DSP algorithms up the beat in in consumer electronics

While most segments of the semiconductor-electronics industry are experiencing contractions and belt tightening and carefully monitoring their order rates, consumer electronics is the place where the boom continues. Silicon integration and high-volume manufacturing are proliferating new generations of consumer electronics equipment, fully-functional at practically throw-away pricing: $69-DVD players, $110-digital camera, and multi-channel audio systems using cheapo plastic speakers yet producing startling full fidelity and sonic imaging.

All thanks to DSPs — dedicated and, increasingly, special-purpose processors that literally "comb-out" a multimedia signal in digital format. The basic DSP operation - the multiple-accumulate - effectively multiplies a digitized analog signal with a modifying number. In that operation, you can make your signal larger or smaller; most importantly, you can extract or change the frequency content.

With this operation, you can suck out the quantization noise produced by a relatively inexpensive A/D converter; you can emulate the digital signal pattern produced by an audio converter with, say, a 12.5-MHz sampling rate; you can extract audiophile sound from stereos sold at Kmart and Radio Shack.

On the picture side, DSPs prevent your snapshots from looking red and scratchy. But did you know that the JPEG photo image compressor keys on color frequencies? Red on one side of the visible spectrum, white on the other — discarding light-colored backgrounds, and compressing the dark-colored foregrounds with same sort of coding schemes used for disk drive data transfers? Amazing what a high portion of the engineering community uses digital cameras, yet hasn't thought through the DSP chain that makes them work.

In this week's In Focus, we've asked contributors to discuss some of the DSP algorithms that form the heart of modern consumer electronics appliances: MPEG4, JPEG (audio and video, encoding and decoding) — as well as the advanced algorithms under development for broadcast purposes: H.264, digital water marking and JPEG2000.

Mike Haidar, product line director at Analog Devices Inc., provides insight into the audio features — not just surround sound, but many special effects — available with the application of 32-bit DSPs to the audio soundtracks of DVDs. And, software manager, Youngjun Yoo of Texas Instruments — who's DSP has a lion's share of designs in digital still camera — offers a contribution on baseline JPEG decoding. New variations include a sequential lossless mode, a progressive DCT-based mode, and hierarchical mode which reconstruct higher quality images. "But all JPEG decoders must support the Baseline version, even if other features within the standard are leveraged," he notes in his article.

Equally informative are the contributions on JPEG compression by engineers, David Katz, and Rick Gentile of Analog Devices and JPEG2000 — a form of motion JPEG useful for compressing digital video — by Jill Hunter of Amphion Semiconductor. Interestingly, the algorithm can be implemented in an Altera FPGA.

On the MPEG-4 front, we have a piece from Chris Basoglu of Seattle-based Equator Technologies. Basoglu's tutorial is especially useful in visualizing how MPEG4 is served by parallel processing capability.

Meanwhile, engineers, Faouzi Kossentini, Foued Ben Amara, and Ali Jerbi of UB Video (Vancouver, British Columbia) breakdown H.264 video encoding in elaborate detail. The H.264 compression will serve digital broadcast and two-way video conferencing. And, ST Microelectronics' Guillaume Petitjean and Jean Nicolai discuss the details of digital watermarking and copyright services.