Re: touchscreen calibration again

Oops,

I still have an issue to fix: sometimes, the pen can bounce, i.e. go up
for a few fractions of a second, and go down again on the same location.

When this happens, the data from the latest pen down state are queued,
and my program records them for the next measure point. This causes this
point to be skipped and wrong values to be recorded for it.

I'm going to fix this by checking time information from the touchscreen
device. I will trash all data recorded before I draw the crosshair
corresponding to the new point.

The program should come soon anyway. By the way, I had to reduce the
number of points to harvest (I had roughly 768 points with a 10x10 pt
linear grid). That was too much work and a risk for accuracy. I made my
grid tighter close to the edges (as Alain found that we've got less
linearity there), and looser at the center.

Lau, does this make sense? I don't know about a possible model, but I
hope it won't hurt to have less samples in the center, if this region is
linear enough.

You can already start playing with my first harvest:
http://michaelo.free.fr/pda/h2210/ts_harvest/ts_harvest_witherrors.out
(caution: there are a few incorrect points because of "bouncing").

    Thanks for your help! I hope we will manage to solve our touchscreen
issues soon!

    :-)

    Michael.

> Hello,
>
> My program is ready! I had to investigate some issues with tslib and
> fix them (I will release my program and my fixes within tslib, of
> course).
> I'll give you a special rootfs file (with a shamcop_ts module without
> Alain's initial linearization code) containing the program.
>
> More soon!
>
> :-)
>
> Michael.

-- 
Michael Opdenacker
http://opdenacker.org/