Any good ideas for implementing the following.....? LED array

[intellijel]

http://mrl.nyu.edu/~jhan/ledtouch/

[intellijel]

This is how the light sensing could work:

http://haze.concord.org/spworks.html

[intellijel]

"My guess is the 8x8 dot-matrix LED block is scanned in rows (see the

start-up "lamp test" at the very beginning of the video) and uses eight

sense amplifiers via a simple switch array that keeps the active row

from saturating the sense amps.  So, a single instance of the sense

action would be, light up a row, then "scan down" the adjacent unlit row

or rows by sampling the sense amps with an 8-channel A/D (probably in a

PIC or AVR part or some other uC).  Then store the values, change to the

next row, sample...and so on.

For noise immunity (ambient light), perhaps the lighted row is pulsed

at some carrier rate, like..oh, 40KHz (like IR remotes) and the sampled

signals are sifted for the carrier.  Some other bit-frobbery ensues" 

[intellijel]

or perhaps:

if X is a LED that is off, and O is an LED that is on:

This would be a small array of LEDs that areoff (5 x 5 matrix):

X X X X X

X X X X X

X X X X X

X X X X X

X X X X X

If you alternated between the following two patterns quickly enough (e.g whatever minimal frame rate to perceive a constant image)

PATTERN A

X O X O X

O X O X O

X O X O X

O X O X O

X O X O X

PATTERN B

O X O X O

X O X O X

O X O X O

X O X O X

O X O X O

This would mean that any LED you have your finger over would have at least 2 (if at the very side) but usually 4 surrounding LEDs that are on - every other cycle.  So if one cylce you had half the LEDs on thisd array lit up, one could simulatenously scan the light sensed readings of the LEDs that are not on, chances are that any LED you have your finger over, will be reflecting light from the four surrounding LEDs into itself.

It would help to put every LED in holders/tubes that limit the diffusion of the LED so that adjacent LEDs can not directly sense the light of their neighours. If you play with the angle (cone shape of the tubes) and the distance from which you hold your finger to reflect teh light (perhaps mount a clear surface that is not too reflective slightly above the LEDs).

Asumming that you are polling the unlit leds for light readings frequently enough, you could use a simple function to determine if the average light level at that point in the circuit is consistant enough that a coordinate is resolved. (simple form of error correction, lots of other things you could implement)

TOUCH surface

===============

\  / \  / \  / \  /

| |    | |  | |    | |  <---  lighty directing cones

| |    | |  | |    | |

O    O  O    O

LEDs