Circles not round! (SOLVED)

[fnaranjob]

Hello everyone,

I ran into a pretty weird problem today, we just finished building 2 cutters with the 2012 DSP, both with the same exact hardware, one of them runs perfectly fine while the other shows the issue you can see in the attached pictures:

Cut at speed 50mm/s

Cut at speed 100mm/s

, the circles don't come out round!, I tested cutting an array of circles of 10mm diameter, perfectly round in the drawing and the result is awful, I tried everything I could think of to solve the issue, I checked for bad grounding, bad connections, switched to a previous firmware, changed encoder resolution, retuned the motors, but no luck, I noticed one thing though, the issue gets worse with higher cutting speed, the cuts in the pictures were made at 100mm/s and 50mm/s, the letters were cut at 100mm/s, if you look close the letters only show defects on the curves, the straight lines cut well, with accurate dimensions. Does anyone know what could be the problem?, as every other build we make this is for a client and downtime is a problem.

Edit, 2013/23/05:

I solved the problem!, it was the Y axis servomotor tuning!, it seems that the autotuning function of the Yaskawa servopack is not that good, I retuned the motor manually and the problem is now completely gone, I can vector cut a 5mm diameter circle at 100mm/sec and it comes out perfect!, thank you everyone for the suggestions though.

Cheers,
Fernando

[Tech_Marco]

On parameters setup, change the Falling edge to Rising edge or the other way around. Save the change and press RESET button on the LCD to take the effect. That should fix the problem

Marco

[fnaranjob]

I'll try that and get back to you with the results

Thank you

[fnaranjob]

Marco,

I tried your suggestion but it didn't work, I unplugged everything from the laser controller and checked the X and Y outputs with an oscilloscope, the signals are coming wrong from the controller!, if I send the controller to cut a triangle I get an uniform waveform in the lines that use only the X axis but when both X and Y move the waveform gets very distorted, this only happens to the X axis pulse signal, the Y axis pulse shows a perfect waveform, I think I may have an old lasercad or firmware version, I have version 6.27, I have 2 DSPs and I tested both, both of them show the problem I described and It has never happened before, this happened only with the last controllers I bought from you, any suggestions on what to do?, should I try upgrading lasercad and the firmware?, my firmware versions are:

Main board: 7.12.8.17
LCD:5.12.2.29

on the other DSP i have:

Main board: 7.12.12.26
LDC: 5.12.2.29

Please help me solve this, my client is getting impatient.
Thank you

[Tech_Marco]

I Jusr sent you the latest OS and LaserCad. Youmay update the OS see if it help. You may want to check the motor driver and its setting. It didn't look like it isthecard issue


Maeco

[fnaranjob]

Marco,

I updated firmware and lasercad, no luck, I even tried replacing the DSP, the power supply, the motor driver, the problem remains, for the record I'm using Yaskawa Sigma 5 servo motors and drives, I read somewhere that someone else is using the same motors, any similar problems?

[baccus61]

On first sight it looks like you have a loose pulley or drive gear but it doesn't look consistent enough for it to be that as the bulge isn't in the same place.

One problem I had was tensioning the belt way too tight and breaking the tensioning strands inside the belt which allowed the belt to stretch. The belt was 8mm wide and 2mm pitch (I think). This was totally out of sight and very hard to detect until I noticed the teeth on the belt were getting flattened going over the pulley. Took me a while to find that one.

I also had some weird problems similar to this and after a few years of putting up with it and changing EVERYTHING on the machine, the last thing I changed was the power supply which fixed it. I now use a switch-mode power supply which turned out to be 1/4 of the price of the previous one. The old one was a toroidal transformer with big caps on it and a rectifier as listed on the Gecko site.

Make sure your earth is connected well also as most of the problems with the board seem to be due to earthing issues.

Can you swap motors around to see if the problem changes?

That's about it for me. I'll think about it some more.

[fnaranjob]

On first sight it looks like you have a loose pulley or drive gear but it doesn't look consistent enough for it to be that as the bulge isn't in the same place.

One problem I had was tensioning the belt way too tight and breaking the tensioning strands inside the belt which allowed the belt to stretch. The belt was 8mm wide and 2mm pitch (I think). This was totally out of sight and very hard to detect until I noticed the teeth on the belt were getting flattened going over the pulley. Took me a while to find that one.

I also had some weird problems similar to this and after a few years of putting up with it and changing EVERYTHING on the machine, the last thing I changed was the power supply which fixed it. I now use a switch-mode power supply which turned out to be 1/4 of the price of the previous one. The old one was a toroidal transformer with big caps on it and a rectifier as listed on the Gecko site.

Make sure your earth is connected well also as most of the problems with the board seem to be due to earthing issues.

Can you swap motors around to see if the problem changes?

That's about it for me. I'll think about it some more.

Thanks for the suggestions,

I swapped motor signals, when I do that the problem changes, it seems to "rotate" so I think it's not a mechanics issue, but it's hard to know, maybe I'll think about a toolpath that helps me determine Wether it stays on the X axis or moves to the Y axis when I swap signals.

One more thing I noticed: This problem happens only when the two motors have to move simultaneously, if I make a square or rectangle it comes out perfect, I think that means its not caused by a loose pulley, I made an array of squares to look for backlash and the sizes and spacing comes out perfect, my machine uses high end belt driven linear actuators, these have nearly zero backlash and are very rigid, if you see the letters I cut the problem only shows on curves or diagonal lines, not on horizontal or vertical lines, I think it has something to do with the the motors not being able to maintain the correct speed and synchronization between each other, I think that since the issue gets a lot worse as I increase the speed, and it is more noticeable with small objects, if I keep the speed low and the objects large the problem is minimal but sometimes my client wants to run a vector cut at high speed, low power just to mark the material and that's where the problem gets worst.

I noticed one last thing, if I engrave something and then cut it the engraving is moved to the right by a couple millimetres only in the X direction, and this offset is constant no matter the size of the artwork, the engraving comes out with correct size and shape but it is not located correctly. Hope that makes sense, English is not my first language.

[Tech_Marco]

What "um" value do you have now? Try to use 2.5 or above but smaller than 20
The DSP seems more picky on Servo motor than a stepping motor.

Marco

[Raman Narayan]

The problem is definitely due to XY steppers loosing steps.
What I believe that is happening is, while you do the squares,
only one stepper is active during the horizontal/vertical move
and the signals are clean for the controller so, everything works
as expected. This is also a sign that your mechanical system is
correct(pulley gear, stepper divider ratios etc).
Try the same with a diamond(rotate square 45 degrees), you will
see a consistent but similar issue, as seen with circles.
Also you mentioned the X axis pulses are distorted, it does seem
to be due to crosstalk with the Y axis pulses getting coupled to
the X-axis(or due to a loose connection in the X-axis ground, or
a ground loop voltage that is modulated by the Y-axis pulses).
So, if you re-check the X-axis stepper connections, and possibly
use twisted cables to provide shielding, you should be able to
correct the issue.

During cutting circles, both X-Y axis steppers are active, and since
the X-axis pulses are being distorted, the controller is missing
transition of few pulses(determined by the rate of Y-axis pulses),
leading to reduced movement of the X-axis, which produces a
oval instead of a circle.

You could also add some low ESR tantalum capacitors(100-1000uf
with voltage rated atleast twice the supply voltage to the driver) right at
the power supply terminals of the X-Y stepper drivers, if the issue is
related to noisy power.

Regards
Raman

[fnaranjob]

What "um" value do you have now? Try to use 2.5 or above but smaller than 20
The DSP seems more picky on Servo motor than a stepping motor.

Marco

I'm using 5um, it was 2.5 previously but I increased it thinking the DSP couldn't process fast enough at high speeds but the problem doesn't change.

The problem is definitely due to XY steppers loosing steps.
What I believe that is happening is, while you do the squares,
only one stepper is active during the horizontal/vertical move
and the signals are clean for the controller so, everything works
as expected. This is also a sign that your mechanical system is
correct(pulley gear, stepper divider ratios etc).
Try the same with a diamond(rotate square 45 degrees), you will
see a consistent but similar issue, as seen with circles.
Also you mentioned the X axis pulses are distorted, it does seem
to be due to crosstalk with the Y axis pulses getting coupled to
the X-axis(or due to a loose connection in the X-axis ground, or
a ground loop voltage that is modulated by the Y-axis pulses).
So, if you re-check the X-axis stepper connections, and possibly
use twisted cables to provide shielding, you should be able to
correct the issue.

During cutting circles, both X-Y axis steppers are active, and since
the X-axis pulses are being distorted, the controller is missing
transition of few pulses(determined by the rate of Y-axis pulses),
leading to reduced movement of the X-axis, which produces a
oval instead of a circle.

You could also add some low ESR tantalum capacitors(100-1000uf
with voltage rated atleast twice the supply voltage to the driver) right at
the power supply terminals of the X-Y stepper drivers, if the issue is
related to noisy power.

Regards
Raman

You're right, I tried a triangle instead of a circle and only the diagonal lines (where both motors move) show the problem, the result is something like this:

result of cutting triangles

Notice the problem is reversed when I change the order in which the lines are cut (clockwise vs counterclockwise) from the results I think its one of two things: a) the X axis is accelerating too fast compared to the Y axis or b) the Y axis is accelerating too slow, lagging causing the line to be distorted, the amount of pulses seems correct as the height of the triangle is of correct size, this means there are no lost steps, I think it is the acceleration/speed that's not in sync or something like that, it actually could be crosstalk as you suggest, I'll rewire the pulse/dir lines with shielded cable just in case, now that I think about it I used the drives software autotuning feature to tune the motors, maybe it didn't do such a great job, I'll plot some graphs to check if the motors are following the command pulses correctly, any other ideas are welcome, this problem is really confusing me!

[Greolt]

I assume the triangles are approx 10mm per side.

Cut them at something like 50mm per side. Same speed etc.

See if the resulting shape is the same or not.

My guess is that the Y axis is lagging but eventually catches up.

Elasticity in the belt drive or similar concept in the servo drive.

When a servo gets behind the drive feeds it more amps until it catches up. Unless it gets so far behind that the drive faults.

Too many things are blamed on "loosing steps". If steps are lost, you generally don't get them back by the end of the move.

[ltran2000]

I have the same problem as yours, I changed stepper, driver, DSP, power supply but still problem,
later I find out that my X axis is twist ( not square to Y axis) readjust it and that solved my problem,
please check yours.
ltran2000

[artofmystate]

This post is marked as solved but there is no posted solution.

How did you fix this problem?

[artofmystate]

Having this same problem, once I changed out the wires from the DSP to the stepper control board to twisted pair (CAT5), shortened the wires as much as possible, and added space between those wires and the limit switch wires most, if not all, of the position error has gone away over long cuts.