Ok, an update on this after a few hours of experimentation.
The initial setup just moving the table back and forth gave backlash figures as follows...
X0.016Y0.034Z0.06 mm
I've now written two programs as follows...
(Check for backlash in clocking position)
(by approaching it from both directions)
G01Z3F100
G01X0.1Y0.1F10
G01X0Y0
G01Z0F100
G01Z3
G01X-0.1Y-0.1F10
G01X0Y0
G01Z0F100
(Check for backlash in clocking position)
(by approaching it from both directions)
(Faster and further for double checking)
G01Z3F100
G01X1Y1F20
G01X0Y0
G01Z0F100
G01Z3
G01X-1Y-1F20
G01X0Y0
G01Z0F100
Running these two programs on a hole clocked to a TIR of <5microns it repeats within 5microns so I think it's fair to say that the original compensation values arrived at by the first method were rubbish.
The new ones that gave the above results are as follows...
X0.006 Y0.010 I've set the Z to 0.02mm as an arbitrary value for the moment.
Frankly I don't believe the figures here either. I've been using this machine with a different controller for years and reckon the backlash is at least twice the figures shown. I'm wondering if the backlash compensation uses twice the value you enter ie a +/- figure is what you enter? It would be good to know.
I've just used the MPG to move first -1mm and then +1mm from the clocked position, and it doesn't repeat, there's an error of 0.01mm
I've also put the figures in using the MDI input and I don't think that repeats either.
So could someone from Mach tell me what the backlash figures mean, and whether the MPG and MDI methods of input use backlash compensation?
If you're just making things from scratch without precision references then it's not going to impact much, the errors aren't huge. However, if you're really trying to make something very accurate then you can't afford to be losing 10-20 microns when you clock something up using a hole as a reference. For the time being, I'm just going to have to use that little program every time I clock up a hole to make sure that it comes back to the centre when approached from either direction.
I'd be curious to know if anyone else has experienced these difficulties.