RE: [iPAQ] iPAQ touchscreen jitters

From: Christian, Andrew <Andrew.Christian.a.t.compaq.com>
Date: Thu Jan 03 2002 - 12:40:17 EST

Well, ideally it would be in user space. The problem is that the raw
data coming out of the touchscreen is noisy, whereas most graphical
systems assume that you have relatively clean data from your input
device. This is a fundamental physical problem with resistive
touchscreens --- you always get a certain amount of noise from sampling
the A/D, variations in screen pressure, electrical noise in the iPAQ,
etc. It's quite a lot worse than what you get out of a mouse, for
example. Because of this, all touchpad/touchscreen devices include some
digital filtering to clean up the signal. We've always had some
filtering in the Linux device driver: what I've done is extend our basic
digital filtering scheme from two steps to three:

Step #1: Ignore the first N samples (where N is normally 0 or 1)
Step #2: Apply a FIR filter of user specification (default: average
                 the last 4 samples).
Step #3: Toss out any events that are within N pixels of the
                last reported position (default: 1) - this prevents
jitter
                which can occur when the electrical noise is on the
order
                of a pixel or two. [new filter]

Actually, we're using a relatively unsophisticated scheme (compared to
what is common in the industry).

        - Andrew

> -----Original Message-----
> From: Keith Packard [mailto:keithp@keithp.com]
> Sent: Thursday, January 03, 2002 11:36 AM
> To: David Woodhouse
> Cc: Christian, Andrew; Disconnect; ipaq@handhelds.org; Keith Packard
> Subject: Re: [iPAQ] iPAQ touchscreen jitters
>
>
> Around 16 o'clock on Jan 3, David Woodhouse wrote:
>
> > > I've updated the iPAQ touchscreen driver to include an anti-jitter
> > > filter
>
> > Wouldn't this kind of filtering be better in userspace?
>
> Yes. I expect we'll want to change this value on the fly; placing it
in
> the kernel makes that harder.
>
> Keith Packard XFree86 Core Team Compaq Cambridge
Research
> Lab
>
Received on Thu Jan 3 09:40:50 2002

This archive was generated by hypermail 2.1.8 : Tue May 04 2004 - 09:44:20 EDT