Skip to contentSkip to conversation

usb: generic calibration support

PATCH In review Holger Schurig 4 participants 6 messages Nov 16, 2006 – Nov 17, 2006 USB

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 usb: generic calibration support

From: Holger Schurig <[email protected]>

Generic calibration support for usbtouchscreen.

---

With build-in calibration support, the "swap_xy" kernel parameter vanishes and usbtouchscreen instead gains a new kernel-parameter which holds 7 integers.

This is used to calibrate the resulting output of the driver. Let x_o and y_o be the original x,y coordinate, as reported from the device. Then x_r,y_r (the x,y coordinate reported to the input event subsystem) are:

    x_r = ( a*x_o + b*y_o + c ) / s
    y_r = ( c*x_o + d*y_o + e ) / s

The default values for (a,b,c,d,e,s) are (1,0,0,0,1,0,1). To simulate swap_xy, one would set them to (0,1,0,1,0,0,1). Once can also use swap_x or swap_y alone, or define other, linear transpositions. The algorithm used is the same as in Qt/Embedded 3.x for the QWSCalibratedMouseHandler.

This interface allows re-calibration at runtime, without restarting the X-Server or any other event consumer.

Please review this patch and schedule it for inclusion once 2.6.19 comes out.

drivers/usb/input/Kconfig +26 -0
@@ -219,6 +219,32 @@
 	  To compile this driver as a module, choose M here: the
 	  module will be called usbtouchscreen.
 
+config USB_TOUCHSCREEN_CALIBRATE
+	default n
+	bool "Calibration support"
+	depends on USB_TOUCHSCREEN
+	---help---
+	  With build-in calibration support, the "swap_xy" kernel parameter
+	  vanishes and usbtouchscreen instead gains a new kernel-parameter
+	  which hold 7 integers. You can also access this with cat/echo
+	  via /sys/module/usbtouchscreen/parameters/calibration
+
+	  This is used to calibrate the resulting output of the driver. Let
+	  x_o and y_o be the original x,y coordinate, as reported from the
+	  device. Then x_r,y_r (the x,y coordinate reported to the input event
+	  subsystem) are:
+
+	      x_r = ( a*x_o + b*y_o + c ) / s
+	      y_r = ( c*x_o + d*y_o + e ) / s
+
+	  The default values for (a,b,c,d,e,s) are (1,0,0,0,1,0,1). To
+	  simulate swap_xy, one would set them to (0,1,0,1,0,0,1).
+
+	  This interface allows re-calibration at runtime, without
+	  restarting the X-Server or any other event consumer.
+
+	  If unsure, say N.
+
 config USB_TOUCHSCREEN_DMC_TSC10
 	default y
 	bool "DMC TSC-10 device support" if EMBEDDED
drivers/usb/input/usbtouchscreen.c +14 -0
@@ -49,9 +49,16 @@
 #define DRIVER_AUTHOR		"Daniel Ritz <[email protected]>"
 #define DRIVER_DESC		"USB Touchscreen Driver"
 
+#ifdef CONFIG_USB_TOUCHSCREEN_CALIBRATE
+static int cal[7] = {1, 0, 0,   0, 1, 0,  1 };
+module_param_array_named(calibration, cal, int, NULL, 0644);
+MODULE_PARM_DESC(calibrate, "calibration data");
+
+#else
 static int swap_xy;
 module_param(swap_xy, bool, 0644);
 MODULE_PARM_DESC(swap_xy, "If set X and Y axes are swapped.");
+#endif
 
 /* device specifc data/functions */
 struct usbtouch_usb;
@@ -499,6 +506,12 @@
 
 	input_report_key(usbtouch->input, BTN_TOUCH, touch);
 
+#ifdef CONFIG_USB_TOUCHSCREEN_CALIBRATE
+	if (cal[6]==0)
+		cal[6] = 1;
+	input_report_abs(usbtouch->input, ABS_X, (cal[0]*x + cal[1]*y + cal[2])/cal[6] );
+	input_report_abs(usbtouch->input, ABS_Y, (cal[3]*x + cal[4]*y + cal[5])/cal[6] );
+#else
 	if (swap_xy) {
 		input_report_abs(usbtouch->input, ABS_X, y);
 		input_report_abs(usbtouch->input, ABS_Y, x);
@@ -506,6 +519,7 @@
 		input_report_abs(usbtouch->input, ABS_X, x);
 		input_report_abs(usbtouch->input, ABS_Y, y);
 	}
+#endif
 	if (type->max_press)
 		input_report_abs(usbtouch->input, ABS_PRESSURE, press);
 	input_sync(usbtouch->input);
Signature · 5 lines
M&N Solutions GmbH
Holger Schurig
Dieselstr. 18
61191 Rosbach
06003/9141-15
#1 lore raw

hi

Quoted 26 lines · Holger Schurig

From: Holger Schurig <[email protected]>

Generic calibration support for usbtouchscreen.

---

With build-in calibration support, the "swap_xy" kernel parameter vanishes and usbtouchscreen instead gains a new kernel-parameter which holds 7 integers.

This is used to calibrate the resulting output of the driver. Let x_o and y_o be the original x,y coordinate, as reported from the device. Then x_r,y_r (the x,y coordinate reported to the input event subsystem) are:

    x_r = ( a*x_o + b*y_o + c ) / s
    y_r = ( c*x_o + d*y_o + e ) / s

The default values for (a,b,c,d,e,s) are (1,0,0,0,1,0,1). To simulate swap_xy, one would set them to (0,1,0,1,0,0,1). Once can also use swap_x or swap_y alone, or define other, linear transpositions. The algorithm used is the same as in Qt/Embedded 3.x for the QWSCalibratedMouseHandler.

Quoted 6 lines

This interface allows re-calibration at runtime, without restarting the X-Server or any other event consumer.

Please review this patch and schedule it for inclusion once 2.6.19 comes out.

sorry, but i have to give you a big NACK on that one:
- no more modparam: it should be per-device sysfs attributes (swap_xy is basically only for touchkitusb compatibility and shoud be converted to per-device sysfs attribute as well. i just never got to do it)
- calibration can be handled in userspace just fine
- even for in-kernel it's in the wrong place. there are other devices that report raw absolute data...so it would belong to the input layer

Cc'ing Dmitry Torokhov as he might have some comments about calibration support in the input layer.

thanks, rgds
-daniel

#2 lore raw
Quoted 1 line

sorry, but i have to give you a big NACK on that one:

Hehe, I actually anticipated this NACK :-)

Quoted 4 lines

- no more modparam: it should be per-device sysfs attributes (swap_xy is basically only for touchkitusb compatibility and shoud be converted to per-device sysfs attribute as well. i
  just never got to do it)

That would be okay for me.

Quoted 1 line

- calibration can be handled in userspace just fine

Yes it can, and that is the most convinging argument against my patch.

However, user-space calibration often sucks. For example, in X11 the calibration values are stored in /etc/X11/xorg.conf, so a calibration program has to parse & write that file. And I've seen binary-only-calibration programs from vendors
use /etc/X11/XF86Config, which doesn't exist here. Anyway, once you've calibrated you have to restart X-Windows, so you need to terminate all running X applications just because of a silly re-calibration. Not nice.

A calibration at kernel level (may it be input level or driver level) doesn't have this problem. I can make the level persistent via udev, modules.conf or other means. Even when I have a frame-buffer based calibration utility, the calibration would be OK for X11 or GTK/DirectFB as well.

For me, this speaks against "Calibration is purely a user-space problem".

Quoted 3 lines

- even for in-kernel it's in the wrong place. there are other devices that report raw absolute data...so it would belong to the input layer

Suppose I have a graphics tabled and a touchscreen connected at the same time. Can I distinguish this easily at the input layer level?

I can very easily distinguish this at the device driver level. After all, the individual calibration parameters are a property of this individual device.

#3 lore raw

Hi,

Quoted 11 lines · Holger Schurig
Quoted 1 line

sorry, but i have to give you a big NACK on that one:

Hehe, I actually anticipated this NACK :-)

Quoted 4 lines

- no more modparam: it should be per-device sysfs attributes (swap_xy is basically only for touchkitusb compatibility and shoud be converted to per-device sysfs attribute as well. i
just never got to do it)

That would be okay for me.

Actually I think having module parameter is OK as long as we have comlementing device attribute. Then user has an option to either specify default value for the parameter at load/boot time or fiddle with an individual device.

Quoted 5 lines
Quoted 1 line

- calibration can be handled in userspace just fine

Yes it can, and that is the most convinging argument against my patch.

Quoted 9 lines

However, user-space calibration often sucks. For example, in X11 the calibration values are stored in /etc/X11/xorg.conf, so a calibration program has to parse & write that file. And I've seen binary-only-calibration programs from vendors
use /etc/X11/XF86Config, which doesn't exist here. Anyway, once you've calibrated you have to restart X-Windows, so you need to terminate all running X applications just because of a silly re-calibration. Not nice.

I believe tslib handles this.

Quoted 13 lines

A calibration at kernel level (may it be input level or driver level) doesn't have this problem. I can make the level persistent via udev, modules.conf or other means. Even when I have a frame-buffer based calibration utility, the calibration would be OK for X11 or GTK/DirectFB as well.

For me, this speaks against "Calibration is purely a user-space problem".

Quoted 3 lines

- even for in-kernel it's in the wrong place. there are other devices that report raw absolute data...so it would belong to the input layer

Quoted 8 lines

Suppose I have a graphics tabled and a touchscreen connected at the same time. Can I distinguish this easily at the input layer level?

I can very easily distinguish this at the device driver level. After all, the individual calibration parameters are a property of this individual device.

Yes, but evdev is the interface between input layer and userspace. If we ever implement in-kernel calibration I think we'll add recalibrate() method to input device structure and have EVIOCSABS ioctl handler call it. But I'd look at tslib first. I believe Qt, Xserver, etc support it.

Signature · 1 line
Dmitry
#4 lore raw
Quoted 1 line

I believe tslib handles this.

The special X server "KDrive" supports tslib, this is used in many embedded projects, e.g. by images created via
http://www.openembedded.org. But mainline X.org server, e.g. what is in Debian unstable (7.1.0), doesn't support tslib.

Also I don't know if X/tslib allows re-calibration on-the-fly, but I guess it does. However, tslib usually does not work via the input subsystem (/dev/input/eventX is just one of them), most devices use proprietary kernel interfaces.

Qt/Embedded for Qt 2 and Qt 3 doesn't handle tslib out-of-the-box, (heck, the don't even know about /dev/input/eventX), but patches exist.

But I think we can stop this thread now :-)

#5 lore raw
Quoted 6 lines · Holger Schurig
Quoted 1 line

I believe tslib handles this.

The special X server "KDrive" supports tslib, this is used in many embedded projects, e.g. by images created via
http://www.openembedded.org. But mainline X.org server, e.g. what is in Debian unstable (7.1.0), doesn't support tslib.

Someone should add tslib support to the main xorg server then ;-). All tslib supported devices will then just work...

Quoted 2 lines

Also I don't know if X/tslib allows re-calibration on-the-fly, but I guess it does.

It does.

Quoted 3 lines

However, tslib usually does not work via
the input subsystem (/dev/input/eventX is just one of them), most devices use proprietary kernel interfaces.

tslib does work via the input subsystem for almost every 2.6 kernel based device I can think of in OpenEmbedded and we made an active effort to get into that position. The only proprietary interfaces are old 2.4 kernels.

Quoted 3 lines

Qt/Embedded for Qt 2 and Qt 3 doesn't handle tslib out-of-the-box, (heck, the don't even know about /dev/input/eventX), but patches exist.

Qt/E 2/3 have a ton of other input issues beside this (e.e keymap problems) so the lack of merged tslib support is only a minor one and as you say, patches exist.

IMO, calibration in the kernel would be a backwards step (especially when tslib can do more besides like filtering and hopefully in the future rescaling).

Regards,

Richard