Scary Math on the Prop? - Touch Screen Calibration

CassLan

A While back I posted about touch screen recomendations..I've settled on using Elo products...mostly because I feel I can buy them 1 year from now and still get almost the same model. Although they are expensive...I have installed them in many POS applications and they have served in rugged enviornments for years.

I have a 15" VGA Elo Touch Monitor (Serial) Hooked up to a C3....I've written a very linear touch screen driver / calibration routine and its working pretty good...but it could be a tad better, and after some research...there is some serious math going into the correction routines...especially needed with larger monitors.

Most of my research has shown the same formulas.....I understand the concepts....but the calculus used to solve them is...well its been a while.

More importantly...is there a better way to do this on the prop without using such vast resources? I imagine implementing these routines on the prop would be pretty costly.

Ideas and Suggestions are much appreciated.

Attached is my current (working OK) Driver Code, and a document which demonstrates the caclulus formulas I'm curious about.

Thanks,

Rick

lonesock

Just a quick hint...your coordinates transformation will be close to "not rotated", and you can get a decent initial value for the solution from a single calibration point. You could then use a simple iterative scheme (e.g. Jacobi Iteration: http://en.wikipedia.org/wiki/Jacobi_method ) to finalize your coefficients, rather than implementing a more robust solver. With a bit of fixed point scaling, this should not take too much Spin code, and you only have to do the math once per calibration. (Note: Gauss-Seidel will be easier to compute, but Jacobi always seemed a cleaner starting point to me.)

Jonathan

CassLan

Thanks Jonathan, I will check out Jacobi