I'm back on this one again, having made some changes and tried a few things. The problem is a lack of accuracy, after a long program the position drifts. I have replaced the computer, as it was playing up - but I'm fairly sure that wasn't influencing the problems. I have also calibrated the backlash in the leadscrews, as I found about 0.02mm of backlash in X and 0.01 in Y. Again, I don't think this is the root of the problem though -because that wouldn't be affected by the microstepping settings of the drives. I'm happy that the power supply is OK to drive the steppers (as long as speeds are kept to 800mmpm or so).
The problem isn't "odd steps" being lost, it seems far more regular than that. Running a program, then changing the microstepping settings on the drives and running the program again seems to point to a pattern of step loss. I'm wondering of the first step after a change of direction is being lost, or something similar.
The closest I've come to a solution is increasing the speed mach3 is running at and setting the drives at a very high microstepping setting. This doesn't cure the problem, but as I seem to be losing the same number of steps it makes the error smaller. Far from ideal, as it still causes problems if the program is large enough.
Any suggestions welcome, also as to a more scientific testing procedure.
Regards Kevin