On Fri, 12 Mar 2004 16:59:29 -0500, you wrote:

>> From:         Mike Harrison[SMTP:mike@WHITEWING.CO.UK]
>> Sent:         Thursday, March 11, 2004 4:40 AM
>> To:   PICLIST@MITVMA.MIT.EDU
>> Subject:      Re: [PIC:] Chessboard (sensor)
>
>> On Thu, 11 Mar 2004 02:19:57 +0100, you wrote:
>
>>>>
>>>> Permanent magnets can have significantly differing strengths due to =
the way
>>>> they are magnetised. Hall effect sensors have significant  tempco, =
and of
>>>> course there is the earth's magnetic field, and local magnetic =
fields to
>>>> contend with. This would also be somewhat expensive.
>>>>
>>>
>>>zero point adjustment is the magic word... but you're right, this is =
awful to
>>>realize.
>>>
>>>> My money's still on my earlier suggestion of a simple tuned circuit =
in each
>>>> piece, and coils etched into a PCB under the board - cheap & easy to=
 build.
>>>> Even if you couldn't figure out a way to do it in a matrix, the =
switching
>>>> would only need nine 8-1 analogue multiplexers (ultra-cheap 4000 =
CMOS) ,
>>>> and the detection hardware would be simple and cheap - PC generates =
the
>>>> frequencies, and a simple detector generates a DC level for the PIC =
ADC to
>>>> detect the presence of a piece - maybe a handful of passives and an =
opamp.
>>>> Can't see how you could get much cheaper than this.
>>>>
>>>> You could even make it retro-fittable to existing pieces, as you =
could usea
>>>> PCB coil with a cutout for the capacitor, forming a simple disc =
shape that
>>>> can be stuck to the bottom of the piece.
>>>>
>>>> Proximity of the piece will not cause false detection (e.g. if piece
>>>> hovers) - you are simply looking for a resonance at one of 12 =
frequencies.
>>>> The position of the piece within the square should also not be much =
of a
>>>> problem. The proximity of the board coil to the (pcb) coil in the =
base of
>>>> the piece would be so close (2-3mm max) that cross-coupling to an =
adjacent
>>>> piece should not be an issue.
>>>>
>>>> Even if PCB coils turn out to not give enough coupling at a suitable
>>>> frequency, small surface-mount inductors could be used - still =
cheap, and
>>>> very good coupling - I've recently used these to do an inductive =
datalink
>>>> into a potted product.
>>>>
>>>
>.I've got another Idea... Resonant  frequencies aren't that perfect (lot=
 of
>>>tuning and choosing the right frequency for each pice without =
constructing
>>>overcutting resonants).
>>>But we can still go forth with coils. We can put a resonant coil (only=
 one
>>>frequency for all) in each figure in parallel to a load resistor.
>>>If we choose different load resistors, then each figure would draw a =
different
>>>amount of power from the coils under the chessboard, this could be
>>>measured...  You have to admit, that this is a little bit easier than =
using
>>>multiple resonant frequencies, that all want to be scanned on each =
square of
>>>the board.
>
>> No it isn't.
>> We are back to all the possible variations of signal strength giving =
false results, in particular
>> distance from piece to coil.
>> With different resonant frequencies, we only have to detect presence =
or absence of loading at a
>> given frequency for each square, which should give a much more =
reliable and definite result.
>
>Why make it so complicated? You don't need mutiple driving frequencies. =
Just excite the coil
>under each square with a step function and look for the ringing =
frequency. Some amplification
>boosts the signal to the amplitude that can be detected with an analog =
comparator and
>timer/capture module.
>
>John Power

That assumes your tuned circuit has enough Q to get enough rings to =
measure the frequency. It is
also possibly more susceptible to noise. Generating a  sweep is easy =
enough  to do, and then you are
looking at a simple yes/no decision based on the coil loading level - you=
 could probably even just
measure  the DC current into the output driver.=20

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