I am using a Smoothie 4x board running my three axis CNC mill with a 52 inch by 52 inch cut envelope. I have one X axis stepper, two Y axis steppers and one Z axis stepper. The X and Y NEMA 23 (3.5 amp) steppers are being driven externally with 2M524 step drivers and the Z axis is being run with the Smoothie step driver.
I have worked very hard to get rid of mechanical backlash with decent results. Now I am fighting something that looks like a software controller issue. On some of the paths the router appears to start or end slightly off path by maybe half the width of the router bit. The GCode is generated by CamBam. None of these overcuts appear in the CAD artwork or when simulated by Camotics.
One other data point to mention is size. These jobs are around 24 inches by 24 inches. I don’t know if Smoothie has any issues with large-sized 2.5D GCode files.
My Smoothie is running the latest firmware as of late November. I updated the config file as well.
I have a Google Drive link that shows some pictures that highlight the issue and the GCode file of the job. For some reason I can’t put the URL in this post. If someone can suggest a way to share these pictures and the GCode file I would appreciate it.
Ken
Imported from wikidot
Hey.
You can post your link here, just change a few of the characters so it’s not recognized as such by the robot.
Smoothie doesn’t have any problems with large work areas, it works on several meters-long machines.
When you say there is an error, is that error cumulative, or is it just localized ?
Cheers.
That’s extremely weird I have never seen anything like it.
What are you using to send the Gcode to the Smoothieboard ? USB or ethernet ?
Can you try Pronterface over a short USB cable ?
I copy GCode files to the micro SD card on Smoothie, unmount the card in Linux, use Pronterface to jog the machine to zero, issue “G92 X0 Y0 Z0” and then a "@play /sd/file.ngc.
I use a 1.5 meter USB cable but do not send any GCode data over the cable.
I use ST-M5045 drivers for X axis and dual Y axis. Standard Chinese stuff. The wiring from the SmoothieBoard to these drivers is maybe 1 meter. The wiring from drivers to steppers (A+/A-/B+/B-) is about 2.5 meters long for Y axis and 3.5 meters for X axis and Z axis.
Have you ever seen issues with these types of external drivers?
I’ve never seen any issue like this, the effect is way too big to be noise on the stepper drivers, and if it were it wouldn’t “recover” after the error and the offset would accumulate.
Are you absolutely sure the Gcode is not to blame ? Cambam does have an option to do this “go further in corners” thing … Can you try another Gcode simulator ?
I can check for corner over cut. Can you suggest another non-commercial (or free to evaluate) GCode sipmulator besides Camotics?
Don’t know if this could be a wire issue. On the steppers I’m using cheap “security wire”. Four 22 AWG copper wires in an unshielded jacket. Like old telephone wire.
Thank you. I’ll try it tonight.
Don’t know if this could be a wire issue. On the steppers I’m using cheap “security wire”. Four 22 AWG copper wires in an unshielded jacket. Like old telephone wire.
Any problems with wires would cause lost steps, which are cummulative, your problem isn’t.
Hey Alan,
I studied the tool paths in Camotics and Cutviewer Mill. Both apps did not show the cut issues. Both apps show that the bad spots are where the tool descends into the workpiece. Maybe the location is off by some amount and then corrects itself once on the cut path. The amount it is off looks to be half the diameter of the router bit. This is a 1/8" downcut spiral end mill.
I have posted on the CamBam forum with the screenshots, GCode and the CamBam file. There’s some pretty sharp machinist there so we’ll see what they suggest.
Thanks,
Ken
Hi there,
You can download ESTLCAM from the official website, and try to generate the same code with this program(you can even simulate it)
is cheap and is free trial for a long number of times, don’t say that is the solution, but this program is very good for the 2.5D, and is generating a very clean G-code.
My two cents…
Regards