(re stepper motor for a digital knob thread)
I've just been playing with a 12v unipolar "surplus"
stepper from an old 5.25 disk drive. Connected it to
the CRO and used a clothes peg for an accurate "knob".

Output voltage doesn't seem to be a problem, even the
slowest one-detent moves make +/-200mV or more, 400mV
P/P, a decent speed (like 0.1 rev/sec) gives +/-1v, and
anything like a spin (>0.3 rev/sec) slams the volts into
the 4.6v safety stop (I used a 4.6v zener in parallel
with each coil).

So using 2 transistors biased at 500mV (just off) you
could sense positive going pulses over 100mV, would give
decent "dial" operation for any real twiddling speed.
Yes it is possible to move the knob if you really try
without triggering a 100mV setpoint, but it's not easy.
Ok, so not as good as optical dial encoder on that point,
but better in terms of bearing/shaft/detent/mounting
quality, etc which matter a lot in high use or high
quality application. And probably price too! :o)

I don't think the problem will be detecting pulses but
in the decoding. Detecting only the + pulses is easy and
will give quadrature encoding (90 degree phase) when the
dial is turning. So turning dial will work well. The
problem will be in tiny fractional moves. There were
200 obvious "detents", but some (half!) produce - pulses.

Looking at one coil only, you can turn the knob one detent
and it makes a + pulse. Turn it another detent and it makes
a - pulse.

I really don't know how to get around this problem. For
any real turning speed the thing will work great, but for
fractional moves there will be 1 detent in 4 that makes
- pulses on both coils, which won't be detected.
This doesn't matter if the knob is turned more than a few
detents and also doesn't matter if the tiny movements
don't need to be measured exactly equivalent by degree.

The very easiest solution would be to detect when both +
pulses occur together, then when one ends before the other
the PIC logs a "pulse" in that direction. This would only
increment one pulse per 4 detents, or 50 pulses per shaft
rotation but work very reliably. 2 transistors, 4 1N4004
diodes, handful resistors.
-Roman


Peter L. Peres wrote:
>
> LM393 with + inputs grounded and motor coils between - inputs and gnd.
> Pullups required on outputs. Or LM358 (still with pullups to achieve Vih).
> Maybe add two 1k resistors in series with each coil to avoid pumping the
> supply if the motor is turned very fast.
>
> Peter
>
>
> >Josh Koffman wrote:
> >>
> >(regarding rotary encoders for DIAL use)
> >> There are ways to use a stepper motor as well, though I've never tried.
> >> The ETC Expression lighting console uses just such an arrangement.
> >
> On Fri, 28 Jun 2002, Roman Black wrote:
> >I *really* like this idea! :o)
> >Having replaced optos (darn unreliable things) in
> >so many VCR optical "dial controls" the idea of using
> >a $1 surplus stepper motor is brilliant. The detent
> >"spring" will never wear out as it is the magnetic
> >field of the motor, and with such low currents and
> >heating the thing will work for about a zillion years,
> >or until the bearings give out. Strong panel mounting,
> >all-metal construction, solid bearings, hardened shaft,
> >etc etc. Now cheap old-style steppers are available
> >everywhere from the junk guys this sounds like a REALLY
> >GOOD digital dial.
> >
> >Now any takers on the simplest circuit to connect one
> >to 2 PIC digital inputs?? :o)

--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.