Software issue or mechanical

Have a 4x8 plasma table im building. Driving with 3/8" wide GT belts at .2" pitch. Heres my problem.

I run the gantry up against my dial indicator and if i program to move 20", 50" or whatever distance away from the indicator and then come back the same amount the machine will come back and stop within .001" every time. If i move it up against the indicator and then program to move in increments of .100" moving the same direction it is not consistent. First move might be .090", next move will be maybe .085", next move will be like .105" and so on. If i reverse direction there is a lot of backlash to deal with, but moving in same direction i would expect the machine to be accurate within at least a couple thousandths. One motor step = .001" at 10 micro steps which im running.

Are there any settings in Mach3 that might need some tweaking?

It sounds like mechanical slop. It could be a pulley that's a little loose - try making a mark where the pulley and the shaft come together, and see if it shifts when you make those little moves.

ONE move out 50" and back and its good..... short moves out and its not coming back properly ? Along with mechanical issues like a loose pulley....

I'd try a test of 100 identical back and forth moves, record the loss on the dial indicator, then run those same moves only 50 times. If you have precisely HALF of the previous record loss of 100 moves, you could be dropping steps on the direction change.... meaning your pulse/dir settings of the software do not work well with what your driver is expecting (direction change durations along with high/low variation).

If your just stepping OUT .100 and it does not get there, but will get out 50" every time ?? is this on ALL axis or just one or two ? If you have one that works correctly, it might also be necessary to swap motors/drivers around to see if you have some bad component(s).

Just noticed you said you do have backlash... So, I'd turn OFF backlash settings immediately and see if you can get back to KNOWN realities, not that your backlash settings are affecting the outcome. With it OFF, you should at least be able to see a pattern within the maximum backlash you really have. If you have backlash, you would be best off trying to eliminate it !

I am making the .100" moves in the same direction to eliminate backlash from the equation. Its just not consistent when making small moves for some reason. It is on only one axis, the long 9' axis. I am tackling one axis and problem at a time. I cant eliminate the majority of the backlash mechanically because if you over tighten the drive belts the motor will miss steps due to the tension.

I replaced the drive pulley and put lock tight on the motor shaft. Its not moving. That was my first thought also.

There are so many settings to adjust in Mach3 i just thought there might be something i am overlooking.

I hooked up the cross axis (Y) and tested it and got the same results. A large move away and back hits spot on every time. But a small move in .1" increments in same direction is not consistent. Ill keep playing with it.

My thoughts are that if the error is not consistent, backlash is most likely not the cause. My guess would also be mechanical slop. I'd double check all your pulley's/gears set screws. Make sure anything that moves or is connected to something that does move is solid. I had a similar issue & once I cleaned and reinstalled all of the set screws firmly with locktite, the problems disappeared. This was after I had already accomplished a walk around snugging them up & the issue remained. Only thing else I could think if it's not mechanical is signal interference or bad/intermittent wire connection. Any software issues I can think of would be repeatable and consistent, but others may be able to offer a more "expert" opinion.

Part of the issue may be the direct drive nature of your belt system. Depending on what you read, some say that you can't rely on the microstepping resolution. You're errors are pretty close to your full step resolution.

I think you'd be much better off with a belt reduction. Also, 3/8" belts are pretty flimsy for a 4x8 machine moving at high speeds. Don't be surprised if you get quite a bit of "bounce" when stopping and cornering.

Heres a good read that explains micro stepping problems. Microstepping: Myths and Realities - MICROMO . I think that sounds like the issues im seeing. I will do some experimenting with the drivers in full step mode and see what happens. I think you are right, a reduction pulley would get me the small step resolution i am looking for but also keep the torque benefits of the full step mode. Man i really dont want to redesign this thing but it may have to happen.

Tried with the drive in full step mode and got the same problem. Moving a single step at a time .010", the first few moves step accurate, then it will lose a thousandth or two and a couple more moves it loses a little more. After 10 full steps (.100") it is consistently short about .005" whether im in micro step mode or full step mode. After a long distance the steps equalize out and ends up where it should but over small distances the step amount just isnt accurate. This tells me its the inaccuracy of the motor steps. I need a big reducer to negate the innacurate motor steps. Or possibly more power to make the motor holding position torque better.

I spoke with Gecko's tech support this morning and he told me no doubt it needs to be gear reduced. I just bought some timing gear pulleys and a .08" pitch belt to reduce the gear ratio to 4:1. Will take some hacking to implement it but shouldnt be a major headache. Im sure that will eliminate the majority of the inaccuracy as a single step at 10 micro step setting will be .0001" and a full step would be .001".

Well the gear reduction was the charm. Wasnt able to use the gear ratio i wanted but ended up with a micro step resolution of .0003". Realisticly it wont be holding tolerances of .0003" but the margin of error is small enough now that it should run well. Taking steps of .010" it now hits exactly .010" 9 out of 10 times and when its off it only misses by .001" and the next step hits right on again. Even taking .002" steps it hits no problem. Backlash on the 8' axis is about .007" without using compensation. Id say its good enough to go.