UCB1300 noise

QNX Newsgroups (Archives) qdn.public.ddk.input
1

I am using the Philips UCB1300 coded, with the QNX supplied touch driver.
When I increase the brightness of my LCD the coupled noise interferrs
drastically with the touch driver. The touch measurements are frequently
railed to the high side. I’ve noticed the software filter that is
implemented in the driver is as such:

  1. take three measurements, one every 10ms
  2. if any two of them are “far” apart, discard all three and start again
  3. take the average of the closest two (not entirely, but close enough :wink:

This filter does not seem to be doing the job very well. Can someone
suggest other algorithms for filtering the touch data?

Thanks,
Mike Kadour

2

What we do for the iPaq (http://qnxzone.com/eQip/) is have a running filter
of the data, I belive it is about 5-10 samples deep, that is reset each
time the pen is pressed/released. This works very well and provides
a system with no jitter. Have not put this into the 38xx touch screen
code yet and the there is a lot of jitter and rapid curved strokes have
jaggies.

chris

Mike Kadour <mjkadour@uwaterloo.ca> wrote:

I am using the Philips UCB1300 coded, with the QNX supplied touch driver.
When I increase the brightness of my LCD the coupled noise interferrs
drastically with the touch driver. The touch measurements are frequently
railed to the high side. I’ve noticed the software filter that is
implemented in the driver is as such:

  1. take three measurements, one every 10ms
  2. if any two of them are “far” apart, discard all three and start again
  3. take the average of the closest two (not entirely, but close enough > :wink:

This filter does not seem to be doing the job very well. Can someone
suggest other algorithms for filtering the touch data?

Thanks,
Mike Kadour

–
Chris McKillop <cdm@qnx.com> “The faster I go, the behinder I get.”
Software Engineer, QSSL – Lewis Carroll –
http://qnx.wox.org/