end there was no variance on the dial indicator.
I reran the code with the spindle running at 22,00 rpm to rule out interference.

My current set up:
fineline automation 48x24"
nema 34 960oz/in 1/2-10 5 start on the z axis
g201x gecko drives
pmdx-126 bob
ESS smooth stepper

All my research on this problem has brought up two definite reasons that as far as I can tell are not related to this issue. The first is a problem with one of the old gecko servo drivers, and second is a parallel port problem.

Has anyone else experienced this and found a answer? I'm left scratching my head as I can not replicate it just testing the axis individually.
Is this something that can be related to mach 3's constant velocity settings?
The other thing I should mention, occasionally the machine shuts down due to the smooth stepper "buffer warning" - could this be related?



Any help or direction would be greatly appreciated!
Thanks!

I've seen this 'problem' when motor tuning has not been done during Mach-3 set-up/configuration. Since you didn't mention 'tuning', I offer this as another item for your check list. You indicated that when driving a single axis (Z) the problem does not exist. This makes it more likely that servicing all three axes simultaneously during the running of the job is resulting in lost steps.

Another thing to consider is the weight of the router. You are loosing steps on the '+' direction, which could result if the lift platform is not neutrally balanced with a heavy router. I've also seen steps lost when there are flakey wires/connections to the stepper motors. You may have already worked through these possibilities, but if not here are a few more ideas to check.

Finally, the 'buffer warning' could definitely be related. If the buffer is overflowing, steps will likely be lost. I don't have personal experience with the Smooth Stepper, but you should seek help at: Index If you haven't already spotted it, you can get doccumentaiton and view the FAQ on the ESS Smooth Stepper there as well.

HTH.

--Rich

[ger21]

Cut your Z axis acceleration in half and see if it makes a difference.

[grubscrew]

Hi there,
Rich
Thank you for the suggestions,
- all 3 axis steps per have been accurately set, as well as the backlash (aprox .002 on each axis)
-Im not sure how the weight of the spindle could be effecting it as at the end of the cut the z axis is higher than the starting point on the material, I as well have a gas strut that should assist in lifting 30 of the 32 pound z axis assembly. the z axis steeper should have a easy life on the machine
-the wires are factory cnc router parts cables, but I will ohm them all out to ensure the ground is not broken on one of them,
-I did not realize warp 9 had a forum, i will defiantly sign up
Thanks!

[louieatienza]

Run your pmdx breakout board straight to your computer parallel port and see if you have the same problem, or if you don't have one you can but a parallel port card cheap.

A better z test would be to plunge 100 holes and see if they're all the same depth. You don't mention how much voltage you're driving that monster z stepper or the drive you're using.

[grubscrew]

Run your pmdx breakout board straight to your computer parallel port and see if you have the same problem, or if you don't have one you can but a parallel port card cheap.

A better z test would be to plunge 100 holes and see if they're all the same depth. You don't mention how much voltage you're driving that monster z stepper or the drive you're using.

Hi there, Unfortunately my control computer is one of those dell minis with no room for expansion. if it comes down to it i will track down a old computer.
-I am running the stepper with a gecko 201x and 48v 21A power supply. ( the CNCRouterParts N34 DIY system)

[louieatienza]

Hi there, Unfortunately my control computer is one of those dell minis with no room for expansion. if it comes down to it i will track down a old computer.
-I am running the stepper with a gecko 201x and 48v 21A power supply. ( the CNCRouterParts N34 DIY system)

Even 48v will limit usable torque for larger steppers under speed. The lower accel can help. I run an 8 start (1 tpi) screw with a 25 lb carriage and I'm only using a 425in-oz stepper, and easily rapid over 250ipm on the z. Have you also tried running without the gas strut assist?

[grubscrew]

Even 48v will limit usable torque for larger steppers under speed. The lower accel can help. I run an 8 start (1 tpi) screw with a 25 lb carriage and I'm only using a 425in-oz stepper, and easily rapid over 250ipm on the z. Have you also tried running without the gas strut assist?

my current z settings under motor tuning is 250ipm with a 30 accel. The problem is strange as when running the z axis by itself (program runs up.5 back to zero, up to .5 back to zero,) 1500 times i see absolutely no change in position, the problem only presents when a larger code is ran.
the tests code was ran at 100ipm as thats what the 2.5d was set at (350ipm 100 plunge)
I just ran 10 minutes of the same code with no gas assist, at the end there was still no variance in positioning. (I even tried leaning on and applying a fair amount of weight to the z carriage to try and make it skip)

right now after reading the warp 9 forums a little bit, one big thing i noticed is that the default setting in the smooth stepper general config page all the axis are set to 256khz . From my understanding if they are higher than needed it effects accuracy.
from the math for my axis :

Steps/Unit * Units/Min = Steps/Min
Steps/Min / 60 = Steps/Second (or Max Step Frequency)

4010*250(max) = 1,002,500
1002500/60 = 16,708
16,708/1000 = 16.708 KHZ
math for x and y was around the 20 khz.

that being said I set all the axis to the lowest setting (32khz) and am in the midst of checking all the axis calibration and backlash ( apparently required after changing frequency however have not found any discrepancy's yet)

[louieatienza]

my current z settings under motor tuning is 250ipm with a 30 accel. The problem is strange as when running the z axis by itself (program runs up.5 back to zero, up to .5 back to zero,) 1500 times i see absolutely no change in position, the problem only presents when a larger code is ran.
the tests code was ran at 100ipm as thats what the 2.5d was set at (350ipm 100 plunge)
I just ran 10 minutes of the same code with no gas assist, at the end there was still no variance in positioning. (I even tried leaning on and applying a fair amount of weight to the z carriage to try and make it skip)

right now after reading the warp 9 forums a little bit, one big thing i noticed is that the default setting in the smooth stepper general config page all the axis are set to 256khz . From my understanding if they are higher than needed it effects accuracy.
from the math for my axis :

Steps/Unit * Units/Min = Steps/Min
Steps/Min / 60 = Steps/Second (or Max Step Frequency)

4010*250(max) = 1,002,500
1002500/60 = 16,708
16,708/1000 = 16.708 KHZ
math for x and y was around the 20 khz.

that being said I set all the axis to the lowest setting (32khz) and am in the midst of checking all the axis calibration and backlash ( apparently required after changing frequency however have not found any discrepancy's yet)

I've tested and advocated that a SS is not necessary unless you have a system that requires a very high step count. Even then an EPP parallel port can have a throughput of 2MB/s, though the real bottleneck is the processor and whatever overhead... As to your stepper methinks 30 accel and 350 rapids might me too high...

[grubscrew]

Louieatienza - There was a typo in my previous post z was at 30 and 250 -however the project is running at a z plunge of 100 so its never seeing the max speed

Haven't had much time to tinker in the shop lately, (the -45c weather this week didn't help the motivation), I have taken a bunch of steps to try and alleviate the problem,
-firstly was changing the smooth stepper khz settings (post 8)
-lubricated the z axis acme screw as it was running dry (used air tool oil as it has been recommended around the forum)
- cut z acceleration from 30 to 15
- turned off the mach 3 tool path display
-checked all wiring and connections, ohmed out the z motor cable, checked the motor polls ohms and compared to x axis

Ran the 2.5d file last weekend after making the changes, and found 3 things
-the smooth stepper has ran out of data error did not show up - probably due to the tool path display being turned off
- drilled a hole at -.5,-.5 before running the project, sent x and y back after the cut was finished and the bit dropped right in, - which should rule out interference as 3 of the 4 steppers are not missing a beat?
-the z axis at the end of the project was +.05 above the z axis start point (material surface) the prior test was +.07 so really no improvement
I think acceleration can be ruled out, as 15 is below the recommended start point for this set up, and if it needs to be lower then projects are going to start taking way to long (just the accel change from 30 to 15 tacked on close to an hour of machine time.

going forward, today after seeing what louieatinza was saying about bottle-necking, the control computer got updated from 1g ram upto 2g .
- I also checked all the wiring inside the control box, didn't find anything major but did find the wires going from the z axis terminal connectors on the pmdx-134 to the stepper cable panel plug were slightly loose but still snug enough they were not falling out.
-as well ensured there was plenty / redid the heat sink paste between the z drive and heat sink.

I will run the same test tonight or tomorrow but have my doubts that anything that has been done will alleviate the problem.

If anyone has any other theory's that should be looked at it would be greatly appreciated, thanks!

[louieatienza]

Did you try reducing the resolution of the Z axis by lowering the microstep resolution at the drive? Judging by your calculations you should have no problems. Under CPU control I was able to lock at 24KHz or so with a 1.6GHz P4, and used the onboard parallel port.

Aside from this, I think your gas strut may be unnecessary for your Z carriage, or at the very least causing problems. Again, I'd recommend also running your machine with it disconnected. But really, only 48V to run that beast of a stepper, you'll lose torque really quick; it's probably capable of close to 100V. Also don't know the inductance of your stepper but the higher the inductance the more voltage you need for it to perform.

I don't think this is an issue with noise... but you can try running to Radio Shack (or Canadian Tire?) and pick up some ferrite chokes for the axis motors, just as an extra safeguard.

Another thing to check would be the retaining ring on the anti-backlash nut (that the spring pushes against.) I've had one get stuck once and it pinched the nut enough that it really made turning the screw stiff. Some nuts run stiff and the remedy would be a makeshift tap out of a scrap piece of screw.

[grubscrew]

the gecko 201x drive is fixed at 10 microsteps.
I understand what your saying about the voltage, and i may go to a bigger power supply as the drives are rated upto 80v, what gets me is this exact setup is used on lots of cncrouterparts machines, so it has to be a tried and proven combination.
I will try a test without the gas strut, im just not a fan of the sounds the screw/nut make without it. with the power off i can rotate the screw to lift and lower the z axis with 2 fingers, with out the strut i cannot turn the screw by hand to lift the z. I just pulled apart the z axis to check everything, weighed it while the screw was disconnected. the entire assembly weighs 34 pounds sitting on the scale. 4 pounds with the strut installed

the screw turns fairly easily through the nut, pulled back the retaining ring to disengage the pressure and felt no difference in turning the screw, so can call the nut good.

I will be ordering a 1/4x1/2 motor shaft coupler tonight for i can try one of the 400ozin steppers from my other machine on this z axis.

[grubscrew]

Got the test ran today. ran the program a bit slower, without the gas strut, and after dismantling the z axis yesterday and inspecting/ cleaning everything. and got the exact same result, was +.05 at the end of the run x and y were dead on. at least its consistent.

was looking further into the stepper vs voltage as well today. From what I understand to find the motor rated voltage : Amps x phase resistance = voltage so:
7*.31= 2.17
Gecko drives recommend running anywhere from 4-25x rated voltage. 2.17*25=54.25v
not much more room to add any voltage.
stepper spec sheet here: http://www.cncrouterparts.com/specs/...ta%20sheet.png

Being a consistent gain would leave one to rule out allot of the looked at possibilities ( interference, lost steps, strut, speed and accel, electrical, control system)

Im wondering if there is a thermal factor causing the troubles. the error is not instant on the test pieces, it only becomes noticeable around the half way mark where some of the rough pass lines start getting left behind as the bit is cutting higher and higher.
My shop is kept above freezing when not in use usually around 10deg c. but before running any projects, the room temp is brought upto 20c, the machine and spindle have a 20 minute warm up program, as well I have a lcd thermometer on the spindle to monitor coolant temp. usually on start up coolant is sitting around 8deg c, before turning the spindle on I run a hairdryer blowing into the cooling rad with the coolant circulating, which gets the coolant temp up to 20 within minutes, then the 20 minute warm-up program,. at the end of the cut today the coolant was around 24deg c. which should rule the spindle out, but maybe the spindle core is taking a long time to heat up?
does delrin (the anti backlash nut) expand and contract?
regardless im about to throw a hammer at it :boxing:

anydirection would be greatly appreciated, thanks!

its amazing what frustration such a small error can cause.

[louieatienza]

The solution may be to get a higher performance stepper. I don't think at your speeds that any thermal issues would cause binding. I had a small router whose Z axis probably weighed 15-17lbs. I tested it at 550Ipm... with a 4tpi (1/2"-8, 2 start) screw and 180in-oz motor wired parallel at 48v, 40 accel, and no lost steps. However this was a Kollmorgen stepper (Danaher) rated at 3000rpm max at 75v. I probably wouldn't run my leadscrews constantly at 2200rpm but for short rapid bursts it's probably fine.

I can't find a torque curve chart for your stepper, but I'm aurw you don't have a lot of torque at 200rpm let alone 600... not at 48v at least.

Add that with the fact thatthe larger stepper has to overcome its own detent torque as well which can be as high as 20% of the rated torque... Also, this test may not be the most fair foe rhe machine. It's normally a bend of x and z or y and z moves that equal your 100ipm. Just plunging up and down at 100ipm with no dwell is tough on most machines, and I doubt you'd be plunge roughing. Maybe. I'd track down an older parallel port computer and retest just to be sure.

[grubscrew]

I should of put that in the last post, i know plunge is only the straight down at the begining, and that the rest of the raster is done at the set feedrate or the max velocity set in motor tuning. -for the last test i has mach 3 z axis velocity set to 100 accel 15.

With everything else ruled out i guess il start looking at stepper options, I was gopong to get away from buying anything for a while as the canadian dollar is the shats right now. But if it aliviates the frustration it will be worth it

[ahren]

I just got an email from grubscrew on this topic -- first I have seen of this. I will say that I'd be surprised if this is related to the motors. Our Nema 34 motors are low inductance (2.1 mH), so run very well at 48V and don't require higher voltage to maintain plenty of torque at higher RPM. I also very much doubt the Smoothstepper is the issue. We sell these standard in all of our pre-wired kits, and see significantly fewer issues than we used to see with parallel port based systems.

The fact that your error is a consistent 0.050" makes me wonder if the issue is related to your machine hitting a clamp or other small mechanical interference on a down stroke at some point -- I'd look carefully to make sure there isn't some point where the path is even slightly impeded on the down stroke by a clamp or other component on your machine. On milling machines, I've seen a gradual loss of steps from people using a non center-cutting bit that just didn't cut well during plunging. While I doubt this is specifically your issue, something like this that is repeatable sounds more likely than anything related to the electronic components you are using. The gas strut, dust collection hoses, cable track -- I'd look for anything that could be getting put in an odd position at some point during this program.

Mechanically, I'd check to make sure your anti-backlash nut isn't unthreading from the aluminum body it is in. This is kept in check by both the jam nut on the back of the nut, and by the cross-drilled set screw on the top of the nut body. If both of these are loose, it's possible for the nut to un-thread slightly. Likewise, I'd check to make sure whatever coupler you are using is tight on both the motor shaft and the ACME screw.

Best regards,

Ahren
CNCRouterParts

[Devastator]

Check bit tightness, check for bearing endplay on router, make sure stepper wires are not bundled up against each other in your wire tracks. All of those things have bit me at one time or another.

[grubscrew]

small update,
after 2 weeks I finally found some time to do some more troubleshooting
Just finished running the exact same g code, in the air spindle and vfd powered down, same result, z axis about .068" this time above the beginning z location. that rules out any interference from the spindle, issues caused by material and bit interaction, and dust collector interference as it was not running as well.

Running the same test right now except with swapped port and pin locations in mach and cables between the z and x gecko 201x. this should rule out any driver problems.
If the driver checks out good, I will fish the z axis cable out of the chain and run the test with the z motor cable isolated from everything to see if that makes a difference...

[louieatienza]

small update,
after 2 weeks I finally found some time to do some more troubleshooting
Just finished running the exact same g code, in the air spindle and vfd powered down, same result, z axis about .068" this time above the beginning z location. that rules out any interference from the spindle, issues caused by material and bit interaction, and dust collector interference as it was not running as well.

Running the same test right now except with swapped port and pin locations in mach and cables between the z and x gecko 201x. this should rule out any driver problems.
If the driver checks out good, I will fish the z axis cable out of the chain and run the test with the z motor cable isolated from everything to see if that makes a difference...

I did have an issue a while back where I actually zapped something in my Y axis drive causing loss of position in one direction. The culprit was a loose Molex connection that caused some arcing between leads. It ended up hosing my drive (a Xylotex.) So if you swap drives, swap cables too to check.

[wizard]

Are you still getting buffer overflows to the Smooth Stepper? This sort of fault bothers me though I have a very hard time seeing how it could be causing your issue. If so I'd look at how communications is set up between Smooth Stepper and the PC. I don't have a Smooth Stepper but this sounds like a fault you should never have. It might not be the cause of this problem but in the end it can't be tolerated.

It is interesting that you have zero problems skewing the Z axis all by itself but the issue returns when you run all three axis together. This has me wondering if you might be getting some cross talk in the system wiring.

The other problem or possibility is the G-Code itself. Back in the day when I serviced a lot of specialized lathes we would run hand coded G-Code to test out the machine for axis repeatability. Hand coding your test routines allows you to understand exactly what is happening on the machine and will eliminate a whole bunch of unknowns. In one instance we had machines loosing fractions of a thousands every cycle of the machine, something seen only after running a hand coded loop as the controller only displayed positions to 1/1000 of an inch. This was traced down to the switching of a relay in a packaged spindle drive. So why am I trying to suggest? Start out real simple with a hand coded loop that runs maybe 100 times, add functionality to it until you see the problem crop up.

For example take your Z axis test program and add Y axis movements to it and test, then add X axis movements and test do this until you see a problem. Don't forget spindle operation and other devices like coolant solenoids. What you have to be aware of is that it is often the switching of a device on or off that causes a problem. The more that device is switched on or off the more likely you are to see a problem. Depending upon the resolution of your system you may need many cycles to see a measurable and repeatable fault.

small update,
after 2 weeks I finally found some time to do some more troubleshooting
Just finished running the exact same g code, in the air spindle and vfd powered down, same result, z axis about .068" this time above the beginning z location. that rules out any interference from the spindle, issues caused by material and bit interaction, and dust collector interference as it was not running as well.

Running the same test right now except with swapped port and pin locations in mach and cables between the z and x gecko 201x. this should rule out any driver problems.
If the driver checks out good, I will fish the z axis cable out of the chain and run the test with the z motor cable isolated from everything to see if that makes a difference...