原创 Audio–the Commodity that Isn’t

Audio is becoming pervasive and is being designed into a wide variety of products. Many new chip designs contain an audio processor just in case it’s needed in future product revisions, since an audio processor can often be implemented in under 100,000 gates on a multi-million gate chip. Some consumer devices contain multiple audio processors – a low-power MP3 processor for long-life music playback and a more robust audio processor for gaming, 3D, pre and post processing effects and speech.

The question to ask: Is audio is becoming a commodity? Clearly there isn’t much reason for a company to invest the time and effort to design a new audio processor or audio accelerator block when there are several available from leading IP suppliers that can be dropped into new chip designs. And most companies don’t have the in-house expertise to design an audio processor from scratch. Furthermore, the audio algorithms themselves are mostly patented or highly regulated software packages that are not subject to value-add engineering by even the most diehard do-it-yourselfers. For instance, Dolby controls their patented algorithms and must certify that all implementations meet their exacting standards. Same for Microsoft (WMA) and many others. Clearly, an audio solution – hardware and software – is best “bought” rather than “made”.

The real challenge comes from the wide number of audio encoders and decoders that must be used –there’s really no one single standard. That’s why hard-wired audio subsystems are impractical and programmable processors are evaluated and chosen based on the breadth of codec support. For example, we have over 20 encoders and decoders available for Tensilica’s HiFi 2 Audio Engine. These include encoders and decoders for AC-3, Dolby Digital Plus 5.1 and 7.1, MP-3, MPEG-4 aacPlus v1 and v2, MPEG 2/4 AAC LC, Ogg Vorbis, WMA, and a wide variety of speech standards. New codecs are being added each quarter. In addition there are a number of specialty audio companies like AM3D, QSound, SONiVOX and SRS Labs, who have deep audio systems expertise who have also ported their special effects software to run on Tensilica’s entire family of audio processors.

But all software implementations of encoders and decoders and post-processing software are not created equal, so when evaluating an on-chip audio subsystem, it’s essential to look into the specs. We recently announced that we invested the engineering resources to get our MP3 decoder to run at the lowest power in the industry, requiring just 5.7 MHz when running at 128 Kbps, 44.1 KHz, and dissipating 0.45 mW in TSMC’s 65nm LP process, including memories. Other solutions might run as high as 20-40 MHz. Just think of the difference in battery life this can make at a time when developers of portable MP3 players are requesting solutions for 100 hours of battery life.