Handset display driver designers get creative

With the multitude of portable systems available on the market today sporting high-resolution displays with rich color depth, it takes effective trade-offs to optimize the design of the display driver ICs. As is the case with most semiconductor development projects, designers of display driver ICs will need to make informed choices if their product is to be commercially successful. They face many interrelated decisions to optimize a design for a particular end application.

For a full archive of articles and related On-Demand seminars, click here

Display drivers are likely the most I/O-intensive chip designs in existence. The dice are typically 10 to 20 times longer than they are wide so that the chips can fit nicely on the display glass beneath the plinth of the phone, as in the recently launched Toshiba W52T handset with wide-VGA (WVGA) resolution. This is the most inefficient die shape for most chips, because there is not much core space available for circuitry. But for display drivers, it is ideal.

Most high-end portable devices currently include a display with quarter-VGA (QVGA) resolution, or 240 x 320 pixels. Providing a unique source driver for each red, green and blue subpixel would require 240 x 3, or 720, source-driver outputs in addition to the 320 gate-driver outputs on the device. Devices with that many I/Os, include QVGA display drivers from Renesas Technology and NEC.

One way to reduce the effective number of I/Os on display drivers is to multiplex the analog output to the red, green and blue subpixels. The drawback, of course, is that motion video on the display will not appear as crisp, since each complete RGB pixel refreshes less frequently. Designers can compensate for that effect with stronger source drivers that transfer the same energy to the subpixel transistor in less time.

Such an approach is almost certainly a requirement as display resolutions increase. The recently launched N903i phone from NEC, for example, sports a full WVGA screen with resolution of 480 x 690 pixels. Without multiplexing the source drivers, the chip would have required 2,550 I/Os just for the gate and source outputs to the screen.

On-chip display RAM
Another challenge chip designers face as screen resolutions increase is how to include sufficient on-chip display memory for an image. Having enough SRAM on-chip to store a single still-image file is attractive to phone makers because of the power savings associated with not having to move bits constantly from the main body of the phone up to the display subassembly. By storing the phone's "wallpaper" image in the display driver itself, the link from the display controller to the driver can be quiet much of the time.

The QVGA display drivers from NEC and Renesas contain sufficient on-chip SRAM to store a full-screen 18-bit image (262,144 colors). This is about 1.5 Mbits. The difference in die area between two parts is largely attributable to the shrink expected by migrating from 0.25 micron to 0.18 µm.

Chips can fit on the display glass beneath the plinth of the phone for the recently launched Toshiba W52T handset with wide-VGA resolution.

Moving from QVGA to VGA or even WVGA places huge demands on the chip designer who wishes to incorporate on-chip display memory. The 0.25-µm WVGA driver from NEC, although large, at 45 mm² (23.8 x 1.9 mm), has only about 144 kbits of traditional six-transistor SRAM. The display memory desired in such a device would be about 5.7 Mbits (480 columns x 690 rows x 18 bits of color). In the process technology used for the chip, an SRAM cell is about 8 µm². Thus, 5.7 Mbits would require about 50 mm² of die area, more than doubling the chip's size.

Clearly, there is much to be gained by migrating to smaller lithography nodes. The problem here is that display driver chips require high-voltage and reasonably high-current output drivers, which are not necessarily available yet at the aggressive process nodes.

One answer is to shrink the color space. Devices are still being launched with 16-bit color (65k colors). For example, Samsung's new Blackjack and RIM's Blackberry still have QVGA resolution and a limited, 16-bit color palette. Those devices are being marketed as business tools, however, so the display specifications fall short of, say, handhelds marketed for streaming video or the family photo album.

Another example of this trade-off is the new Mitsubishi D800iDS, which has two QVGA screens. The primary screen uses 18-bit color, but the second screen--which really is the device's touch-sensitive keypad-- uses only 16-bit color. Apparently, a palette of 65k colors is deemed sufficient for a keyad application.

Another approach to packing a complete image's worth of display memory would be to use one-transistor SRAM technology. That shrinks the cell size by about half. To date, however, Semiconductor Insights has not seen a display driver that uses anything other than a typical, six-transistor cell design.

Bits through the hinge
Clamshell phones have stimulated intense development of the electrical interface between the handset's main body and the display assembly. There are many reasons for such development, including minimizing the width of the connector, reducing electromagnetic emissions, increasing the data rate and, of course, reducing cost.

It appears the few thousand transistors required to implement low-voltage differential-signaling serdes blocks on the display controller and the driver IC are far cheaper than extra wires in the connection between the two. As the speed of these interfaces increases and their power consumption drops, it will be less problematic to transmit images continuously to data display drivers that lack on-chip memory.

There are several such initiatives in place. NEC's Current Mode Average Differential Signaling competes with schemes such as the Mobile Display Digital Interface embraced by Qualcomm, Seiko-Epson and Toshiba. The first such initiative, National Semiconductor's Point-to-Point Differential Signaling, has been embraced by STMicroelectronics, Magnachip, Sharp and HiMax.

The existence of multiple standards is not so problematic, since, in the end, a mobile handset is a closed system. Consumers are not plugging new display drivers into their handsets at home.

Looking ahead, one trend to watch for is the integration of electronics into the display itself. At Semiconductor Insights, we evaluating more and more devices that deploy low-temperature polycrystalline silicon displays. The advantage of LTPS is that more of the electronics can be incorporated into the glass itself, beginning with the drive transistors.

LTPS displays with integrated drive transistors are now found in many Sharp displays, for example. (Sharp has branded its version of polycrystalline silicon "continuous-grain silicon.") The electron mobility in polycrystalline displays is now such that the integration of more electronics can be considered, such as the D/A converters and perhaps display memory.

Although the "glass computer" is still a little futuristic, the trends are interesting to watch. A great quantity of transistors can be hidden in something the size of a typical display-- even a portable one. n

Rob Hilkes is lead analog/ mixed-signal analyst at Semiconductor Insights (Ottawa), a leading technical-analyst company focusing on microelectronics.

 See related image