On Wed, 27 Apr 2011 14:34:24 -0700 (PDT), Mathew Newton wrote:
On Apr 27, 10:12 pm, Owain wrote:
On Apr 27, 8:58 pm, Mathew Newton wrote:

Based on the operation of my automated cat feeder (http://www.newtonnet.co.uk/catfeeder) I've been asked to develop a similar
device for dispensing nuts to birds. ...
What I want to do is be able to do is trigger only a single serving
and, given the somewhat inexact method by which this paddle delivery
method operates, I have decided that my best bet is to somehow detect
the ejection of a bunch of nuts (there'll be less than 10 per serving
i.e. per section of the paddle) once they have left the dispenser.
This detection can then stop the motor turning and await the trigger
for another serving.

Would it be possible to preload the nuts into individual service
portions?

snip perfectly sensible and valid idea!

I should've mentioned already, the decision to re-use the cat feeder
design principle has already been made on the grounds of limited time
for (re)development, parts availability and the proven concept. If the
endless protoyping and trialing taught me once thing it's that there
designing this sort of thing in theory is once thing but turning that
theory into practice unearths all sorts of issues that weren't
immediately obvious! The most common problem is nuts getting trapped
(matron) and this design has proven itself to be completely immune
from that issue. Also, for what it's worth, whilst I called it a bird
'feeder' it's actually going to be used as part of a semi-automated
bird-training programme (action-reward type of thing).

The suggestions based upon paddle position detection (e.g. stepper
motors, flap detectors, shaft angle detection etc) have been
considered but discounted based on the way the dispenser works in
practice. There is a fair amount of paddle distortion to accommodate
the random fall, size and position of feed and the point at which the
food is dispensed really does vary continuously. It's really quite
amazing how such a slow moving paddle (4RPM) can seemingly hold on to
the nuts and so what would've dropped in one position may repeat quite
the same next time. It might be worth further experimentation however.

I've found some microswitches with a 0.078N operating force which
might be low enough to be able to detect the dropping of the nuts
(there I go again) and so might grab a couple of them to see if
they're of any use. Alternatively, they might serve well with a cam
approach should that idea seem like a goer.

Mathew

Just to be clear: this is intended ONLY for indoor use?

I'd consider an acoustic detector. Either an electret microphone or
a piezo insert. Have that attached to the thing the nuts will fall
onto and detect the noise they make. It wouldn't have to be hi-fi,
just an op-amp and a high-pass filter.