Multi-touch: Many questions to one desire....

joerg joerg.twinklephone at
Wed Mar 5 05:04:10 CET 2008

Am Mi  5. März 2008 schrieb Marco Trevisan (Treviño):
> joerg ha scritto:
> > I'm investigating to squeeze some multi-touch properties out of the GTA02 
> > 4wire-resistor touchscreen, by hacking the basics and probing the ts in a 
> > couple non-recommended completely different manners. There still is a 
> > little  hope...
> What's your hope exactly?
> Do you have some "secrets" to share? :P

Well, it's basically a 4wire (2 metal coated transparent plastic foils) 
design: you apply + and - at X+ X- which creates a linear voltage gradient 
over the one foil, and you messure the voltage at any of both Y of the point 
where the second foil touches the first. (and vice versa x<>y, for other 
coordinate). So much for the classics.
However there are a couple of other permutations to apply + and minus and/or 
leave open and where to probe for voltage, given you have 4 wires and 3 
states (+,-,probe+open) for each one of the 4 to choose from. Hard to explain 
without some graphics showing a simplified circuit diagram.

And then, if all this fails, there still remains the dynamic approach 
(pulse-response), for there are two planes creating a R-C-R-C-R... network, 
resembling somewhat like e.g. a coax cable. You know you can analyze coax 
(10Base2 ethernet, even BaseT) for position of sharp bends with an analysis 
of pulse response. Somewhat like hitting a drum and telling from resulting 
sound where there are the 2 stones on the drum membrane.

All this *VERY* theoretical anyway for now, for i got scope etc, but even 
don't have real hw to test with it :-( (who was the guy with a fried NEO and 
no idea what to do with it? I had a real use for it, even better when it's 
broken), and i was too lazy (or too poorly gifted) to do the math yet.
 What i expect/hope: At least i now got diagrams, so i know what _can_ or 
_must_not_ be done with given hw in GTA01/02. Just dreaming for now... ;-) 
But i'm quite sure it's feasible to get at least kinda force feedback for the 
single touch. Maybe more... Amazing what 12 formulae with 5 (10) unknown 
variables(R) can reveal (ok, this are just numbers for instance, didn't the 
math yet).
When it comes to dynamic pulse response analysis - with all sorts of filters 
and the whole scary stuff (any volunteers?), we probably even get the weight 
of user-C touching the screen ;-) In an ideal world. A/D converters of CPU to 
sample the response are not _so_ bad.

All this [patent pending, (C), smells like me etc]:jOERG ;-). At least mention 
me, and send me a beer each day, and this is prior art. Ehrrm.
This week CeBIT, let's see what's my schedule for next week.


More information about the community mailing list