I have a particular interest here. My pet PIC project is the control
of my house lights(, fans, solar water heating etc.) on a centralised
basis.  This includes the touch/ proximity sensors I propose to install
instead of conventional light switches, which will be PIC-based and
execute an algorithm similar to the popular "stud-finder"s.

  One of these latter, instead of using a little pot with an edge-wheel
to set the reference level, has a simple "zero" button.  Probably
analogue sample-and-hold circuitry, but a feature dead easy to implement
with a PIC.  A 12C508 would be perfect for this application?

  My proximity plates will be expected to auto-zero over a time constant
of some minutes (altough a shorter period would most likely do fine) to
compensate for humidity variation, presence of furniture, component
drift, ease of set-up and any other presently unquantified reason.

  Now I have ALSO had embedded in the driveway when it was built, three
arbitarily sized single-turn loops.  These are yet to be connected by a
conduit and boxes alongside (indeed, I have presently mislaid where the
wires emanate from the slab), but I have been thinking on the circuitry
to use.  Note that the loops are actually attached to the reinforcing
mesh in my case, but I think that is common enough practice.

Mike Keitz wrote:

> Offhand I think of two major approaches to building such a detector.
> The first one would make the loop part of an oscillator and detect
> changes in frequency of oscillation.  The second would make the loop
> part of a filter and detect changes in impedance to a signal from a
> constant-frequency oscillator.

  Getting back to the "stud-finder"s, these use a simple R-C time
constant determination directly comparable to the STAMP "pot" function.
This can be very sensitive indeed where long-term stability is not a
concern.  The equivalent L-R version should surely be a third, if not
first candidate for this job?

>  Most likely the loop characteristic is tuned in when the loop is
> installed and any change assumed to be from a vehicle near it.  Some
> sort of auto-calibration could be used but it might get fooled by
> heavy traffic.

  Obviously, a microcontroller can implement various self-calibration
algorithms digitally and these can be defined in a logic, rather than
exponential time domain.  In particular, a digital sensor could
distinguish and note the presence of a car parked near a sensor as well
as other cars passing over it.  Note that for loop sensors, both
increase (ferrous object) and decrease (conductive) in inductance
represent valid events.

> Which naturally leads to a potentially marketable project idea: some
> sort of active or passive device (other than a large block of Fe) that
> could be mounted on a bicycle or motorcycle to help it trigger a
> traffic detector.

  The only passive device coming to mind is a horizontal conductive
shorted turn.  This being rather inconvenient, perhaps a PIC-based
transponder which detects the search coil frequency and generates a
phasing pulse timed to "bend" it.

> if the project  adapts to all the various frequencies and circuits in
> use it would work most everywhere.

  Microcontroller-based for sure.

> Obviously one would need to build up the actual detector or a
> reasonable copy of it to develop this project, since testing in the
> street isn't very practical or safe.

  Safe?  Well, maybe not if you insist on sitting in the middle of the
road, but with some fancy talking, council approval and a quantity of
witches hats...  Practical?  Well, without access to the actual direct
sensor output in the control box, a very SLOW test process, even if you
do it after midnight with no traffic present.

  Cheers,
        Paul B.