Touchpad/Testing

From OLPC
< Touchpad
Revision as of 03:50, 2 February 2012 by Martinlanghoff (talk | contribs) (Variables)
Jump to: navigation, search

Touchpad calibration and testing is a complex area, not very well documented in the industry. This page collects useful information when testing and debugging touchpads.

OLPC has shipped 3 touchpads

  • ALPS (in XO-1 CL1 model)
  • Synaptics (in XO-1 CL1A model, XO-1.5, XO-1.75)
  • Sentelic AVC (in XO-1 CL1A model, XO-1.5, XO-1.75)

Right after boot, use "dmesg | grep -i psmouse" to see what protocol is in use. This usually reveals the tp in use.

Variables

These are the variables we find influence TP performance. Our tuning work aims to find a good balance that makes the TP work correctly for a wide range of cases.

Keep in mind that these variables can be pushed to points where the TP misbehaves -- make your finger wet enough, or your power source dirty enough, and no TP will work correctly regardless of vendor and tuning.

Experience indicates that configuration changes that consider only a subset of these variables... are going to break some of the untested cases.

  • Dirty power: if the power supply is connected to a dirty/noisy power strip (ie: one with several other power supplies connected), this can cause issues with the TP input; control for this by also testing on battery power.
  • Approach speed: some TPs calibrate as soon as they sense the finger on the TP -- if your finger approaches the TP very slowly, the calibration loop might catch signal of your finger _near_, but not in contact with the TP, leading to erratic tracking; control for with very slow approach to the TP as well as fast & firm approach to the TP.
  • Firmness of touch: very light contact with the TP can lead to erratic tracking; control for this using a very light touch, as well as firm contact with the TP.
  • Hand placement: the ideal condition for the TP is where only one finger contacts the laptop; so try
    • strictly one finger
    • second hand holding the unit near the antennas / speakers
    • second hand holding the corner of the base assembly
    • second hand near (but not touching) touchpad
    • thumb resting on buttons
  • Lap vs table: body contact /proximity with the laptop can affect TP behaviour -- test with the laptop on your lap, then on a plastic or wooden table.
  • Humidity and temperature: test in hot+humid, hot+dry, cold+humid, cold+dry environments; the top risk is hot+humid.
  • Dust and humidity: dust, specially combined with humidity, poses a hard problem for TPs, test with talcum powder in a humid environment.
  • Sweaty/wet finger: a drop of water or a wet finger are expected to confuse TPs. Sweaty fingers are a hard problem for TPs but should be able to operate passably.
  • Finger size: different finger sizes look very different to the TP software -- test with users with various finger sizes. Ensure you test with children!
  • User variation: Other variations in users bodies and usage patters can affect the TP. Ensure you test with a number of users.

Tests

Standing finger

  • Place your finger in one spot, keep it there for at least one minute. The pointer should not move.

This test is useful to reveal very dirty power sources.

Circles

  • Put your finger on the TP and move the pointer in circles -- for example follow the Sugar Activity icons circle/spiral loosely for at least one minute.
  • Optionally, try to finish the stroke pointing to a specific icon or element on screen.

After a few seconds, a faulty TP will start tracking the circle motion erratically.

Rolling finger

  • Place your finger in one spot of the TP, in a 45 degree angle
  • Without lifting the finger, roll it left and right repeatedly, for at least 1 minute

After a few seconds of finger rolling, a faulty TP will start reporting erratic position jumps.

This test exercises the small finger motions used for precise pointer movements.

Spirals

  • Paint spirals from the outside in, trying to avoid crossing an already-painted line.

This is a good test for precision.

Corner to corner sweep

  • Move the pointer to a corner of the screen
  • In one fast stroke, you should be able to reach the opposite corner of the screen.

This test controls for acceleration -- ideally the configured acceleration allows you to both have precision with slow movements, and enough accel that swift movements can cover the whole screen corner-to-corner distance.

See also