So Who Is A Computer Programmer? I Got A Thought.

[strantor]

I've been having synapse firings along these lines for a long time now. "Poor man's DRO." If you havent seen it already, there's a method to pull the position data out of the calibration port on even the cheapest (think harbor freight) of digital calipers.

http://m.instructables.com/id/Reading-Digital-Callipers-with-an-Arduino-USB/

There are more elegant poor man solutions like machine mountable "DRO scales" (calipers with the teeth cut off) for sale on cdcotools.com.

But I think we can do better than all that, and cheaper.

Check out the LDC1000 from Texas Instruments. It's a 24bit Inductance to digital converter. It can measure position change via a change in inductance. That means a spring can now be a position sensor. Think about that; no fragile Glass scales, impervious to dirt (not necessarily metallic debris, kind of a gotcha but not insurmountable ), not affected by temperature, lots of "pros" most importantly "cheap" - if you can't afford a spring just rob one off your screen door.

I've purchased the LDC1000 development kit and screwed around with it. Pretty awesome, but I noticed it's pretty sensitive to external reluctance (if you grab the spring, or just act like you're about to grab it, the position reference changes). But I think this can be cured with external metallic shielding (put it inside a pipe). I'll try that when I get a chance and let you all know how it goes.

Failing that, I think something as simple as a length of nichrome wire could potentially be a reliable position sensor. Think about a potentiometer, open it up, stretch it out the length if your lathe bed (or mill table) and attach the wiper to your carriage. The main problem with nichrome is the temperature coefficient. Differences in temperature from one end of the wire to the other can skew the reading. But take 2 pieces in parallel and apply opposite polarity to each, sum the outputs, and I think that should cancel out the nonlinearity. Or use AC excitation. By my math, on my lathe with 36" of carriage travel, I could get position resolution of .0005" with a run-of-the-mill mill 16 bit ADC. For a few bucks more, a 20 bit ADC gets me down to .00003". Only problem is keeping it clean. It would have to be shielded from debris which might be problematic.

I'm going stop talking and start testing any year now, stay tuned.

 

[chevydyl]

It's a good idea but what about interpolation? That's where your gonna have an issue making radii that are I think tangent to one another, like Jim said the rate at which it can compensate for the backlash, how long does the other axis have to wait for it to comp so they can move together again, and will you see that step? most likely, but it could be very minute

 

[Firebrick43]

The issue of back lash is not programming.(drive/control programming)

Large machines use scales (similar to what your thinking) and encoders. But it's not for backlash. It for thermal compensation (growth,movement of ball screws and ways as they heat up and cool). Some really large machines can be laser measured in a grid pattern and the coordinates loaded in the controller so even if the machine is not perfectly in alignment the tool path (as controlled) will be. So the basis of you idea is out there, just not for the problem your trying to solve.

The problem with back lash is going from plain milling to climb milling. Doing complex curves, pockets, or radius past 90 degrees is going to result in a reversal of directions. When back lash rears it's head in the middle of a cut the tool get broken, parts scrapped, and even machines destroyed.

Even the best of servos and drives can not react to the sudden and violent forces when thrust reverses and backlash is present.

While technically possible to program a machine to put up with backlash, it unnecessarily complicates the cnc program, and takes away many of the possibilities of cnc. You would be better suited just to put power feeds on a manual mill.