Am Mittwoch 18 Juni 2008 15:07:51 schrieb Andy Green: > Somebody in the thread at some point said: > |> I'm going to have a look at the touchscreen improvement in kernel stuff, > |> it seems like it should make some kind of solution. > | > | Cool, please also include tslib in your analysis. The dejitter and > > variance > > | plugins do not work for our data -- it seems they need more "base > > accuracy" > > | from the ts, so they would be a candidate to improve as well. > > The problems aren't coming from tslib so I leave that to someone else. > > Our raw data sucks so bad that even though we average it in the driver > over 32 samples before we publish it down the input device, the "tall" > axis averaged data still shows bad behaviour. By the look of it > rejecting excursions alone in the driver should really help that 32-way > averaging to do its job, if it doesn't add excessive delay or > quantization of movement, so tslib massaging might not even be needed.
Ok, even better! :M:
