On Sat, Mar 14, 2009 at 9:09 AM, David Brownell [off-list ref] wrote:

On Thursday 12 March 2009, Eric Miao wrote:

quoted

quoted

This should be a workaround for broken hardware and I don't expect
this sync time on every boards.

The sync time depends on the moment of this sync between two HSYNC,
I don't expect that to be very long,

But how long is that?  1/(50 KHz) or so?

For Sharp Zaurus, let's say a poodle, with PXA255 and QVGA LCD
portrait, a typical of 320 lines per frame (although counting in blank
VSYNC will make this a bit larger), and 60Hz frame refresh rate:

1 HSYNC = 1/(60Hz * 320) = 52.08us

An average sync time would be half of this, say, 26us.

Having another interrupt to trigger the sampling will be a bit
trickier and considering this average of 26us on a 400MHz
PXA255, and another interrupt latency, I don't see too much
benefit of doing that.

And my personal feeling of the final result with the GUI seems
quite responsive, though I didn't stress test that.

quoted

yet the hardware is broken enough to
endure this latency.

Actually I heard from someone else this morning that it's not just
"broken" hardware that sees touchscreen ADC sampling noise related
to hsync.  So this patch makes sense ... although there's no way to
guarantee the SPI requests won't have queueing delays before they
arrive at the touchscreen controller, possibly hitting the next
hsync interval.

It seems that it's surprisingly tricky to get clean samples from
touchscreens.  There's physical oscillation to compensate, which
changes capacitance until it damps down.  There's also electrical
oscillation from charging the various sense lines.  Add to that a
bunch of honest-to-goodness noise associated with hsync and vsync...
yeech!  It would be a Good Thing if someone were to split out the
sample filtering logic from this driver so it could be used with
other touchscreen drivers.

Yes, sample filtering can actually be made input device wise.

There are actually some user-land solution to this, e.g. tslib.
Yet this sync issue seems annoyingly unavoidable even with
a fine grained filter, or maybe I didn't try my best to tweak those
filters to a maximum potential yet, and I gave up on that finally.

- Dave

-- 
Cheers
- eric
--
To unsubscribe from this list: send the line "unsubscribe linux-input" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html