Good afternoon,

I have a big problem on my CNC with Mach3.
My CNC is equipped with 3 precision screw ( zero backlash ) and 3 industrial servocontroller for stepper motor.
The power supply is controlled by a continuity group.
The PC is a P4, 2.6 GHz, Win XP Pro with all services shut down.
The PC is not connected to Internet ans is not connected on local Ethernet.
All functions on Windows are programmed to have the maximum performance of program ( no stand-by, no screeen-saver, minimum color on graphic controller, no other program installed ).
I have enabled the "enhanced pulsing" and the "sherline mode" on Mach3.
The motor has 800 step/rev , the screw is 5 mm , the maximum speed programmed is 1000 mm/m, the acceleration is 100.

There are not problem if I load a simple program : I can go to a whichever position and a whichever feed rate and the position is ok ( I have used a digital probe to test the position and the machine have no errors ).
I have also tried to make 100 cycle of go-return from 0 to 1000 mm and the final position is OK.

When I load a program with subroutine, circular interpolation, and number of lines higher than 200-300 I have a big problem:
After a complete cycle ( 40 minute of work) the position of 3 axis is not correct !!!!
Each axis have a error from 0.01 to 0.8 mm and depend on how many "step are lost".
I have noticed that every some minute a little "tic" there is on one axis ... the axis lost some steps and the correct position is lost.
I have tested the machine for 3 days and I have not found the cause.

I have also installed an other CNC software but the problem with the other software there isn't.

Another strange thing ......
If I make 3 times the same cycle the error is not the same .... sometimes is bigger, sometimes is lower.

I have tried to test the machine with driver test of Mach3 and is all OK.
The Kernel is set to 25000 Hz.
The Diagnostic window tell me that all is OK.


Have you some idea to solve the problem ?

Thanks
Maurizio

I think you shuld try fitting encoders. maybe your hardware cant keep up with the speed of your program, try runing it on lower speeds.

I have tried at low, medium and high speed.
The result is always the same.
If I work at low speed sometimes is a little worst situation .... the time of cycle is higher and the number of step lost are higher.

If I work at maximum speed sometimes I haven't an error !!!!!!

Maurizio

You seem to be confident that your machine is working accurately. So try this test.

Try moving the cutter round in a circle several times and see if the machine is correct after that. All positions on the circumference of a circle must be approximate (since PI has no finite end) and I am wondering if the approximations are always the same.

Hi jimpinder,

I have tried also this test.
If I make only 10-15 circle the start position and the stop position are the same.

If I make 1000 circle the start position and the stop position are different.... after 100 circle error is 0.02 .... after 200 circle error is 0.08... after 500 circle error is 0.3 mm after 1000 circle error is 0.7 mm.

So if I make 1000 pieces with the same hole in the same position the last piece will have a center of hole different of 0.7 mm from the first piece.
This problem is not visible if I work with other software .... this is the reason that I presume the problem is the Mach3.

From that I doubt the first 10 - 15 are the same - it is just that you can't measure it. Also the fault appears to get worse the more repetitions you do - as though the fault compounds itself.

I cannot think of a simple explanation, because I do not know how the maths stacks up. I know Art in the tutorials says Mach3 works to over 12 decimal places.

The only explanation I can come up with is that when rounding up or down you must make an approximation. If you start from a position of X1 Y1, then after the first sequence the final position might be X 1.000086324874 Y1.000000986362. These might be small enough not to move the steppers an extra step.

If we set off again (without resetting the DRO) then all calculations start from the new position (not X1 Y1) and it may be eventually they trigger off extra steps (or a missing step) in the movement.

I do not think Art would have the DRO's go to zero, because as far as the machine knows it is where it should be.

If at the end of each phase of your test piece (when you are back to where you started) you add in a line G0 X1 Y1 this would reset the DRO's to the original reading. The axis would not move because the amount is too small, but calculations would still start from the same position everytime. It might even be better to include this at the start.

Try this and see if it cures the circular problem - I would be interested to find out.