While 4-wire resistive touchs are starting to feel definitively 1990s, they’re still by far the most accessible and affordable way to add interactivity to your LCD-based projects, and that’s unlikely to change any time soon.

Technically all you need to read data from a touch screen is three ADC pins, and a fourth GPIO (digital I/O) pin.  In reality, things are a bit more complicated.  You also need to map that data to the individual pixels on the LCD screen beneath the touch screen, and you can have variation from one touch screen to another due to manufacturing differences, rotation or offset of the touch screen, etc.

So how do you accurately and consistently map data from your touch screen to an underlying X/Y pixel?  Read on to find out more in the latest installment of EE Bookshelf!

While working on a project a while back with a TFT LCD screen and a built in resistive touch screen, I naively wrote my own touch screen calibration code measuring the four corners of the display, calculating the offest between top and bottom and then dividing some values to guess the location.  It sort of worked, but was off by up to 10 pixels or so because the screen response wasn’t linear.

Being a bit ashamed of the results, I starting wondering how other people calibrate resistive touch screens, and came across a couple really helpful appnotes from Cypress and TI:

• AN2173: Touch Screen Control and Calibration — Four-Wire, Resistive (Cypress)
• Calibration in Touch Screen Systems (TI)

Don’t get too scared away by all the calculations, but both of these app note explain the way that most systems correlate touch screen values to underlying pixels: linear algebra and matrix multiplication.  (If you can’t remember how that works, I’ll sneak in a real book in this review that has a good refresher on this and just about everything else you’ve forgotten since high school or university: K.A. Stroud, Engineering Mathematics 6th Edition, ISBN 0831133279).

Basically, the formula described in these app notes will take into account the rotation of the screen, and variation in sensitivity between the top and bottom of the resistive panel, etc., and deliver a reliable translation between the touch screen and the underlying display.  In my own case, using the methods described in the app notes above, I went from being off by as much as 10 pixels or more to being within 1-2 pixels or where I was touching.

As a compliment to the above app notes, Carlos E. Vidales wrote an article way back when that includes some source code to get the matrix multiplication out of the way.  The accompanying article also explains the same concept if the app notes were a bit hard to digest, but you’ll definately want to have a look at the source code available in a link here: EE Times – How to Calibrate Touch Screens.

If you just want to see some code quickly (based on the work from Carlos Vidales and the app notes above), you can see how I used this to improve my own touch screen results here.