A Very Confusing Coontroller ? Problem

[MoonlightMedia]

I hope this is posted in the right place - If not, please advise.

I am running Mach 3 with a fast Win2K System
I have tried to resolve this on three different computer systems figuring
maybe they were not fast enough.

The current condition is as follows -

1- Straight line cuts work fine on both the X and Y Axis.
2- Diagonal line cuts work fine.
3- The Z Axis seems to respond appropriately

My problem is that I cannot cut a circle !

There is a disturbance in certain parts of the circle where it looks as if the
X or Y Axis was actually going the wrong way.

To troubleshoot this, I have done and observed the following.

All of the following was done with a simple 6" circle plot.

1 - I thought it might be power related so I beefed up the power supply
No positive results.

2 - I thought it might be cable noise so I reduced down to a certified 3 ft
cable.
No positive results.

3 - Ran Mach 3 at varying speeds and settings
The problem persisted.

4 - Disconnected the X Axis and ran the Y Axis only.
Observed that the back and forth motion was somehow disturbed at the
end of its' range. Basically when it was stepping at its' slowest point on
the Y Axis.

5 - Repeated item 4 while the Y Axis was disconnected and observed the X
Axis only.The disturbance was repeated at the extreme ends of the X
range.

6 - Removed the motor covers and observed that the motors chatter in their
direction during the disturbance. You can also feel the reversal with your
hand on the router.

7 - Selected a point of reference (home) a full 20" in X and Y from the
beginning point and repeated steps 4 thru 6.
No positive results.

8 - I thought that perhaps the diection line was seeing false signals so I
disconnected the X Axis direction line and grounded it. I repeated step
5.
No positive results.
The process was the same but as might be expected the plot only ever
went left to right other than the disturbance. The change in direction
during the disturbance STILL OCCURRED even though the direction line
was grounded.

9 - I reran all of step 8 but brought the direction line to +5V. The plot ran
the other way but the results were the same. At the ends of the plot
where the movement is minimized, the steppers chatter and reverse
directions for a brief period, maybe a 1/4 second.


As you might imagine this has me most perplexed. My customer is growing
very restless and wants to know when I will be done. At his point, I can
only suspect that the boards are not operating right but for the life of me
I can't see how or why !


Any help you can give or light you can shed would be much appreciated !


Bill Cox
Moonlight Media LLC
Higganum, CT

[Torchhead]

What kind of interface (breakout) card are you using? What motor drives?

[randyf1965]

Can you post a picture of the plot? maybe a scan of it?

[ger21]

Sounds like backlash, and lots of it.

[MoonlightMedia]

for Torchead

The breakout board is a simple one with no active components. If you read the text
for step 8 and 9 , you will see that nothing is moving except the step line on a three
foot parallel cable.
I am reluctant to state the drive vendor because I don't want to give them a black
eye unjustifiably. I'm awaiting some tech support and if it isn't forthcoming, then I
will state the vendor and the level of support I got from them. In reality, does it
matter? I can tell you that the motors are being driven at 4 amps 24VDC unreg. This
is well within the motor spec.

for randyf1965

I have posted a picture of the circle plot although the failure mechanism isn't
very evident until you follow the troubleshooting steps I took. You can watch this go bad in a linear straight line on a single axis.

for ger21

This is my first encounter and I am not aware of the results of backlash. Could it be
so consistent as to repetatively produce the circle plot I supplied. This is the same
when you do a circle inside a circle inside a circle etc. It always happens at the same
angle which equates to the same speed in a linear single axis line plot.

Also note that this does not happen when simply plotting a straight line but rather
when the speed of the plot is changing as in a circle.

[NC Cams]

Cutting a circle is a VERY difficult animal. Why?

Because in one axis of movement, you have the drive trying to send the motor at its maximum speed potential. Concurrently however, you are asking for relatively no motion on the other axis. This axis is now trying to bounce back and forth between two relative speeds (very slow and ultra slow and ultimately stop and immediately change direction) but movement is essential for you to make the circle.

Chances are, all your problems are at 12, 3, 6 and 9 o'clock. These are the hardest points for the CNC to move thru. You're next problem are at the 45 deg points as both axis are moving at essentially equal speed - if not, footballs, not circles.

Take the 3 oclock direction change for an example (the others are the same only rotated in a different positon).

As you go thru the 3 oclock point, the Y axis is moving at the max speed that it will see during this portion of the range of motion. Yet the X axis is moving less and less in the positive direction and it preparing to stop and move with the same velocity path in the opposite direction - hopefully quickly, smoothly and in proper time with respect to Y axis motion.

Two things are giving you fits:

1. The X drive is probably being asked to work at a fraction of its minimum step potential. IT probably wants 1/16th step to navigate properly but that ain't gonna happen due to the step function you're working with. It could be bouncing back and forth due to rounding errors in the processor.

2. IF you have any lack of rigidity, slop or increased friction or backlash in the drive axis (especially in lead screw more so than a ball screw) the motion you instantaneously need can't/won't happen. Result: the table dwells, shudders, jerks, doesn't move then WHAM takes off because you shoved and it finally moved.

We struggled for months with the servos on our Eztrak to make it cut a decent circle. Never happened.

That's when wee found that our system had a third problem - the axis didn't return to 0,0 after making the bazillion cut steps we had to go thru to make the perfect arc of travel. FInally after a huge expenditure of time and money, we got the mechanicals iron out.

We still had the return to 0,0 issue to resolve. A visit by the Bridgeport service guy fixed it (bad read ahead issue with cache, improperly formatted H/D, bad power supply settings and servos poorly tuned and also trying to move TOO fast). The faster we moved the worse the problems became. Slow gradual moves worked better.

We've seen guys try to grind cam profiles with steppers as opposed to servos. They NEVER can get smooth finishes. WHY? no matter what they do, the motor is stepping. The steps generate chatter that stems from the cogging inherent to stepper operation. SImple, use more gear reduction. Sadly, this solves one problem but generates another - you lose the speed needed to move fast enought to generate the profile during the max velocity of motion points of axis travel.

Easiest way to understand what's happening re: the stepping is take a magnifying glass and look at the computer screed and try to make the mouse pointer move in a PERFECTLY DIAGONAL line.

NO can do. It is a step function and you're trying to generate something that requires smaller steps of motion than you can EVER possibly generate. With 2:1 gearing and 2000ppr counters, gearing, an analog system will ultimately do better because each rev of the screw is 1 step in 4000 but even so, it will have minute problems at the direction changes.

We eventually got our direction change blips/flat spotting/chatter down to 0.0001'~-.0003". And it only took $1650 in ball screw bearings, 1.5 days of service work by Bridgeport tech (plus expenses), in concert with WEEKS of work on our part eliminating hysterisis and friction in the system. Forgot to mention #1300 in servo motor service.

Simple task, cut a circle.

Sorry no. The generation of a sinusoidal velocity profiles (which is what you're doing when you cut a circle and in two directions simultaneously) is not anywheres near the easiest thing for a CNC to do....

GER21: yes backlash is a factor but when we got it down to amost nothing with preloaded ball screw and hugely preloaded ball screw bearings, we still had it. That's when the problem got serious because we knew we'd be into a realm of the machine where the skill of the person who wrote the code would become strongly tested. Fortunately, our code worked pretty good. The mechanical and electrical gremlins are what bit us.

Hope this helps...

[MoonlightMedia]

for NC Cams - Thanks for the response. I appreciate that you had lots of problems and
time will tell if they relate to what's going on in my situation. The table (mechanics)
is an older Gerber but it's pretty tight. If you view the plot I sent, you will see that
the disturbances happen in different places than you mentioned. It's fairly plain that
I'm not speaking about a little resolution error or differences between the computer
plot and the table but a rather large divergence from a reasonable circle.
I'm not sure I understand your first paragraph at all. Are you saying that the
X and Y plot directions and speeds are affecting one another ???
After all is said and done, do you have any tests, suggestions or observations that
might lead me toward fixing my problem ? I appreciate any constructive input you
can offer. Again, thanks for the response!

[lucas]

To me this looks very much like (mid band?) stepper resonance, you have the problem in circles but also in straight lines and every time only at the same low speed.
I assume the motor's are noisy when the problem occurs?
Are you using microstepping? if not this could solve it.
Try the same straight line without cutting and spindle off as this one could trigger the resonance.
Do you have another type, brand of motor? If yes just connect it (no need to mount it on the machine) and listen to it's noise when running the same straight line.

[NC Cams]

Looking at your diagram, the X axis is going to have to move at a much lower speed with respect to Y from B-to-C than from A-to-B or C-to-D. It repeats on the other side of the circle in true mirror image fashion. Predictable and repeatable which is an important clue.

The fact that the 12, 3, 6 and 9 o'clock motions are pretty decent, you may not have a problem with hysterisis in the mechanicals at these critical turnaround points. That's good. You may simply be seeing the transition points from stepping to micro-stepping which is why the thing is glitching SO repeatably and predictably.

You may have a coding error which is showing up at these points. Milling a circle is NOT easy due to the fact that you're essentially asking the machine to go thru all four quadrants with both axis under exacting velocity control with respect to each other so as to make the proper/necessary motions. Simultaneous motion is NOT so easy to do, especially if you have a "look ahead" problem with your controller/software.

If there is a way to slow the process down - microstep the whole time, go to a slow inch/minute feed, these might be some things to try.

I don't think it is noise - it it is, it is WAY too predictabe and WAY too repeatable and at EXACTLY the same position. I'd contend it is either a software glitch or a oddity that you found between the program and your axis drivers.

I'd suspect resonance to cause highly erratic motion - this is anything but erratic. Perfect symmetry and repeatable form. Looks like the control is TELLING it to do what its doing.

Yes, when you cut a circle the X and Y axis have to be in PERFECT sync with each other. If not, you get footballs, ovals, round cornered squares, anything but a circle.

Try doing point to point milling instead of arc miling, like 3600 flats instead of 2 or 4 arcs. If you do point to point and the tangents all fall on the circle, you net out a circle and avoid the software glitch you may have found. That's how we got around our problem in milling our cam lobes (circles with bumps on one segment of them).

[ger21]

Can you post the g-code for that?

[lucas]

Originally Posted by NC Cams I'd suspect resonance to cause highly erratic motion - this is anything but erratic. Perfect symmetry and repeatable form. Looks like the control is TELLING it to do what its doing.

I've been struggling with stepper resonance lately and did a lot of tests, these have shown that this resonance is very repeatable and always occurs at exactly the same speed.

Originally Posted by NC Cams Try doing point to point milling instead of arc miling, like 3600 flats instead of 2 or 4 arcs.

He also experiences the same vibration at the same speed on straight lines...

Another possible test and avoid the circle problems would be to move at a small angle like:

G00 x100 y001
G00 x0 y0
G00 x100 y002
G00 x0 y0
G00 x100 y003
G00 x0 y0
G00 x100 y004
G00 x0 y0
etc..

These will move the Y very slowly and could produce the resonance.

I could be wrong because I'm very focussed on this resonance issue wich has recently shown to be a problem on my PCB drill, it occurs only when doing these "very small angle moves".

[NC Cams]

Can you factor OUT the area of motion just before and just after the glitch?

Then do the motion on a step by step, pulse by pulse basis. If its resonance, sound should give it away. If it is a true signal issue, the glitch should repeat at slow or high speed.

Fortunately, it occurs at a reasonably repeatable point. Go there and see what's happening that shouldn't be or vice versa. Easier than looking for a random loss of motion repeatablity which is what we were chasing.