jonker@amc.uva.nl (06/20/91)
To construct a real tiny shaft encoder (might cause problems with
the optical part of the encoder) reduce a printout (need not be
very small) of a radial wheel by photograping it (on b/w film?),
and directly using the film as a shaft encoder wheel. That's major
resolution. Perhaps you can make an interference pattern with a
laser diode, and such a wheel. Two optical diodes can detect the
motion of the interference pattern and with that, the direction
of the wheel turning: # laser diode
---------------------- wheel
rotation axis ^ P two photo diodes
The photodiodes must be on the same radial distance to the axis,
close to eachother, perhaps with a slit on top of them ?
any comments, catches or flaws ?
ardjm59@prism.gatech.EDU (MILLS,JOHN M.) (06/21/91)
In article <1991Jun20.094757.237@amc.uva.nl> jonker@amc.uva.nl writes: >To construct a real tiny shaft encoder (might cause problems with >the optical part of the encoder) reduce a printout (need not be >very small) of a radial wheel by photograping it (on b/w film?), >and directly using the film as a shaft encoder wheel. That's major >resolution. This has been an interesting thread. I like the PostScript approach, and have heard of using PostScript to make interference (i.e., holographic) optics by defining the grid pattern, printing it at 300 or more dpi, and then reducing it photographically. What sorts of resolution are people talking about in this thread? 4/rev? 512/rev? .. > Perhaps you can make an interference pattern with a >laser diode, and such a wheel. Two optical diodes can detect the >motion of the interference pattern and with that, the direction >of the wheel turning: # laser diode > ---------------------- wheel >rotation axis ^ P two photo diodes One technique used in encoders is to sandwich a moving and a fixed aperture disk, with different numbers of "slots" in each. If you have 128 slots in one of the disks, and 129 in the other, the "clear spot" between them will go once around the disk as the moving shaft moves only _one_slot_. ("Moire fringe" effect) The light beams should be broad enough to cover at least a couple of slots, and positioned in quadrature (90 Deg. of phase) with the moving "clear spot." You can't use just "any" multiplication factor, since the two beams must each cycle on and off exactly once as the moving disk rotates by one slot. ("In quadrature": second diode should come "on" half way through the first diode's "on" period; first goes "off" half way through the second's "on" period, etc.) Fully decoding the quadrature multiplies the pulses/rev by four, and gives you direction output. HP sells a chip for this, at least they used to. My example probably has too much resolution for a first- cut home project, but the idea is as applicable to home brew as to commercial designs, especially since it makes the gadget _less_ sensitive to precise mechanical and optical alignment. Also, most encoders provide a 1/rev index pulse. This is probably worth having, somewhere on the disk. # diode1 # diodes 2,3 (at 90 deg.circ) ///________________________ ________..._/// fixed disk w/index hole _...____________________ ________..._ moving disk w/index hole (radial grid lines) ^ pivot O sensor1 O sensors 2,3 (at 90deg.circ) (index) (motion) (We are pushing the envelope of my ASCII drafting, but I'm not much better with a pencil [8*P). >The photodiodes must be on the same radial distance to the axis, >close to eachother, perhaps with a slit on top of them ? BTW, there is no reason to use _disks_ if you would rather use a transparent _cylinder_ shape. Circumferential resolution is the limiting factor (along with alignment), but the encoder can be laid out along the shaft if you prefer. Reflective/non-reflective layouts could also work, but you need one transparent element to achieve the Moire interference. The original post restricted itself to encoders, but I believe the context was manipulator readout. A/D converters and multiplexers, and one- or multi-turn pots are an alternative that is cheap/axis, and requires _no_ tricky mechanical development. Unimate uses pots for coarse feedback on the Puma (I think). They don't have the wear-free attraction of encoders, but have some advantages: (1) absolute output (2) "missed-pulse" errors don't persist (3) fairly easy to analyze I know they have problems, too, but this simple alternative seems to have gone unmentioned. >any comments, catches or flaws ? I'm sure there are some in my post, but I wanted to get the general idea out. -- MILLS,JOHN M. Georgia Institute of Technology, Atlanta Georgia, 30332 uucp: ...!{decvax,hplabs,ncar,purdue,rutgers}!gatech!prism!jm59 Internet: jm59@prism.gatech.edu