Probing direction error compensation

[adatesman]

Had to put this down a long while to cool off and then life got in the way with Pandemic.

Given how simple it was for me to quantify the kinematic error, I’m not sure that compensating for it would be that terribly difficult. After all, there was a cheap Mach3 plugin for it.

In any event, I’d be curious how Centroid’s new probe fares. If someone would be willing to run the same wall following routine as I did, I would be interested in running the numbers. Iirc I did 10 runs at 0.050” step over on a calibrated ~2.000” master ring, with the probe rotated 90 degrees and another 10 runs done, etc.

I suppose it may get quite complicated if backlash compensation wasn’t thought through very well though. I could see that getting difficult.

Which of course begs the question of the backlash compensation accuracy... has anyone ever independently qualified that with a “good” probe? Quite the rabbit hole to go down.

But long story short, I believe that so long as the error is repeatable and quantifiable, a software solution is appropriate. Ymmv. Have long since fixtured the parts on the rotary and probed single direction to avoid the issue. Only reviving this as it came up on the FB group.

[Muzzer]

I'm not convinced that creating a macro would be quite as simple as you seem to think. You've twigged that backlash has a part to play in this.

So perhaps if you are probing a cylindrical bore and know where the axis is, you know which direction is normal to the bore, so can move the probe further than the backlash, approach normally to the surface and add / subtract it the backlash from each probed position. But when you are trying to follow a non uniform wall, you are going to have to probe the next contact in order to know where the normal is likely to be. Then approach it at a normal. But wait, you need to know where the wall goes next in order to know where the normal direction is likely to be. Aaaagh! So you may manage to create a cylindrical bore / boss macro. Beyond that, you can't do a lot of probing with any accuracy.

My solution was to buy a used Renishaw probe. Rather than costing me 100+ hours and $100, it cost me ~$200. When new they claimed ~1um accuracy. That's an order of magnitude better than most of our machines could manage to machine. Don't forget that probing to ridiculous accuracy is pointless on a machine with measurable backlash, not least when you actually apply machining forces to the spindle, table etc.

Can't help thinking there is a danger here of trying to polish a turd. I try to recognise the limitations of my machine and my patience and design around them once the machine has been set up reasonably well. There aren't many things I design and make that require silly precision - what are you trying to make on yours?