Laser accuracy correction

Has anyone thought of the idea of using lasers to automatically try to correct inaccuracies in the Cartesian mechanics? What I mean by that is to have lasers tell the computer exactly where the nozzle is aiming regardless of where the step count has placed it, then if it is off a little to try to shift it back into the right position through a lot of trial and error. You could give it a set number of chances to reposition it or a percentage of accuracy to aim for, and then for speed purposes have it move on. Yeah this would slow things down by A LOT, but I’m thinking that you might be able to get much cleaner pieces this way.

The idea came to me the other night after watching a Nova program about the search for black holes, where the technology was mentioned of using lasers to correct telescope images naturally distorted by looking through the atmosphere.

Edited 2 time(s). Last edit at 10/20/2009 03:31PM by Tornadoboy.

... for a tracking accuracy comparable with 0.1mm and a building area of say 50x50x50cm you need two high precision cameras with estimated 30 megapixel and very good optics or a really sophisticated image-correction-program ... that's 'slightly' above the budget of most reprappers ><

Then linear encoders or triangulating strings with radial encoders sounds better

Viktor

Use rotating mirrors and an optical sensor on the printing head and one on a fixed reference point. Then you can time the duration between hitting the fixed sensor and the head sensor, and then compare that to the duration of an entire period. By using a pyramid made of 4 flat mirrors, and a 6000 rpm motor to spin this assembly, there would be 400 sweeps per second. So, 90 degree sweep is 2500 microseconds, or 40,000 counts of the system clock.

Sketching a setup would suggest the angle of the sweep over a square working area would be approximately between 35 and 70 degrees or so, or between 15000 and 30000 timer counts. So, at the far end, with a build area of 500x500mm, the travel speed of the beam would be 36µm per clock tick. With up to one tick rounding error at both sensors and little less than one tick for calculations, this would just about get us to 0.1mm accuracy. See attached sketch.

However, rotation speed would have to be very constant, the light sensors would have to react very quickly and reliably, mirrors would have to be flat and in general, the laser beam would have to be narrow and reasonably focused and the AVR would have to be able to solve the math quickly and accurately enough. Resolution could be increased a bit by having the mirrors rotate more slowly, Likely we won't need the full 500x500mm build area or could tolerate lower resolution off-center. The interesting observation is that accurate measuring by timing a rotating laserbeam seems (in theory) possible.

The significant amount of complexity involved here for XY positioning seems unwarranted. Would be fun to do though, and mechanically doable, as the exact alignment of the mirrors is not required. I'm just worried that in solving the trigonometric equations, there will be too much accuracy loss on the AVR.

Now, if we could somehow use these rotating laserbeams for 3D scanning as well, that would be interesting.

-Geert

-Geert

There's a company that builds a metrology system that uses this principle: http://www.arcsecond.com/ Not sure if they have patents covering it.

---

Larry Pfeffer,

My blog about building repstrap Cerberus:
[repstrap-cerberus.blogspot.com]