Touchpad accel, spirals and xset

Paul Fox pgf at
Tue Jan 19 16:52:20 EST 2010

martin wrote:
 > On Mon, Jan 18, 2010 at 8:47 PM, Paul Fox <pgf at> wrote:
 > >  > According to the man page the 2nd parameter is the threshold at which
 > >  > the acceleration value is applied so I'm not surprised that you don't
 > >  > see any difference between 0 and 1.  Even 4 might be hard to see.  I
 > >  > think that to see a difference for fine motor this value will have be in
 > >  > the 5 to 10 range.
 > Right -- I did apply Paul's recommendations. The manpage isn't
 > necesarily going to know all the interactions between our TP, kernel
 > driver and xorg input driver. Experimentation and source review carry
 > more weight.
 > I assume Paul did both -- he's been looking at this for a while.
 > Now, in my experimentation with the values he supplied, it did not
 > make any difference.
 > > martin isn't seeing a difference, which surprises me, because i
 > > definitely saw a difference between 0 and 1.  from the wiki page
 > Ok. Gentlemen, maybe it's a better idea to grab 2 XOs, setup 802 on
 > one, 802B1 on the other, and run a quick test with the stock settings,
 > and other settings you think might help.

okay, i've now done so, and i now think i see what you mean about
    xset m 7/4 0
    xset m 7/4 1

for large, faster movements, the latter ('1') is far more
controlled -- cursor doesn't go as far.  this is the property
which made it attractive to richard and me, when we realized that
large touchpad jumps were being amplified by the acceleration

however, for very small movements (like drawing a spiral), i see
that the former (i.e.  "7/4 0") setting is better -- the cursor
doesn't tend to "get ahead" of the finger as much.  (hard to
describe -- i guess it's just a matter of the cursor moving too

having now observed this, i guess now think our X server actually
does implement what the latest xset man pages describe:

    If the `threshold' parameter is provided and 0, the
    `acceleration' parameter will be used in the exponent of a
    more natural and continous formula, giving precise control
    for slow motion but big reach for fast motion, and a
    progresive transition for motions in between.  Recommended
    `acceleration' value in this case is 3/2 to 2, but not
    limited to that range.

i think setting acceleration to 7/4 as we're doing is a little
aggressive, and i might choose a slightly smaller value (e.g.,
165 or 160/100 instead of 175/100) -- this is just one guy's
opinion.  but i certainly agree now that the threshold arg should
be '0'.

 paul fox, pgf at

More information about the Devel mailing list