I happened to stumble on to a Youtube video demonstrating a homemade laser interferometer, what surprised me was how simple the setup was and how accurate the sucker was.
For those that don't how laser interferometry works. Basically a laser of known wavelength creates a 'reference' beam (a beam thats not actively being interfered with). There are a number of configurations, but they all share the same common idea, the reference beam is split (or another beam is used) and bounced off a target object. The two beams interact with each other, the constructive/constructive interference from the beams is then typically bounced to some sort of detector. Conceptually, its 3 separate beams, the first is the reference beam beamed to the target, the second is the reflected beam from the target, and the third is the combination of the reference and reflected beams.
... So, uh, how does it help measure stuffs? Lets say the object being measured is exactly one wavelength away from the 'emitter' (the point where the reference beam 'ends') the reflected light will be directly in phase with the light from the reference beam, so they add together, and suddenly you have a light at the same wavelength that is twice as bright as it was when it was emitted. If the object is the moved for 1/2 the wavelength, (its not 1.5 y away from the emitter), the reflect light would be 180 degrees out of phase, thus canceling out the light from emitter entirely, the resulting beam would be non-existent. Each pulse (from full brightness to darkness) is 1/2 the wavelength of the light, so if the wavelength is known, and the number of pulses is known and their 'direction' you can precisely tell the position of an object down to 1/2 the wavelength of the light used ...aka really frig'n accurately!
I don't think it would be ... too ... hard to construct a laser interferometer, they've been built since the early 1900's to things in to perspective. It would require a laser of some known wavelength, a beam splitter, a photo-diode sensitive to the light wavelength, and some control electronics.
Anyone think we can design one of these things to help position a mill or lathe?
I don't believe the mechanical aspects are terribly difficult for someone to put together but the electronics are going to be the make or break point. A 1080nm laser would have a "resolution" of 0.00054mm, and every vibration from a machine tool is going to show up. However, assuming you can count all of the vibrations, they should cancel out in the long run. The real challenge is being able to poll the photo-diode fast enough so there is enough data to remove the noise from the machine vibrating.
For example at 1080nm, there would be 47,038 'pulses' per inch. At 100 IPM, that becomes 78,397 poll/second just to keep up with the movement of the object. Now, consider that your milling with a 8 flute end mill at 10k rpm (just throwing out worst case numbers), thats 80,000 impacts per minute, or 1,333 per second. By this point, almost 80k polls/second are needed just to be reasonably sure of the current position. As a factor of safety, it would be nice to increase the polls/second to somewhere around 160k-400k.
160khz - 400khz+ is not really that fast in the micro-controller/microprocessor world, but, in order to 'poll' the photo-diode you're going to need a really fast analog to digital converter, is this even feasible?
Thanks for reading, its kind of long post!
[note]Haha, just noticed my 'diagram' has the resultant beam and photo-diode on the wrong side of the beam splitter..[/note]