LED Sensor Array

[borkus]

I know you've been working on this for awhile.. I just discovered this led finger pad project a few days ago.

Seems to me your first attempt with several pics, although very educational, was a big waste of uC power
I2C is very handy I use it extensively in some robotics and uC projects.

The LED matrix is now getting a big ways towards using less uCs and more LEDS.  The other thought I had, although I haven't gotten to an assembly phase (a little too busy at the university) is possibly to use a multiplexer and demultiplexer connected to the uC to allow an even larger array to be used.

If you're still around I'd be interested in knowing what you're up to with these, or if the fad has died away.

[justDIY]

Welcome to the forums Borkus

I know you've been working on this for awhile.. I just discovered this led finger pad project a few days ago.

Seems to me your first attempt with several pics, although very educational, was a big waste of uC power
I2C is very handy I use it extensively in some robotics and uC projects.

I knew from the get go that it is a waste, but at that time I didn't really care.   Depending on which threads you have / have not read (here and on my blog),  I admit that using "slaves" is not the effective solution, at least, not to recreate the application seen in Jeff's video.

I will have to argue the point however, that a matrix is the way to go for all applications.   Using touch sensing as a replacement for analog inputs such as switches, sliders and knobs, far fewer sensor inputs are required, and using slaves allows for a degree of autonomy in different sub-units of a sensor, for example one slave handling 10 switches, one handling 2 sliders, one handling a knob.

If you're still around I'd be interested in knowing what you're up to with these, or if the fad has died away.

The problem with using LEDs as human interface sensors is they are extremely sensitive to noise and interference.   Jeff's video was shot in a dark room, and I have done much of my development in a similar dark room - this is because with a moderate to high level of ambient lighting, the sensitivity of the LEDs is greatly diminished.   This really is a death blow for their use in any sort of mainstream applications.

The fad as you put it, has died away.   I have learned how it is done, which is my original goal.   I have chosen to not do any more development because I fail to see any practical use for the technology.

[borkus]

By in no means was I trying to diminish the "slave" exercize, i saw immediately the possibilities of doing such a thing, not to mention using i2c in this manner is of much use

I have to give it a shot, its a new thing to me.  I was wondering about the external light issues, which I noticed immediately, because the video was shot in a dark room. 

I have some ideas for constantly calibrating routines that might make a touchpad workable even with changing lighting conditions.

At the moment I have a robotics project, an APRS (amateur radio) digital systems project, and this one is going to be another side project.  It doesn't look like it'll really require more time than the other two so it might get done in parallel

The other problem I see in a main stream application is the power consumption.  I can't believe that a 50+ matrix of leds consumes less energy than a current laptop touchpad   But the fru-fru factor would be a great eye catcher.. A PS2 or USB interface would be a must due to the diminished number of laptops with serial ports.

[justDIY]

I have some ideas for constantly calibrating routines that might make a touchpad workable even with changing lighting conditions.

this is almost mandatory ... what I ended up doing was performing one of two simple alogrythims ... the first was a an attemp to 'subtract out' the ambient light, subtracting the current reading from a contanstly updated "no object" reading.  the second method involves subtracting one reading from the previous reading of that sensor, looking for a "big enough" change.

both of these methods worked to some degree, but only to a point, indoors under low to moderate lighting.   neither would work outdoors, where even a cloudy day was so bright that the sensors were swamped and always read 0 - ok, I didn't actually go outdoors, but I tried it in a very sunny room, which I'd equate in brightness to a cloudy outdoor day - same result.

The other problem I see in a main stream application is the power consumption.  I can't believe that a 50+ matrix of leds consumes less energy than a current laptop touchpad   But the fru-fru factor would be a great eye catcher.. A PS2 or USB interface would be a must due to the diminished number of laptops with serial ports.

there's no conceivable reason to replace existing notebook capacitve sensors with large array of LEDs - you'd loose too much resolution and sensitivity, and greatly increase costs.  you have a good point about adding LEDs to an existing sensor platform, to increase the eyecandy value - however, even this would be detremental to sales of said product ... the "geek" market that would be impressed by such nonsense likely would not have the capital to pay for the extra "useless" feature ...  the "geek" market is a pinhead in size compared to the mainstream market of youthful adult to middle aged white business men, and to a lesser degree women and other racial demographic markets... they simply want a product that is reliable and and fairly priced - they don't want a huge 64 to 128 LED matrix as a touchpad interface

quantum research group, qprox, has some awesome capacitive sensors, which would serve as the main interface on just about any application (they even do knobs and sliders now).   leds (and or LCDs) would be combined with these sensors to provide a visual feedback as to the state of the input - check out the demo videos on their website www.qprox.com

purely from an engineering standpoint, the power required to use a matrix as a sensor is only going to cost you the power for 1 led and the power for your microcontroller.   I suppose there is a way to count the reverse bias current, but that is so small, it matters not.

this is good debate on the technology - I hope you can find an application that would sway me!

[benofxtown]

I've been playing around with this LED touch sensor idea... you've done far better than me to press!

I've been using the pic18f2455 and usbhid to get my data to the pc.

Only problem is, unless its the LED's i've tried (r,g,b,y,w,o ultra bright & regular red) is that using the ADC is a bit crappy. i'm getting about 3+- in a 10bit integer when I put my hand over it, not practical at all.

My room is lit with a 40watt incadesent.

I've tried all sorts to get a *bigger* margin.

My 2 most important findings!

a) IR Leds are really good, > 30% margin.
b) Reading back digitally gives a much better result, but its slow...

I will experiment with reducing the reverse bias time, and see if i can get a quicker digital reading, I imagine that the discharge curve would be the same.

Has anyone found any breed of LED to be better?

Another development, but i'm not sure how the PIC would respond to noise etc... if i take a few samples and give myself a reference band based on these, instead of 0v & 5v maybe that would help, i did try manually adding a few potential dividers to close the gap but the noise was killing it.

I also noted the longer the LED is left in its high-z / 0v state, the bigger the noise, more perculier was that i was picking up a perfect 2.333...hz sine wave? Very odd.

Looking forward to hearing of you're guys & gals developments!

ps: add a ceramic cap to the adc input, but don't ground it, it acts like a proximity sensor... if one could disconnect one side of the LED entirely, i think this would increase the sensitivity.

[benofxtown]

I wonder if OLED has the same effect for both light sensing and electron swapping

Oooh wouldn't that be cool, flexible, interactive, multisensory displays!

You could have roll-away interactive whiteboards...

[omega]

I've hacked up a simple circuit and code to try to test this, and I do indeed see the effect, but like at least one other person here I'm only getting a very tiny fraction of the range, nowhere near enough for anything useful.

Part of what I'm unsure of is how long each of these forward, reverse, discharge, and measure cycles are supposed to take.  The implication is that you can forward bias, then immediately (e.g. a few cycles at 16MHz) reverse it, then immediately measure the voltage.  I've put delays anywhere from a few cycles to a few milliseconds into the process in various places, and so far have succeeded in nothing other than moving the origin of the very small range.

Hardware: ATmega8 at 16MHz with one LED+resistor attached to PORTC.0 and .1 (both analog capable), another LED+resistor lit solid next to it.  LEDs are red 10cd spider/superflux from lsdiodes.com running ~10mA (5V, 270R), measured as 10.2mA.  Cheapo 3mm LEDs either don't work or are so far out of range as to be useless.  I'll have to try some others I have around (superbrights from besthongkong and jeledhk via ebay).

However, what I really want to experiment is the idea of taking this measurement technique into the multi-touch screen realm.  First, watch http://youtube.com/watch?v=UcKqyn-gUbY.  Now, the technique as shown very minimally at http://mrl.nyu.edu/~jhan/ftirtouch/, seems to be a relatively straightforward application of reversed-led sensing.  Sure enough, it's done by the same guy.

The idea appears to be that of putting an array of LEDs around the edge of a piece of plexi or glass, overlayed on top of an LCD (or rear projection as in the video).  Frustrated Total Internal Reflectance (FTIR) basically dictatest that a third medium (a finger) interfering with the boundary between the plexi and the air will cause some scattering of the light.  By firing each LED in turn and measuring the value sensed at each of the others, you can do some kind of mathematical magic and derive the location of every simultaneous touch on the surface.

However, I need a bit more sensing resolution first...    Additionally, a to get a decent refresh, you need faster converters.  I have a couple Philips LPC2102 chips waiting to be played with, and they can do ADC at 400KHz, which is enough for a fairly large number of LEDs (sqrt(400,000 / 50Hz) ~= 90) to wrap around the edges of a surface.

Anyway, before I can think much more about that I need to get what I have working better.  Any hints would be greatly appreciated.

TIA,
   Omega
   aka Erik Walthinsen
   omega@vcolo.com