I was doing some googling and searching, and found this interesting article:
http://www.qsl.net/oe5jfl/encoder.htm
I thought maybe somebody might be able to use it in the future...
Cracking project!
A friend used to print encoder disks onto transparency film, but he had some problems with them. They were being used by students, and took some handling, which may have been the problem.
Then he got these 64 slot encoder disks made in stainless steel:
They are etched, like PCB's. Once the tooling is made, they cost about 1 GBP. He is very pleased with their reliability and robustness. If they were used in a 16:1 gearbox, they'd give 10 bits, but not absolute position.
Gray-code is wonderful for giving absolute position.
Lovely.
GB
[edit]I believe I was wrong, they should be 60 slots/rev.[/edit]
Absolute encoder looks ambitious 
A compromise would be to add a second photodetector with a single slot to the slotted-wheel version, then once the thing has rotated enough absolute position can be detected and then maintained in software.
Adding more slots to the second detector spaced out by differing amounts enables less rotation before determining position, as you can count the spacing between these slots and thus know which pair you've seen once you've seen two of them. For a 64-slot wheel you could space at 6,7,8,9,10,11,13 slots and only need to turn about 70 degrees to establish position.
A compromise would be to add a second photodetector with a single slot to the slotted-wheel version, then once the thing has rotated enough absolute position can be detected and then maintained in software.
Yes. That would be an interesting compromise.
Those encoders are for maze solving micromouse robots, which is okay with relative distance, and hence has no index hole for absolute position. It wants to know where it is in the maze, and uses other sensors for that.
Providing the program doesn't drop events (not trivial), and it knows the number of slots, then relative position is okay. The wheels slip a little bit, especially when fast cornering, so the mouse needs to re-calibrate its absolute position anyway by measuring the position of walls and the posts (walls slot into) which are on a known grid spacing.
Finer resolution is useful. When it is accelerating hard, or braking, position and velocity are critical.
I'm told world class mice can accelerate close to g, i.e. 9.8m/second, and can turn in average speeds of 4m/second e.g. http://www.youtube.com/watch?v=gu2vX886Oj8&feature=PlayList&p=8AB7668B1EFC1FDF&playnext_from=PL&playnext=1&index=100
This encoder is on the motor end, with a gearing of 6 to 1, so it gives 6*60 slots/wheel rev = 360/rev.
I think the wheel size was chosen to give 2 or 3 slots/mm of movement across the floor (can't remember).
By detecting both the metal-to-hole, and hole-to-metal transitions, the encoder gives 260 reference pulses. By using a 2nd photo detector, slightly offset, it gives 2(2*60) reference pulses, or 1440 pulses/wheel rev.
Adding more slots to the second detector spaced out by differing amounts enables less rotation before determining position, as you can count the spacing between these slots and thus know which pair you've seen once you've seen two of them. For a 64-slot wheel you could space at 6,7,8,9,10,11,13 slots and only need to turn about 70 degrees to establish position.
We have debated similar schemes. The encoders are made using an analogue process, like making PCB's, so there is a lot of flexibility.
The mouse rotates the encoder 6 or 9 revs/maze cell, and the wheels slip a bit. Also, sensing the distance to walls, and re-calibration from posts gives some absolute position information. So I have never gone any further with it. It costs over 100 GBP for tooling, which puts me off too.
I'd like to use optical encoders in a low-cost DIY robot gearbox for kids we're working on.
Thank you for the suggestion MarkT.
I had been persuaded that it wasn't worth doing a smarter encoder, but maybe the pedagogic value alone makes it well worth re-examining, and the ability to get absolute position might be useful in other types of robot.
You have got me re-motivated!
Cheers!
GB