I’d be curious to hear your experience. How do you come to the conclusion that the hardware is not great?
I ask because the research I’ve been doing lately using data directly from the Linux kernel seems to indicate that actually, the accuracy and resolution of the PixArt PCT3854QR touchpad are very high. There’s a possibility that palm detection is not great coming from the firmware–I’m still investigating. [Update: it turns out that a key remapper I had opted to use was neutring libinput’s “disable while typing” for touchpads!]
However, other “annoyances” I’ve had with the touchpad experience in Linux are happening higher up in the software stack. For example, two-finger scrolling is ALWAYS happening correctly from the data coming from the kernel, but occasionally X11 decides not to scroll for some reason. I suspect the heuristic regarding “two fingers in the same direction, while not intending to click” is subpar.
I’m curious about whether this is happening at the
libinput level, or perhaps another part of the X11 / Wayland stack.
Here are some notes I’ve been taking on the “touchpad” stack in Linux:
How do Touchpad events go from physical movements to app events?
- Finger movements affect capacitive touch circuitry
- Firmware for
Yet to be open sourced by Framework?
- Multitouch events sent to /dev/input/event8
(use udev to find out which file corresponds to the touchpad, e.g.:
$ udevadm info --export-db | grep -5 ID_INPUT_TOUCHPAD=1
- mtdev - type A to slotted type B protocol conversion
- libinput - “common denominator” for all touchpads
- GTK/KDE etc.
- Gesture detection (Touchegg, libinput-gestures, fusuma, gebaar)
- Application responds to touch events (movement, clicks, gestures etc.)
If anyone else is interested in learning and / or hardware hacking with me so we can improve the Linux experience on Framework, please DM me or [edit:] join the https://linuxtouchpad.org effort.
Here’s a video of cleartouch looking at kernel multitouch data: