The Sony XEL-1 television is a technological marvel. Released in 2007, this 3mm-thin OLED beauty boasts an incredible 1,000,000:1 contrast ratio in a laughably small 11″ form factor. Nevertheless, at the time it was released the panel in the XEL-1 was nearly twice as large as the next commercially available OLED panel, and it represented a major leap forward for Sony.
Organic LEDs (OLEDs) provide several advantages over other display technologies such as TFT LCDs. Each picture element, or pixel, in an OLED is actually a very small LED emitting monochromatic light. This means that when a pixel in an OLED displays black, zero light is emitted. In contrast, the pixels in a TFT LCD operate by selectively blocking light that is emitted from a CCFL or LED backlight. When a TFT LCD displays black, the pixels block the backlight, but only partially. The light that gets through causes the display to appear lighter and, well, less black.
The dark blacks and brilliant whites of an OLED are described by its contrast ratio. The contrast ratio of any display is the ratio of the luminance of the brightest color to the luminance of the darkest color. A typical TFT LCD contrast ratio is about 3,000:1, meaning that the darkest black is 3,000 times dimmer than full white. As I mentioned earlier, the XEL-1 OLED has a contrast ratio of 1,000,000:1! Talk about the blackest of blacks! How much more black could it could be? The answer is none. None more black.
OLEDs also benefit from their potential as low-power displays. An OLED consumes power in direct proportion to the number of pixels turned on. If the OLED is used to display light text on a dark background or some other similarly sparse image, relatively few active pixels are required. This results in significant power savings compared to a TFT LCD, which must drive the backlight at a uniform brightness across the entire display. Of course, it’s possible for an OLED to consume a large amount of power if all of the pixels are turned on simultaneously. This is called “flashlight” mode. It’s more of a secondary feature.
The future of OLED technology is not difficult to imagine. So when a client came to MindTribe recently with an idea for an advanced concept product, naturally we focused on an OLED display. An uncommonly large OLED, in fact. And despite weeks of searching, we could only find one reliable source for an OLED of the right size…
A Panel Apart
Once past the hesitation of chopping up a $3,000 television for parts, the process of reverse engineering the interface to the XEL-1 OLED panel proved to be an enjoyable challenge. The XEL-1 consists of two primary pieces – the panel and base. Each piece has its own PCB, and the functionality appears to be divided as follows: the panel is responsible for generating the correct voltages and drive signals for the OLED pixels given power and video data, and the base is responsible for everything else. There is a full teardown available over at Bunnie Studios, so we’ll only focus on the most significant bits (MSBs) here.

OLED Video Cable with paired LVDS conductors
The panel connects to the base via two shielded flat flexible cables (FFC). Some very helpful visual cues on the panel PCB, including the components near each cable connector, suggested that one of the cables carries power, while the other carries video. Furthermore, by following the traces from the video connector to a THine THC6LVD104 LVDS receiver, we determined the video is transmitted in a 7:1 LVDS format. This discovery corresponded nicely with the observation that the conductors on the video flex cable are arranged as differential pairs, with ground connections between each pair.
On the output side of the LVDS receiver, the video data feeds into a large Cyclone II FPGA via a 35-bit parallel bus running at just over 37MHz. The 35 bits are divided as follows: 10 bits of color information per channel x 3 channels (RGB) = 30 bits color, 3 timing control bits (HSYNC, VSYNC, Data Enable [DE]), and 2 unused bits. Some simple probing with an oscilloscope revealed the control bit assignments, allowing us to determine the video signal timing. Interesting fact: although the XEL-1 is listed as having a resolution of 960×540, the panel is actually driven at a resolution of 976×548. A close look at the panel—sans metal bezel—exposes the extra pixels masked by a printed border.
A simple test image helped betray the color bit assignments. The image consisted of three vertical stripes, one red, one blue, and one green. Since the video signal refreshed left to right, top to bottom, each color bit strobed during the portion of each horizontal refresh that corresponded to the color of that bit. For example, each blue bit would strobe during the final third of the horizontal refresh interval, whereas each green bit strobed during the middle third of the same interval. Varying the intensity of the colors allowed us to further determine the relative significance of each color bit.

Color bit assignment test image (in case the description was confusing)
Armed with a full mapping of the bits in the video signal, we turned our attention to the non-video signals carried over the video flex cable. Probing with a DMM seemed to indicate that these signals were unused, but a closer inspection with an oscilloscope showed otherwise. Two of the signals collude to establish a full-duplex, 115,200 baud asynchronous serial channel, used in a base-initiated command-response pattern. Two other signals appear to function as standard logic-level control lines driven by the base.
The power cable held no mystery, with significantly fewer conductors arranged into three obvious groups. Five of the conductors are used for common, two for +5 volts, and three for +16 volts. Interestingly, the power brick for the XEL-1 produces a regulated +16V output, and the output of the power brick feeds directly through the base to the panel. Sequencing here was pretty trivial, as the +16 volt rail is on anytime the unit is plugged in, and the +5 volt rail turns on or off with the panel.

OLED panel power connector with obvious pin groups.
It’s ALIVE!
We constructed a test system for driving the OLED using an Atom™-based single board computer (SBC) and a small ARM7 development board from Olimex. The SBC was selected because it provided direct LVDS video out through a Hirose DF13 connector. An adapter cable jumpered the LVDS output of the SBC to a proto-board with a flex connector. The microcontroller was also patched into the flex connector to drive the control signals and serial interface.
Mimicking the video timings of the original base electronics involved creating a custom display driver using Intel’s Embedded Graphics Driver (IEGD) kit. The IEGD kit is basically a tool for building driver packages for fixed-mode displays, the kind found in ATMs or vehicle navigation consoles. Using the kit, we created a driver for generating a precisely-timed signal on the LVDS port and mirroring that output to the DVI port for debugging. Finally, all that remained was an install process with multiple, unnervingly long display blackouts and one final reboot…

Complete test setup driving OLED and cloned display. You can see on the bench supplies that the panel by itself is consuming just over 12W.
Clearly, this article would not have been written if it hadn’t worked. That’s not to say that it worked the first time. Indeed, it did not, and there were plenty of setbacks that have been omitted here for the sake of brevity (and to make us sound more deft).

Closer shot of the test setup. You can see the Olimex board (top) and the SBC (right) patch into the cable adapter (center), which connects to the panel (left).
It’s clear that OLED technology is shaping up to be the future of displays, though it will be some time before the really large panels become economical for consumer devices. Nevertheless, more and more new products are using OLED technology now for sharper graphics and better power efficiency. Organic Displays. Coming soon to a farmers’ market near you.