Anyone who has already tried to use some kind of tablet device for writing should know that there are fundamental differences between screen types.

The most common is the capacitive type, where you use a finger or some kind of conductive pen to write on a glass surface, while the touchscreen device captures movement of the capacitance change through a grid of transparent electrodes on the backside of the glass. This works, but it sucks for writing precise text or drawing sketches. You can find these screens in almost every modern smart phone, tablet PC or kitchen appliance. They are cheap!

Next is the resistive touchscreen, where a small, hard point presses down on a plastic surface. The touch element is composed of two pieces of clear foil, coated with a conductive material. While the two layers stay isolated when there is no pressure applied, the pen forces them together in a certain point, forming a conductive path. By knowing the specific resistance of the surface coating, the circuit can determine the position of the pen tip by measuring path resistance from different edges of the screen. This type was pretty popular in PDAs (which have by now been fully replaced by smart phones, what a shame ;-) ) as well as the almost equivalent navigation assistants – and is not that common anymore. Writing performance is fair but not exceptional, though.

The third kind is the most interesting one. Real tablet PCs (the ones with the flip-over display) have this normal-looking pen with the nylon tip, which you can use to accurately write on the glass/plastic display surface. Many even feature some buttons on the pen, some kind of eraser on the backside – and they are damned accurate! They have another thing in common: Most of them use technology by a company called Wacom, also producer of digital writing and drawing pads for artists.

This type is called a “digitizer screen”, and it uses a sensing panel *behind* the actual display to recognize and track the pen. The digitizer panel contains an amazing set of surface coils to provide an alternating magnetic field through the screen. Inside the pen, there is a resonant circuit which uses the field energy to transmit the button states and even pressure on the pen tip back to the coil. By monitoring the strength of the resonance through different surface coils, the digitizer then calculates the position of the pen above the surface. In other words, you get a high-res info about the pen position (easily above 25.000 points resolution along the surface edge, depending on the digitizer type!), you know the pressure applied, button presses on the pen and even where the pen is when it is not yet touching the surface.

I recently disassembled a trashed tablet PC (Toshiba Portege) with a broken motherboard for interesting parts, and came across this:

The LCD panel is a LTM12C328T type. Attached to the backside is a SU-010-X01 tablet pen digitizer, and the marking on the ICs clearly suggests that it is made by Wacom. This would make a fine graphics tablet – but how to attach it to any other PC?

Further examination shows two Wacom ASICs (W8001) and a board number (PWB-A542-X). No pin markings, though. And, as usual, no datasheets available.

The digitizer was connected to the original mainboard through this 14 pin connector (pin 1 on the top end). All wires end at the LVDS display connector on the motherboard, which is the singular connector for all display-related components in this tablet device. No pinout information is available, however – except for some speculation on the internet.

What we have is: 14 pins total, 3 wired red (1, 13,14), 7 wired blue (6 through 12) and 4 unconnected (2 through 5). I already removed the original wires when I took the pictures, sorry!

From the coloring of the cables it is pretty obvious, that some are different. My assumption about the red ones being power while the blue ones carry data can be confirmed by analyzing the layout of the interface PCB. Pin 14 attaches to the large ground plane – ok. 13 leads through some SMD parts, probably coils, to two different paths, decoupled by capacitors. Probably VCC, ok. The voltage is unknown however, so I just assumed it to be either 3.3V or 5V. Any higher stabilized voltage is uncommon for this type of circuit.

That leaves pin 1, which is also colored red. This one leads directly to the interface chip, but there is a solder bridge (open), which connects it to the GND plane – which makes it GND, but probably some kind of semi-isolated GND for the interface. I closed the solder bridge to nail this to ground. Then I applied 3.3V from my lab supply to the board and probed all remaining pins with the scope.

The first thing I found was that there is no data activity going on with the pen away from the display. In this case, all pins are low except for #12, which has an on-board pullup resistor of 22 kOhms. Maybe some kind of reset signal.

When the pen gets near the display, stuff happens. While it is still a short distance from the surface (even through 15mm of wood table!), pin #6 jumps to high. At the same time, a continuous data stream appears on pin 9. No other activity can be detected though, so I suspect a serial interface. If the pen touches the surface and pressure is applied to the tip, pin 5 also goes high. By now I know that this was only a lucky coincidence, because the panel not necessarily transmits stylus position data right away.

That brings me to a preliminary pinout:

1. GND (logic)
2. ?
3. ?
4. ?
5. TOUCH (active high)
6. PRESENT (active high)
7. ?
8. ?
9. (Serial?) TX
10. ?
11. ?
12. RESET?
13. VCC (3V3)
14. GND (common)

Another assumption needs to be made about the serial interface. First, a back channel is likely to exist, as the PC can usually query the tablet for parameters. This can be deduced from linux Wacom device drivers for the W8001 chipset. Also, the typical baud rate for serial tablets seems to be 19200 or 38400. I am not sure, however, if the tablet supports any type of flow control. The extra connected pins suggest that at least two such lines exist (probably DTR/RTS), but the scope shows no activity.

Now I connected the panel to my linux PC through a MAX232 IC and a cheap Prolific PL2301 serial-USB converter. Pin #9 goes to the PC RX, and I tried pin #10 as PC TX, because the pair #7/8 somehow seemed much more probable to be flow control lines by layout. Next, I managed to find out something about the serial protocol. There appear to be several, which are quite well documented. I guessed the panel to be of the ISD V4 type, because it was previously connected to an ACPI internal serial port – so, ISD probably stands for “Internal Serial Device” or something like that. These devices feature a reduced command set without any response to presence detection, which means that no auto detect is possible and it therefore won’t work with Windows – even if you use tricks like editing the registry for serial tablet detection. The driver needs to KNOW that a panel of this type is present, and where it is. Wacom had a special driver for such tablets, but it is no longer available.

Care has to be taken with the MAX232 though. The panel seems to be 5V-tolerant, but I did not want to take chances. I dropped the MAX232 supply voltage as far down as possible, which is around 4.5V. Depending on the components used, that may or may not work. A proper USB/3.3V serial coverter is the correct choice.

More info about the protocol:

http://sourceforge.net/apps/mediawiki/linuxwacom/index.php?title=Wacom_Protocol_Overview

http://sourceforge.net/apps/mediawiki/linuxwacom/index.php?title=ISDV4_Protocol#Format_of_Stylus_Query_Response

Using a serial terminal, I set the port to 38400 baud and tried to send the ASCII “*” character, which queries the stylus info. No response. The same happened on 19200 baud. This is not surprising though, because the panel only responds to “*” if it is in the stopped mode, so a “0” (zero) has to be sent first. I found that out by probing around on different baud rates and combinations. Finally, the panel responded with 10-byte sequences and could also be switched to stylus streaming mode.

At this point I tried to hook the archlinux XServer onto the device. After installing xf86-input-wacom, I created a new config file in /etc/X11/xorg.conf.d which I named 99-wacom-isd.conf:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16

Section "InputDevice"
Identifier "WACOM ISDV4 stylus"
Driver "wacom"
Option "Device" "/dev/ttyUSB0"
Option "Type" "stylus"
Option "Button2" "3"
Option "AutoServerLayout" "on"
EndSection
 
Section "InputDevice"
Identifier "WACOM ISDV4 eraser"
Driver "wacom"
Option "Device" "/dev/ttyUSB0"
Option "Type" "eraser"
Option "AutoServerLayout" "on"
EndSection

This tells X to try and add this tablet, whenever a new hardware device is plugged in. Note that it does not use a device class, because the tablet will not broadcast its type! After a reboot, the tablet is recognized by default and usable as a pointing and drawing device. If it does not work, chances are that the panel requires a reset (read on below) or that the serial port is not setup correctly (should be 19200-8-N-1, no flow control). This leaves a final pinout (highlighted are all pins used in simplified operation):

1. GND (logic, can be shorted to common by solder bridge)
2. ?
3. ?
4. ?
5. TOUCH (active high)
6. PRESENT (active high)
7. Flow control (?)
8. Flow control (?)
9. Serial TTL TX
10. Serial TTL RX
11. ?
12. RESET (?)
13. VCC (3V3)
14. GND (common)

Some questions are left:

• Is there flow control? What about pins #7,8,11?
• How is the reset handled? Often the tablet will not react to serial commands after power-up and will therefore not interact with the driver. In this case, pin #12 needs to be pulsed low to make it work. It appears as if there is no internal power on reset circuit, so this was probably triggered by the tablet driver after booting.
• What about the pins #2-5? Some devices seem to have an USB port on these, but mine does neither have any kind of pullup on these pins (needed to select USB speed) nor does it react to applying a USB bus.

Now, all that is missing to my very own LCD graphics tablet is the LCD part. Panelook.cn tells me that it is a Toshiba 12.1″ 1024×768 TFT display with CCFL backlight, but again – no datasheet. The interface is probably LVDS, judging from the cables and mainboard trace layout. I still have one of these VGA-LVDS converters laying around, so I will use it for this project. There is also a model with DVI in, to which I will upgrade later on – or replace the panel, something like that.

What is missing is – again – the pinout for the 20 pin panel connector. From examining the panel input header and measuring conductivity, I decided on the following layout:

1. VCC 3.3V (red)
2. VCC 3.3V (red)
3. GND (red)
4. GND (red)
5. LVDS O0- (blue)
6. LVDS O0+ (white)
7. GND (red)
8. LVDS O1- (blue)
9. LVDS O1+ (white)
10. GND (red)
11. LVDS O2- (blue)
12. LVDS O2+ (white)
13. GND (red)
14. LVDS O3- (blue)
15. LVDS O3+ (white)
16. GND (blue, to shielding)
17. N/C
18. N/C
19. GND (blue, to shielding)
20. Unknown signal (blue, to shielding)

This layout is pretty much the standard LVDS header layout, with the ground lines in between and all, so I am pretty confident with this. As for the data format, this can be guessed: 1024×768 is obviously the resolution, and color depth is usually 6 bit if we are talking about a portable screen. 8bit are only common in desktop displays. The converter needs to be jumpered to these values. And – after hooking the panel up to the LVDS converter – it works like a charm!

A little hard to see, but we have picture! This needed some wine until it worked…and the backlight is not yet plugged in.

Finally, a little calibration is needed on the software side to fix the digitizer mapping to the screen surface (xsetwacom) and the DIY paperless scratchpad is finished! Well, it still needs a case of some kind, but some CNC work will fix that.

 Appendix:

• Calibration:

  #!/bin/bash # WACOM SU-010-X01 calibration data   xsetwacom --set "WACOM ISDV4 stylus" area 0 0 24000 18300

  #!/bin/bash # WACOM SU-010-X01 calibration data xsetwacom --set "WACOM ISDV4 stylus" area 0 0 24000 18300

  This script needs to be run after attaching the tablet. It tells the driver to map pixel position 24.000 in X direction (right border of tablet) to the right border of the screen, and the same for the vertical axis. All values are approximated by trial and error, but they fit pretty well. Note that the aspect ratio of your configured active area should be the same as that of the screen, so you may lose some space if you want to write on a (extra) widescreen display or else the drawing might be squeezed slightly. If the attached display is used, everything is fine.

• Backlight:

  

  Again, this time with the backlight enabled. The display is not the best in terms of contrast, I will try to get my hands on a Lenovo X230 IPS tablet screen. In the meantime, it is sufficient as a secondary display. Even though the writing looks kind of crude from the picture, the digitizer picks up every last nuance of the pen movement.
  By the way, the CCFL inverter is a special type. It uses a piezoelectric transformer instead of a standard magnetic one, probably to avoid emissions near the digitizer. Its pinout is #1/2=VCC (>=5V), #3/4=GND, #5=Dimming (connect to VCC), #6 unknown, #7 unknown