[sdiy] Rotary Encoder Design

[Ben Bradley]

I thought up something a few years ago that may be useful for the
currently-discussed sequencer, or really any synth device where you
want to read several controls using high resolution with a
microcontroller. I still haven't made a prototype, but there's no
reason for this not to work.

Background:
Available rotary encoders are basically two types:
1. cheap detented mechanical (quadrature output, the whole four cycle
output per detent) with 12 to 24 detents. These are used in car radio
volume and tuning controls, and 90s CRT brightness-and-constrast
controls. Often have a putbutton feature but that's not relevant here.
The price on these is good, in the $1 to $2 range.

2. nice no-detent easy-turning shaft optical quadrature output with a
few hundred counts per revolution. These are used for
higher-resolution needs like (1990s) logic analyzers and digital
oscilloscopes for the "master incremental controller." The prices I've
seen on these are like $30, where you can barely justify one for a
master controller. They give good resolution, but using several would
be too costly to justify.

My Idea:
Use a couple of linear Hall-effect devices fixed at 90 degrees to each
other, right behind a magnet on the end of the shaft encoder. These
are arranged so that as the knob is turned through one rotation, one
output would output one cycle of a sine wave, and the other would
output one cycle of a cosine wave. Reading these in the A/D input of a
microcontroller and doing the appropriate trig function (basically
atan2) would give the angle of the shaft. You could (with appropriate
multiplexing, etc.) read a large number of such encoders at what would
feel like real time with a modern microcontroller. Even cheap plastic
3d-printable parts should get resolution to a couple degrees. Cost of
parts might be $2 or $3 per encoder (not including microcontroller and
multiplexing, which would be shared among the encoders).

The parts I got for this a few years ago are TI DRV5053. It just
occurred to me that perhaps similar are available with a digital
interface, and looking on Digikey finds AS5510 with an I2C interface.
These aren't necessarily optimal or the cheapest, just the first I ran
across, but an I2C interface would be easier hardware-wise than
multiplexing analog voltages.

This doesn't cover displaying the value controlled by each encoder as
it's turned, but that's a "separate function." For the
currently-discussed sequencer you might only need to hear the
oscillator for controlling pitch, and for duration use a two-digit
display.