"By the end of 2005, our customers will have shipped tens of millions of units that include our first-generation Xtensa HiFi Audio Engine," said Chris Rowen, president and CEO of Tensilica. "This new generation product supports additional audio standards and is based on our leading-edge Xtensa LX processor, so it's faster, lower power, smaller and more flexible."
This new 24-bit embedded audio processor has inherent advantages over common 16-bit audio processors in use today. It provides superior sound quality of compressed files due to the increased precision available for intermediate calculations in the compression and decompression algorithms. And 24-bit audio is fully compatible with all of the popular audio standards.
Because the Xtensa LX processor with the HiFi 2 Audio Engine is fully programmable, multiple audio standards can run on the same hardware, allowing the same silicon to be used for multiple applications or the same device to play or record audio in different standards.
Sixteen Audio Packages
Tensilica offers sixteen audio packages for the Xtensa HiFi 2 Audio Engine, so these functions can be integrated into a SOC design, with no additional codec software development required.
-- Dolby Digital AC-3 decoder: implements the complete Dolby Digital AC-3 audio decoder with multichannel 5.1 decoding. It supports all standard down-mix modes and the full range of bit and sample rates.
-- Dolby Digital AC-3 Consumer encoder (DDCE): implements the consumer version of the Dolby Digital encoder, supporting 1- and 2-channel encoding over the full range of bit and sample rates.
-- Dolby Digital Plus Consumer decoder: implements the complete Dolby Digital Plus Consumer decoder with multichannel 5.1 decoding as well as the Dolby Digital converter for compatibility with existing Dolby Digital 5.1 receivers. It supports all standard down-mix modes and the full range of bit and sample rates.
-- QSound microQ components: implements QSound's microQ audio components - a MIDI wavetable synthesis engine with a scalable sample set, a multi-channel stereo mixer, a wide range of processes to control frequency response, dynamics and spatial enhancements, as well as full 3D positional audio for games.
-- MP3 decoder: implements MPEG-1, MPEG-2, and MPEG-2.5 (extensions of MPEG-1) layer 1/2/3 (MP3) decoding with support for constant bit rate, variable bit rate, and free formats. Also, it is fully compliant with ISC/IEC 11172-4.
-- MP3 encoder: implements MPEG-1, MPEG-2 and MPEG-2.5 (extensions of MPEG-1) layer 1/2/3 (MP3) encoding with support for joint stereo coding for better efficiency and support for the constant bit rate format.
-- MPEG-4 aacPlus v2 decoder: implements aacPlus v2 decoding which includes support for Parametric Stereo (PS) and SBR and is compliant with the 3GPP specifications for Enhanced aacPlus.
-- MPEG-4 aacPlus v2 encoder: implements aacPlus v2 encoding which includes support for Parametric Stereo (PS) and SBR and is compliant with the 3GPP specifications for Enhanced aacPlus.
-- MPEG-4 aacPlus v1 decoder: implements aacPlus v1 decoding which includes support for Spectral Band Replication (SBR) and is compliant with the 3GPP specifications for Enhanced aacPlus.
-- MPEG-4 aacPlus v1 encoder: implements aacPlus v1 encoding which includes support for Spectral Band Replication (SBR) and is compliant with the 3GPP specifications for Enhanced aacPlus.
-- MPEG-2/4 AAC LC decoder: implements MPEG-2/4 AAC decoding compliant with ISO/IEC 13818-4 and 14496-4. It supports the decoding of MPEG-2/4 AAC LC and LC multi-channel files as well as the decoding of ADIF, ADTS, MPEG-2 and MPEG-4 file format data streams.
-- MPEG-2/4 AAC LC encoder: implements MPEG-4 AAC LC and MPEG-2 AAC LC 2-channel encoding. It supports Constant Bit Rate, Variable Bit Rate, TNS, PNS, Joint Stereo, etc. It also supports ADIF and ADTS encoding formats.
-- WMA decoder: implements decoding of high-, mid-, and low-rate WMA data streams as well as the decoding of ASF (Advanced Streaming Format) data streams and is fully compliant with Microsoft WMA test specifications.
-- WMA encoder: implements encoding of high-, mid-, and low-rate WMA data streams as well as the encoding of ASF (Advanced Streaming Format) data streams and is fully compliant with Microsoft WMA test specifications.
-- AMR narrowband speech codec: implements AMR narrowband speech decoding and encoding and is fully compliant with the 3GPP specifications for the AMR NB codec.
-- AMR Wideband speech codec: implements AMR wideband speech decoding and encoding and is fully compliant with the 3GPP specifications for the AMR WB codec.
Audio-Specific Hardware Extensions
The Xtensa HiFi 2 Audio Engine employs over 300 audio-specific DSP (digital signal processing) instructions, created specifically by the engineers at Tensilica to optimize the Xtensa LX processor for audio. It features a dual MAC (multiply accumulate) architecture supporting both 24x24- and 32x16-bit arithmetic on both MAC units.
The Xtensa HiFi2 Audio Engine employs Tensilica's FLIX (Flexible Length Instruction Extension) technology to provide an efficient architecture to minimize cycle consumption and power for audio applications. This extends the base Xtensa LX configurable processor with a 64-bit instruction format containing dual multiply/accumulate and other DSP operations as well as loads and stores to and from two register files specific to the Xtensa HiFi 2 engine. Numerous features such as saturation, fractional arithmetic, and signedness can be selected with the multiply/accumulate and arithmetic operations to ease codec porting and implementation effort. Also provided are operations to facilitate bit stream access and variable-length (Huffman) encoding and decoding.
The Xtensa HiFi 2 Audio Engine's native 24-bit data precision provides a superior trade-off between sound quality and codec performance compared to many common 16-bit audio processors in use today, while support for 32- and 16-bit data items is also provided. And the HiFi2 Audio Engine is fully compatible with all industry standard audio coding formats, and delivers noticeably superior sound quality even when decoding prerecorded 16-bit encoded music files.
Through a combination of significantly lower power per MHz, and architectural optimization of the new Xtensa HiFi 2 Audio Engine instruction set, the Xtensa HiFi 2 Audio Engine delivers dramatic improvements in energy efficiency that lead to increased battery life for portable and wireless applications. Based on TSMC's 0.13 nm LV process, running at 200 MHz, static power for the Xtensa HiFi 2 Audio Engine is 0.40 mW, dynamic power is 17.8 mW, and total power is 18.2 mW or .091 mW/MHz. Total power for the first Xtensa HiFi Audio Engine was .207 mW/MHz in the same process. Tensilica's first HiFi Audio Engine's worst case required MHz for the MPEG-4 AAC LC encoder is 85 MHz, while the worst case required MHz for the MPEG-4 AAC LC encoder with the new Xtensa HiFi 2 Audio Engine is just 38 MHz. The total energy used for this example (MHz required times power-per-Mhz) is a nearly 5-fold improvement in energy efficiency.
Lower Gate Count
Through careful engineering, and by fully exploiting the new features of the Xtensa LX architecture, Tensilica was able to reduce the overall gate count of the Xtensa HiFi 2 Audio Engine, compared to the first version, resulting in a 20 percent smaller size when targeted to 200 MHz. Total gate count, with the 300 audio-specific instructions, is approximately 78K gates.
Performance to Spare
Because of the efficiency of the Xtensa HiFi 2 Audio Engine architecture, it requires only a fraction of the available MHz to perform the audio decoding and encoding functions. Targeting high-performance applications, it can support speeds greater than 280 MHz in TSMC 0.13 um LV, leaving lots of headroom for other control, media and signal processing tasks.
Pricing and Availability
The Xtensa HiFi 2 Audio Engine is available now and priced at $75,000 for the audio instruction set. The following software codecs are available now: MP3 encode/decode, MPEG 2/4 AAC LC encode/decode, AMR WB encode/decode and the microQ MIDI synthesizer. Others will be available over the next two months. Codecs are priced separately, starting at $37,500 for object code implementations of the MP3 decoder. Use of the Xtensa HiFi2 Audio Engine requires a license for the Xtensa LX processor.
Tensilica was founded in July 1997 to address the growing need for optimized, application-specific microprocessor solutions in high-volume embedded applications. With a configurable and extensible microprocessor core called Xtensa, Tensilica is the only company that has automated and patented the time-consuming process of generating a customized microprocessor core along with a complete software development tool environment, producing new configurations in a matter of hours. For more information, visit www.tensilica.com.
-- Tensilica and Xtensa are registered trademarks belonging to Tensilica, Inc. aacPlus is a trademark of Coding Technologies. All other company and product names are trademarks and/or registered trademarks of their respective owners.
-- Tensilica's announced licensees include Agilent, ALPS, AMCC (JNI Corporation), Astute Networks, Atheros, ATI, Avision, Bay Microsystems, Berkeley Wireless Research Center, Broadcom, Cisco Systems, Conexant Systems, Cypress, Crimson Microsystems, ETRI, FUJIFILM Microdevices, Fujitsu Ltd., Hudson Soft, Hughes Network Systems, Ikanos Communications, LG Electronics, Marvell, NEC Laboratories America, NEC Corporation, NetEffect, Neterion, Nippon Telephone and Telegraph (NTT), NVIDIA, Olympus Optical Co. Ltd., sci-worx, Seiko Epson, Solid State Systems, Sony, STMicroelectronics, Stretch, TranSwitch Corporation, and Victor Company of Japan (JVC).