Thanks for the tip FRIDGE. Yeah, rack and pinion would definitely be the way to go in the long run. Can't afford it so it looks like I'm stuck using belt with The RAT for the moment.
Good luck with your build.

 

Back after a little hiatus. Backlash is getting me down. Need to try and find it. There are so many factors to consider and I feel like I'm getting bogged down with it.
I feel like the belts are where most of it it originating. I recently went to GT3 from GT2, seems to have helped a bit.
Maybe it's to do with the actual setup in Mach3, maybe I'm trying to go to fast.....?
Here are some of the settings currently running..
Accel: 100mm/s/s
Velocity: 8000mm/minute
Steps/mm: 33.33333333 (x and y)

Feed: 970mm/minute
Speed: 2-2.5 on a makita 700 series trim router ~14000RPM
DOC: 2-3.5mm
Single flute 1/4" up spiral cutting plywood.

Here is a picture if a little test pattern that I just did.
Might help pinpoint some of the problems.
Red dot is the start point. Numbers represent start point of each cut. All at 2mm depth, except for 10, being the outside/last cut, done at 3.5mm per pass.
Seems to be a lot (maybe 1/2 a millimetre or so either way) of movement occurring on entry and exit of material.
Any ideas or questions are welcome.

 

Still trying to figure this out. Now the question is, backlash or missing steps?

 

Checking grub screws now.

 

They look fine. Here is a little one minute video of the machine doing air cuts. I sped the whole thing up to around 2910mm/s and accel up to 150mm/s/s from 100mm/s/s and 970mm/s, to see if I could get it to obviously miss steps. After marking the shafts on the steppers while at zero, running the gcode through and then returning to zero, all motors seem to be in exactly the same spot. So to my as yet untrained eye, it isn't missing steps.

 

Hmmm, video didn't load. Maybe this will work.

 

Just noticed something. You can kind if see it in the video. It's almost like the z stops at zero on its way down from the 20mm safe z. In this case it's a 1mm DOC. Is this normal with mach3?

 

Might explain the bad entry's and exits, that is if either the x or the y start moving before the z has reached the end of its movement......

 

Hi Larry, not sure if I quite follow what you are saying. Are suggesting I go over the motor tuning again? If so what recommendations do you have?

 

Nema 23s (269 Oz-in) all round. Gecko G540@48v.

 

Ok, so I did some testing. Changing acceleration, rapid and feed rates for every lot of six squares.
The numbers at the top right of each lot of six squares represents the acceleration, rapid and feed rates from top to bottom as you see them. The bottom right one is supposed to be 10000 not 1000..
Looking at the bottom right square (the one with the circle in it) is telling me that there is backlash in the y axis. Same with the bottom right corner of that particular square....
So, next thing to do will be to investigate if there is any way to reduce the backlash in the y axis.

All in all I'm quite impressed that there is little or no difference as I wind up all the rates..
Left to right is the y axis, top to bottom is the x axis.

 

Thanks Nick. It's driving me insane.
Well, I decided to rotate it all trough 90 degrees to see what the difference was.
Not sure exactly what it's telling me yet.
Anyone?

left to right is the x axis, top to bottom is the y axis.

 

Still think the y axis is the one with the most backlash.......
Tried tightening the belt, checking for binding. No dice....

 

Just so you know the gantry is the y axis in this case. Currently I have a 1200x600 bit of MDF that I've levelled as my working area for the moment, which is roughly centre.
So by doing what you are saying, are you trying to remove the possibility of flex in the gantry?
I guess I'll move it to one side over and re-level it?

 

You have a very good point Tweaky. I noticed this after I did the 90 rotation.
Just getting the gcode out of the computer now. Will post it in a minute.

 

here it is.
Cleaned it up a little when I rotated it. Took out one of the circles and one of the V's.
Hope you find the culprit in there.

 

Thanks Tweaky! I think we are getting somewhere now..
Guess I'll be pulling it all apart to try and find that **** flex, of which there is a little in the z in the x and y direction. need to change the z limit switch anyway.
Learning, learning, learning! Silver linings and all.
Thanks agin for time and patience.

 

I'll give it a go. Report back in a couple of hours.