Digital video is fast becoming the new killer app in signal processing. From surveillance cameras to personal video recorders, from HDTV to the home PC, digital video seems to be everywhere.
This boom in video-capable products is good news for programmable chip vendors.
Video compression algorithms like MPEG-2 and H.264 are extremely computationally demanding. In the past, system designers met this demand with fixed-function hardware, such as off-the-shelf MPEG-2 decoder chips. Although this is a perfectly reasonable (and efficient) approach, it has become less viable in the last couple of years. Why? Because of the increasing flux in video algorithms and product features, particularly in consumer electronics equipment.
Historically, encoding formats used in consumer equipment—such as the VHS video tape format—have remained stable for decades. This is no longer true, however, now that content is readily distributable via the Internet, where new digital video formats are cropping up with alarming regularity. Consumers don't want to hear that their playback device is incompatible with their favorite content; they want to be able to play any content on any device.
If you're selling a DVD player that relies on a hardwired MPEG-2 chip and your customers start hankering to play DivX or Windows Media video clips they've found on the Internet, your product is going to require a lot of rework to support those additional formats. If your product is based on a programmable chip, however, you may be able to add that support quickly, without changing your hardware.
The rapid growth in digital video applications comes at a perfect time in the evolution of processor capabilities. Processors only recently began offering the speed needed for digital video at prices that are low enough for high-volume products. One secret behind the increase in speed is the vast increase in parallelism offered by embedded processors. In some cases, this parallelism comes from features specifically tailored to the needs of video processing. For example, several DSP and general-purpose microprocessor vendors have added video-specific instructions or coprocessors to their products. In other cases, parallelism is more generic, as in an FPGA. In either case, these highly parallel features are typically flexible enough to meet the ever-changing demands of digital video applications.
DSPs, general-purpose processors (GPPs), media processors, and FPGAs are all benefiting from a burgeoning trend toward using programmable technologies for digital video. Each technology has its unique pros and cons and is suitable for some applications but not others. For example, if cost and power are tightly constrained, DSPs often have a strong advantage. GPPs provide high-quality compilers and good support for common operating systems; DSPs often don't. FPGAs can provide outstanding performance, but at the price of lengthy development times.
In this supplement, we explore some key video applications and examine how they are shaping current programmable chip technologies. I believe that digital video is the most significant emerging application for DSP-capable chips and that the next few years will witness a huge upswing in demand for chips with strong video processing capabilities. It's a great time to become educated about the algorithms and programmable chips used in video applications.
[ Jeff Bier is the general manager of Berkeley Design Technology Inc. (www.BDTI.com), the digital signal-processing technology analysis and consulting company. Jennifer Eyre of BDTI contributed to
this column. ]