tslib consists of the library libts and tools that help you calibrate and use it in your environment. There's a short introductory presentation from 2017.

If you have problems, questions, ideas or suggestions, please contact us by posting to our mailing list.

Visit the tslib website for an overview of the project.

• setup and configure tslib
• filter modules
• the libts library
• building tslib

tslib runs on various operating systems, including GNU/Linux, FreeBSD or Android/Linux. See building tslib for details. Apart from building the latest tarball release, running ./configure, make and make install, tslib is available from the following distributors and their package management:

• Arch Linux and Arch Linux ARM - pacman -S tslib
• Buildroot - BR2_PACKAGE_TSLIB=y
• Debian - apt-get install libts0 libts-bin libts-dev

TSLIB_TSDEVICE          TS device file name.
                        Default (inputapi):     /dev/input/ts
                        /dev/input/touchscreen
                        /dev/input/event0
                        Default (non inputapi): /dev/touchscreen/ucb1x00
TSLIB_CALIBFILE         Calibration file.
                        Default: ${sysconfdir}/pointercal
TSLIB_CONFFILE          Config file.
                        Default: ${sysconfdir}/ts.conf
TSLIB_PLUGINDIR         Plugin directory.
                        Default: ${datadir}/plugins
TSLIB_CONSOLEDEVICE     Console device.
                        Default: /dev/tty
TSLIB_FBDEVICE          Framebuffer device.
                        Default: /dev/fb0

• On Debian, TSLIB_PLUGINDIR probably is for example /usr/lib/x86_64-linux-gnu/ts0.
• Find your /dev/input/eventX touchscreen's event file and either
  • Symlink ln -s /dev/input/eventX /dev/input/ts or
  • export TSLIB_TSDEVICE /dev/input/eventX
• If you are not using /dev/fb0, be sure to set TSLIB_FBDEVICE

This is just an example /etc/ts.conf file. Touch samples flow from top to bottom. Each line specifies one module and it's parameters. Modules are processed in order. Use one module_raw that accesses your device, followed by any combination of filter modules.

module_raw input
module median depth=3
module dejitter delta=100
module linear

see the section below for available filters and their parameters.

With this configuration file, we end up with the following data flow through the library:

driver --> raw read --> median  --> dejitter --> linear --> application
           module       module      module       module

Calibration is done by the linear plugin, which uses it's own config file /etc/pointercal. Don't edit this file manually. It is created by the ts_calibrate program:

# ts_calibrate

The calibration procedure simply requires you to touch the cross on screen, where is appears, as accurate as possible.

You may quickly test the touch behaviour that results from the configured filters, using ts_test_mt:

# ts_test_mt

You need a tool using tslib'd API and provide it to your input system. There are various ways to do so on various systems. We only describe one way for Linux here - using tslib's included userspace input evdev driver ts_uinput:

# ts_uinput -d -v

-d makes the program return and run as a daemon in the background. -v make it print the new /dev/input/eventX device node before returning.

In this case, for Qt5 for example you'd probably set something like this:

QT_QPA_GENERIC_PLUGINS=evdevtouch:/dev/input/eventX
QT_QPA_EVDEV_TOUCHSCREEN_PARAMETERS=/dev/input/eventX:rotate=0

For X11 you'd probably edit your xorg.conf Section "InputDevice" for your touchscreen to have

Option "Device" "/dev/input/eventX"

and so on. Please see your system's documentation on how to use a specific evdev input device.

Remember to set your environment and configuration for ts_uinput, just like you did for ts_calibrate or ts_test_mt.

Let's recap the data flow here:

driver --> raw read --> filter --> (...)   --> ts_uinput  --> evdev read
           module       module     module(s)   application    application

Linear scaling - calibration - module, primerily used for conversion of touch screen co-ordinates to screen co-ordinates. It applies the corrections as recorded and saved by the ts_calibrate tool. It's the only module that reads a configuration file.

• xyswap

  interchange the X and Y co-ordinates -- no longer used or needed if the linear calibration utility ts_calibrate is used.

• pressure_offset

  offset applied to the pressure value

• pressure_mul

  factor to multiply the pressure value with

• pressure_div

  value to divide the pressure value by

The median filter reduces noise in the samples' coordinate values. It is able to filter undesired single large jumps in the signal. For some theory, see Wikipedia

Parameters:

• depth

  Number of samples to apply the median filter to

Pressure threshold filter. Given a release is always pressure 0 and a press is always >= 1, this discards samples below / above the specified pressure threshold.

• pmin

  Minimum pressure value for a sample to be valid.

• pmax

  Maximum pressure value for a sample to be valid.

Infinite impulse response filter. This is a smoothing filter to remove low-level noise. There is a trade-off between noise removal (smoothing) and responsiveness. The parameters N and D specify the level of smoothing in the form of a fraction (N/D).

Wikipedia has some theory.

Parameters:

• N

  numerator of the smoothing fraction

• D

  denominator of the smoothing fraction

Removes jitter on the X and Y co-ordinates. This is achieved by applying a weighted smoothing filter. The latest samples have most weight; earlier samples have less weight. This allows to achieve 1:1 input->output rate. See Wikipedia for some theory.

• delta

  Squared distance between two samples ((X2-X1)^2 + (Y2-Y1)^2) that defines the 'quick motion' threshold. If the pen moves quick, it is not feasible to smooth pen motion, besides quick motion is not precise anyway; so if quick motion is detected the module just discards the backlog and simply copies input to output.

Simple debounce mechanism that drops input events for the specified time after a touch gesture stopped. Wikipedia has more theory.

• drop_threshold

  drop events up to this number of milliseconds after the last release event.

Skip nhead samples after press and ntail samples before release. This should help if for the device the first or last samples are unreliable.

Parameters:

• nhead

  Number of events to drop after pressure

• ntail

  Number of events to drop before release

Variance filter. Tries to do it's best in order to filter out random noise coming from touchscreen ADC's. This is achieved by limiting the sample movement speed to some value (e.g. the pen is not supposed to move quicker than some threshold).

There is no multitouch support for this filter (yet). ts_read_mt() will limit your input to one slot when this filter is used. Try using the median filter instead.

• delta

  Set the squared distance in touchscreen units between previous and current pen position (e.g. (X2-X1)^2 + (Y2-Y1)^2). This defines the criteria for determining whenever two samples are 'near' or 'far' to each other.

  Now if the distance between previous and current sample is 'far', the sample is marked as 'potential noise'. This doesn't mean yet that it will be discarded; if the next reading will be close to it, this will be considered just a regular 'quick motion' event, and it will sneak to the next layer. Also, if the sample after the 'potential noise' is 'far' from both previously discussed samples, this is also considered a 'quick motion' event and the sample sneaks into the output stream.

The following example setup

       |--------|       |-----|      |--------------|
x ---> | median | ----> | IIR | ---> |              | ---> x'
       |--------|    -> |-----|      |    screen    |
                    |                |  transform   |
                    |                | (calibrate)  |
       |--------|   |   |-----|      |              |
y ---> | median | ----> | IIR | ---> |              | ---> y'
       |--------|   |-> |-----|      |--------------|
                    |
                    |
             |----------|
p ---------> | debounce | -------------------------------> p'
             |----------|

would be achieved by the following ts.conf:

module_raw input
module debounce drop_threshold=40
module median depth=5
module iir N=6 D=10
module linear

while you are free to play with the parameter values.

The API is documented in our man pages in the doc directory. Check out our tests directory for examples how to use it.

ts_open()
ts_config()
ts_setup()
ts_close()
ts_reconfig()
ts_option()
ts_fd()
ts_load_module()
ts_read()
ts_read_raw()
ts_read_mt()
ts_reat_raw_mt()

Wikipedia has background information.

Usually, and every time until now, libts does not break the ABI and your application can continue using libts after upgrading. Specifically this is indicated by the libts library version's major number, which should always stay the same. According to our versioning scheme, the major number is incremented only if we break backwards compatibility. The second or third minor version will increase with releases. In the following example

libts.so -> libts.so.0.3.1
libts.so.0 -> libts.so.0.3.1
libts.so.0.3.1

use libts.so for using tslib unconditionally and libts.so.0 to make sure your current application never breaks.

If a release includes changes like added features, the second number is incremented and the third is set to zero. If a release includes mostly just bugfixes, only the third number is incremented.

A tslib tarball version number doesn't tell you anything about it's backwards compatibility.

• libc (with libdl if you build it dynamically linked)

• libevdev - access wrapper for event device access, uinput too ("Linux API")
• libinput - handle input devices for Wayland (uses libevdev)
• xf86-input-evdev - evdev plugin for X (uses libevdev)

• ts_uinput - userspace event device driver for the tslib-filtered samples. part of tslib (tools/ts_uinput.c)
• xf86-input-tslib - outdated direct tslib input plugin for X
• qtslib - outdated Qt5 qtbase tslib plugin

This is a complete example program, similar to ts_print_mt.c:

#include <stdio.h>
#include <stdint.h>
#include <stdlib.h>
#include <fcntl.h>
#include <sys/time.h>
#include <unistd.h>

#include "tslib.h"

#define SLOTS 5
#define SAMPLES 1

int main(int argc, char **argv)
{
    struct tsdev *ts;
    char *tsdevice = NULL;
    struct ts_sample_mt **samp_mt = NULL;
    struct input_absinfo slot;
    int32_t max_slots = 1;
    unsigned short read_samples = 1;
    int ret, i, j;

    ts = ts_setup(tsdevice, 0);
    if (!ts) {
            perror("ts_setup");
            return -1;
    }

    max_slots = SLOTS;
    read_samples = SAMPLES;

    samp_mt = malloc(read_samples * sizeof(struct ts_sample_mt *));
    if (!samp_mt) {
            ts_close(ts);
            return -ENOMEM;
    }
    for (i = 0; i < read_samples; i++) {
            samp_mt[i] = calloc(max_slots, sizeof(struct ts_sample_mt));
            if (!samp_mt[i]) {
                    free(samp_mt);
                    ts_close(ts);
                    return -ENOMEM;
            }
    }

    while (1) {
            ret = ts_read_mt(ts, samp_mt, max_slots, read_samples);
            if (ret < 0) {
                    perror("ts_read_mt");
                    ts_close(ts);
                    exit(1);
            }

            for (j = 0; j < ret; j++) {
            	for (i = 0; i < max_slots; i++) {
			if (samp_mt[j][i].valid != 1)
				continue;

			printf("%ld.%06ld: (slot %d) %6d %6d %6d\n",
			       samp_mt[j][i].tv.tv_sec,
			       samp_mt[j][i].tv.tv_usec,
			       samp_mt[j][i].slot,
			       samp_mt[j][i].x,
			       samp_mt[j][i].y,
			       samp_mt[j][i].pressure);
                    }
            }
    }

    ts_close(ts);
}

If you know how many slots your device can handle, you could avoid malloc:

struct ts_sample_mt TouchScreenSamples[SAMPLES][SLOTS];

struct ts_sample_mt (*pTouchScreenSamples)[SLOTS] = TouchScreenSamples;
struct ts_sample_mt *ts_samp[SAMPLES];
for (i = 0; i < SAMPLES; i++)
        ts_samp[i] = pTouchScreenSamples[i];

and call ts_read_mt() like so

ts_read_mt(ts, ts_samp, SLOTS, SAMPLES);

tslib is cross-platform; you should be able to run ./configure && make on a large variety of operating systems. The graphical test programs are not (yet) ported to all platforms though:

This is the hardware independent core part: libts and all filter modules as shared libraries, build on the following operating systems.

• GNU / Linux
• Android / Linux
• FreeBSD
• GNU / Hurd
• Haiku
• Windows
• Mac OS X (?)

This makes the thing usable in the read world because it accesses your device. See our configure.ac file for the currently possible configuration for your platform.

• GNU / Linux - all (most importantly input)
• Android / Linux - all (most importantly input)
• FreeBSD - almost all (most importantly input)
• GNU / Hurd - some
• Haiku - some
• Windows - non yet

Writing your own plugin is quite easy, in case an existing one doesn't fit.

• GNU / Linux - all
• Android / Linux - all (?)
• FreeBSD - all (?)
• GNU / Hurd - ts_print_mt, ts_print, ts_print_raw, ts_finddev
• Haiku - ts_print_mt, ts_print, ts_print_raw, ts_finddev
• Windows - ts_print.exe, ts_print_raw.exe ts_print_mt.exe

help porting missing programs!

This can be any third party program, using tslib's API. For Linux, we include ts_uinput, but Qt, X11 or anything else can use tslib's API.