Hey,

I am having a problem calibrating my axis'.  When using the auto calibration feature of Mach3 I can't get it to be accurate to within 30 thousandths of an inch.  I don't think that its my measuring apparatus because it is accurate to one thousandth.  Also, it acts weird in that multiple calibrations will yield better results up to a point (about a 30 thousandth as an error) and then it gets increasingly worse and the error increases.  I tried messing around with some things and couldnt get better results.  What are some tips to getting better accuracy, i.e. configuration settings and geneally good practices? 

Thanks,
Tyler

It is not a good idea to use a measuring system to calibrate your axis. The result can only be as good as your measuring. I cannot get down below 1/1000 of an inch with mine, which would leave Mach totally inaccurate for my purpose - which is turning steel. At 1/1000th of an inch over, a steel axle will not fit a steel wheel.

The only method of calibrating is to calculate your steps per unit.
The first thing to do is to decide on which units you are going to settle on and set the machine accordingly on Config/Select Native Units. This does not mean that you can only use what you select (G20 [inches] or G21[millimeters] takes care of that when running programs) but it is the base system on which you set up your machine and all the subsequent tool and program offsets.

It is normally tied to your leadscrew. If you have an imperial leadscrew  e.g. 10 turns to the inch, or a metric leadscrew e.g. 5mm pitch generally determines which units you must use. You now calculate and enter enter this in Config/Motor Tuning/Steps per unit.

The calculation then goes like this - AND IT IS EASY, AND ACCURATE

Your motors, if standard steppers, will use 200 pulses for 1 rev. (200). Your drives will have microsteps 8 or ten are usual - my Gheckos have 10 (200 x 10 = 2000). You may have step down gearing to your leadscrew, mine is 3 to 1 (2000 x 3 = 6000) and lastly your leadscrew. I had an imperial leadscrew at first - 10 turns to the inch (6000 x 10 = 60,000) I have just changed to a metric ball screw one turn to 5 mm (6000 / 5 = 1,200).

So my answers for steps per unit were 60,000 steps per inch, or latterly 1,200 steps per mm. The two are NOT the same since changing leadscrew altered the whole dynamics of the lathe table and having changed, I had to set up my offsets and tooling again.

This number DOES NOT CHANGE. You cannot alter it to improve accuracy, it is a matter of fact. If your calculation is clearly way out, you can use a measuring system to check it, but the calculated system is the one that is accurate. i.e. in my case 1,200 steps per mm, not 1,195 or 1,204 or any near number - 1,200 is the correct answer.

Once you have caluculated and entered the figure, then you can check if your table is moving correctly. If it is not, then some other factor is affecting it, but not steps per unit. This may be backlash, you may be missing steps for some reason, or whatever, but deal with that when you have your steps per unit set accurately.

Hi Jim

The one flaw to your theory is when you do not know the thread pitch or the thread pitch is not accurate.

With slip gauges it is perfectly OK to use the Mach Calc Function & get perfect results to the nearest 0.0001"

Phil_H

I have an accurate DRO on my X and Y and my calculated steps per never match the DRO. I may have to add or subtract a few steps to be accurate. I also make a very long run, say 24" and compare that to the DRO.
Given that ballscrews do have an error per foot, you never know what or where that is exactly. I find that pitch diameter is more accurate than steeth per pulley also. Taken from the Browning catalog, a 60 tooth has a PD of 7.162 and a 20 tooth is 2.387. You don't know if these numbers are rounded, but dividing 60 by 20 is a 3:1 ratio. If you do the PD math you get 3.0004189 ratio. Perhaps not worth mentioning, but if looking for 0.0001" accuracy, every step and rounded number matters.