Tensilica HiFi  Audio/Voice DSP IP

Home: IP Portfolio > Tensilica IP > HiFi DSPs for Audio/Voice/Speech

Tensilica HiFi DSPs for Audio, Voice, and Speech

Audio, Voice, and Speech Processing Couldn’t Be Easier 

Looking for the best quality audio, voice, and/or speech processing core for your next design? With all the software already ported for immediate deployment?

With the largest selection of audio DSPs, software packages and ecosystem of partners, adding audio to your next SoC couldn’t be easier.  Make the choice that more than 90 licensees have already made — select Cadence® Tensilica® HiFi DSPs for all your audio-, voice-, and speech-processing requirements.

Largest Partner Ecosystem with Expansive Range of Audio Software

Cadence has your audio software needs covered with over 125 partners in its comprehensive audio ecosystem. These companies range from industry giants like Dolby and DTS to smaller companies providing innovative sensor fusion, always-on, sound enhancement, and noise reduction software. 

We have more than 300 audio, voice, speech recognition, and voice enhancement software packages already ported to the HiFi DSP architecture. This means you can get up and running very quickly, and can easily port your own proprietary software, completely in C, while also maintaining or surpassing the performance of assembly on other DSPs.

List of 300+ Software Packages for Audio, Voice, Sensor Fusion and IoT

Table of Partner Audio, Voice, Sensor Fusion and IoT Software

HiFi DSP Customers Are Designing Exciting Products

Tensilica HiFi DSPs are at work in many popular products. Learn more in our Customer Profiles Section

Choose from Five HiFi DSPs for Audio, Voice and Speech

  1. HiFi Mini DSP - Smallest, lowest power DSP for always-listening voice trigger and voice recognition
  2. HiFi 3 DSP - Efficient 32-bit processing for audio enhancement algorithms, wideband voice codecs, and multi-channel audio
  3. HiFi 3z DSP - Most energy-efficient DSP for object-based audio, super-wideband voice codecs, and neural network-based automatic speech recognition (ASR)
  4. HiFi 4 DSP - 2X HiFi 3 performance for DSP-intensive applications like multi-channel object-based audio, digital assistant front-end processing, and neural network-based ASR
  5. HiFi 5 DSP - Performance leader for AI speech and audio Processing with up to 4X the NN performance and 2X the audio DSP performance of the HiFi 4 DSP. Ideal for digital assistants, infotainment, and other voice-controlled products.

Features for a Wide Range of Audio, Voice, and Speech Processing

The Features You Need for Designs Ranging from Ultra-Low Power to High Performance

Audio DSP Plus Control

All Tensilica HiFi DSPs are click-box additions to our customizable Xtensa® LX processor and include the base Xtensa instruction set. In other words, they are excellent targets for control applications as well as audio.

Customizable to Meet Your Requirements

Because our HiFi DSPs use our Xtensa LX processors, you get all of the customization advantages that our processors are known for, along with incredible flexibility, including (but not limited to):

  • Complete flexibility to add/configure caches and local memories
  • Option to add direct RTL-like interfaces via Ports, Queues, and Lookup Interfaces
  • Option to add custom instructions with full compiler support

See our Xtensa product section for more details.

 

Optimized for Audio and Voice

Our HiFi DSPs are optimized solutions for digital audio, voice and speech. The performance levels of these DSPs approach those of fixed-gate implementations. They offer the flexibility, fast time-to-market, and lower risk of a processor-based solution, while simultaneously reducing silicon area and power consumption.

Our HiFi DSPs utilize Xtensa LX processor technology with custom instructions (better code density, fewer cycles), flexible datapath widths (higher efficiency per cycles), multiple instructions per cycle (VLIW—higher efficiency per cycle), and simultaneous data operations (SIMD—parallelization). A comprehensive Eclipse-based software development environment is provided, with everything from a source code editor, debugger, and ISS to an optimized compiler for exceptional code density.

Optimized for Always-Listening Voice Trigger and Recognition

The HiFi Mini DSP is our smallest, lowest power DSP supporting always-listening voice trigger and speech command modes. The HiFi Mini DSP enables a hands-free experience. Cadence is working with Sensory and other software partners who provide innovative voice activation, speech command recognition, voice pre-processing and noise reduction products—all optimized on the HiFi Mini DSP.

We optimized the HiFi Mini DSP to be as small and power-efficient as possible for voice trigger and voice recognition in two major ways. First, the DSP uses compact 40-bit encoding, which significantly improves code size. Second, we added efficient 16-bit instructions that are optimized for voice and audio codecs.

Efficient for Digital Assistants

The HiFi 3z, HiFi 4, and HiFi 5 DSPs efficiently support front-end audio processing like beamforming and noise reduction, as well as neural network-assisted speech recognition (ASR) processing.

 

All the Digital Radio Codecs You Need

Our HiFi DSPs run all the decoders required around the world for digital audio, including:

  • HD Radio™—From iBiquity™ Digital
  • DAB—In the MP3 decoder
  • DAB+—Certified by Dolby®
  • DRM—Certified by Dolby
  • XM-Satellite—From XM® Satellite
  • T-DMB—In the BSAC decoder
  • ISDB—In the AAC, HE-AAC decoders

Fully Programmable in C

All HiFi DSPs are fully programmable in C, with no assembly required. So if you have proprietary algorithms to port for a HiFi DSP, it’s easy and quick.

Ready for Immediate Design-In

All HiFi DSPs are pre-packaged, ready to drop into SoC designs. They provide programmable support for over 225 audio, voice, and sound enhancement packages, which are pre-ported and ready to use, with no software porting required.

Capable of the Best Sound Enhancement Algorithms

Audio requirements for everything from smartphones to home stereo receivers are being driven by audio post-processing.

For voice, smartphone requirements are moving from narrowband to wideband codecs with improved noise suppression, noise dependent volume control, and other algorithms to improve sound quality in crowded environments.

For home and car audio, audio enhancement software is significantly enriching the user experience, creating important competitive advantages.

And gaming requirements in all environments are driving requirements for up to 32 streams for immersive play.

Efficient for Home Audio Including Set-Top Boxes, Sound Bars, and TVs

Whether we are running the latest object-based audio codec or the latest and greatest audio post-processing algorithms, the HiFi DSP family is the right choice.

The need for DSP horsepower in the home audio space is expected to increase as more and more devices begin to directly integrate voice assistant services.

The HiFi 3, HiFi 4, and HiFi 5 DSPs offer a scalable set of choices to build any home audio SoC.

Performance to Spare

Because of the efficiency of the Tensilica HiFi DSP architectures, only a fraction of the available MHz to perform audio decoding and encoding functions is required, leaving extra headroom for other control as well as media and signal-processing tasks.

Tensilica HiFi DSP Family Specification

What are the differences among the HiFi DSPs?

HiFi24 ISA

HiFi32 ISA

ISA Subset Variant

HiFi Mini Fusion F1 HiFi 3 HiFi 3z HiFi 4 HiFi 5
v24 V32(c)(f) v32(vf) v32(vf) v32(vf) v32(vf)
VLIW Slots 2 2 3 3 4 5
VLIW Bundle Instructions Sizes 40-bit 40/48-bit 64-bit 40/48/64-bit 48/88-bit 64/128-bit
Fixed Point MACs per Cycle
32 x 32 n/a 1 2 2 4 8
24 x 24 2 2 4 4 4 8
32 x 16 2 2 4 4 up to 8 16
16 x 16 2 2 or 4 4 up to 8 up to 8 16
Accumulator 56-bit 64-bit 64-bit 64-bit 64/72-bit 64/72-bit
FPU (optional) Scalar (independent) FPU Integrated
2-way SIMD VFPU
Integrated
2-way SIMD VFPU
Two integrated  2-way SIMD VFPU Two integrated  4-way SIMD VFPU
ITU Intrinsic Support No Yes
Circular Buffer Support None 1 1 1 2 3
Bitstream VLE/VLD Support

Yes

Optional

Yes

Yes

Yes

Yes

User-Defined Instructions Limited Yes Yes Yes Yes Yes
Other Notes Includes 32x24 precision MAC Optional connectivity protocol acceleration 32 MACs/cycle NN MACs (optional) half-precision FPU

 

ISA Variant Definitions

v24 - Base HiFi instruction set, 24-bit and 16-bit data types
v24c -Enhanced precision 32x24 support, circular buffer
v32 - Superset of v24 with additional operations on 32-bit types
v32(f) - Single-precision integrated IEEE floating point
v32(vf) - Single-precision integrated IEEE vector floating point
v32(c) - Communications protocol acceleration (AES, convolutional encoding....)

HiFi Mini DSP for Always-On Voice Recognition and Sensor Fusion

Need Always-On Functionality? HiFi Mini DSP Is Great for Voice or Face Recognition and Sensor Fusion

Low-power always-on wake-up is now a must have feature for mobile applications. Voice trigger functions have been widely adopted. Face trigger is becoming more popular. And sensor fusion also requires low-power always-on DSP processing. The HiFi Mini DSP is ideal for all these applications, so you can use one core for all.

Always-On Wake Capabilities

Low-power always-on wake capabilities will continue to evolve. Right now, this is commonly used for voice trigger, face trigger, and ultrasonic motion detection. For these evolving applications, you need both ease of development and performance headroom. You need the HiFi Mini DSP.

The HiFi Mini DSP's low-power audio buffer sharply reduces MIPS requirements, up to 8X lower than other solutions. The RAM includes phrase memory and data buffers (1.6-1.8x).  ROM includes recognizer code and constants for acoustic model and vocabulary.

Sensor Fusion

Sensor fusion combines control and signal processing. The HiFi Mini DSP can easily support both, as it's based on the Xtensa 32-bit RISC CPU. It is efficient in both control and DSP-intensive tasks and features smart power management. Cadence works with several software partners to bring the best software for sensor fusion to the HiFi Mini DSP. 

Watch HiFi Mini DSP Demonstration from RealTek


Watch Ty Kingsmore from Realtek Semiconductor as he demonstrates his new products at the Consumer Electronics Show (CES) 2015 in Las Vegas. This uses the HiFi Mini DSP for always-on, always-listening capabilities, significantly lowering power by offloading the host processor.

HiFi 4 DSP for Ultra-High-Performance Audio

Designing for the Future? HiFi 4 DSP Offers the Highest Performance for Next-Generation Demands

Whether designing for the next generation of audio standards or for a voice digital assistant, the HiFi 4 DSP provides the right level of high-performance audio digital signal processing power. This allows digital TV and set-top box makers to easily support the new object-based audio codecs that are replacing the previous generation of channel-based codecs. Additionally, OEMs creating voice-controlled digital assistants are embracing the HiFi 4 architecture for its audio preprocessing capabilities as well as its efficiency at running neural network-based keyword-detection and voice-recognition software.

Optimized for the Most Demanding DSP Algorithms

Our HiFi 4 DSP is a highly optimized audio/voice processor geared for efficient execution of audio and voice codecs and pre- and post-processing modules as well as other demanding DSP functions in areas such as wireless communications. Its key features include:

  • Support for four 32x32-bit multiplier-accumulators (MACs) per cycle with 72-bit accumulators, more than double the performance of other audio DSPs for computationally intensive functions such as fast Fourier transform (FFT) and finite impulse response (FIR)
  • Support for eight 32x16-bit MACs per cycle under specified conditions
  • Four very long instruction word (VLIW) slot architecture capable of issuing two 64-bit loads per cycle
  • Optional vector floating-point unit available, providing up to four single-precision IEEE floating-point MACs per cycle
  • Software compatible with the existing HiFi DSP family consisting of over 225 HiFi-optimized audio and voice codecs and audio enhancement software packages.

HiFi 5 DSP for AI Speech and Audio Processing

The Age of the Voice UI

Voice UI capabilities are becoming more and more important to OEMs. In the home, numerous appliances from smart speakers to set top boxes to microwave ovens all have built-in voice assistant capabilities. The trend is even continuing to the automobile with new natural language processing (NLP) and voice assistant services being adopted by leading brands.

Voice UI processing typically falls into two categories: audio pre-processing and speech recognition/keyword detection.

Key features of the HiFi 5 DSP include: 

  • Five very long instruction word (VLIW)-slot architecture capable of issuing two 128-bit loads per cycle 
  • 2X MAC capability versus HiFi 4 DSP for pre- and post-processing, includes:
    • Support for eight 32x32-bit or 16 16x16-bit MACs per cycle 
    • Optional eight single-precision floating-point MACs per cycle
  • 4X MAC capability versus HiFi 4 DSP for NN processing, includes:
    • 32 16x8 or 16x4 MACs per cycle
    • Optional 16 half-precision floating-point MACs per cycle
  • New HiFi NN library offers a highly optimized set of library functions commonly used in NN processing (especially speech), which can easily be integrated into popular machine learning frameworks
  • Software compatibility with the complete HiFi product line, totaling over 300 HiFi-optimized audio and voice codecs and audio enhancement software packages

Audio, Voice, and Speech Software—Over 300 Packages Available Now

List of 300+ Partner Software Packages Already Supported on HiFi DSPs

The following is a list of audio and voice enhancement software packages available for Tensilica HiFi DSPs.

Audio Codecs

  • AEC Wideband
  • AMR Wideband+ Decoder
  • AMR Wideband+ Encoder
  • DAB/MP2 Decoder
  • DAB+ Decoder
  • Digital Radio Mondiale (DRM) Decoder
  • Dolby AC-4 Decoder
  • Dolby Digital AC-3 Decoder, 5.1 ch
  • Dolby Digital AC-3 Consumer Encoder (DDCE) 2, 5.1 ch
  • Dolby Digital Compatible Output Encoder 5.1 ch
  • Dolby Digital Live Encoder
  • Dolby Digital Plus Consumer Decoder 7.1 ch
  • Dolby Digital Plus Decoder/Converter 5.1 ch
  • Dolby Mobile 3+
  • Dolby MS10 Multistream Decoder
  • Dolby MS11 Multistream Decoder
  • Dolby MS12 Multistream Decoder
  • Dolby ProLogic II and IIx Decoder
  • Dolby TrueHD Decoder 7.1 ch
  • Dolby Volume 2, 5.1 ch
  • DRA Decoder, 7.1 ch
  • DTS Boost
  • DTS Broadcast and DMP 5.1 ch
  • DTS Core Decoder 5.1 ch
  • DTS Envelo Headphone and Speaker
  • DTS-HD (Master Audio), Express, Transcoder, and Neo:6
  • DTS-HD PP
  • DTS Interactive Encoder
  • DTS M6 Decoder, 5.1 ch
  • DTS Neural Surround
  • DTS Symmetry
  • FLAC Decoder, 7.1 ch
  • MP3 Decoder
  • MP3 Encoder
  • MPEG-2 Layer 2 Decoder, 5.1 ch
  • MPEG-4 aacPlus v2 Decoder 2, 7.1 ch
  • MPEG-4 aacPlus v2 Encoder
  • MPEG-4 aacPlus v1 Decoder 2, 7.1 ch
  • MPEG-4 aacPlus v1 Encoder
  • MPEG-2/4 AAC LC Decoder 2, 7.1 ch
  • MPEG-4 AAC LC Encoder
  • MPEG-4 BSAC Decoder
  • MPEG-H Audio Decoder
  • mSBC Bluetooth Decoder
  • mSBC Bluetooth Encoder
  • mSBC Decoder
  • mSBC Encoder
  • Ogg Vorbis Decoder
  • RealAudio 8, 9, and 10 Decoder
  • Sample Rate Converter
  • SBC Bluetooth Decoder
  • SBC Bluetooth Encoder
  • WMA Decoder
  • WMA Encoder
  • WMA 10 Pro Decoder 2, 7.1 ch
  • WMA 10 Pro Lossless Decoder 2, 7.1 ch

Voice Codecs

  • AAC-ELD
  • AMR Narrowband Speech Codec
  • AMR Wideband Speech Codec
  • Bluetooth CVSD Decoder
  • Bluetooth CVSD Encoder
  • G711 Speech Codec
  • G722 Speech Codec
  • G723.1 Speech Codec
  • G726 Speech Codec
  • G.729AB Speech Codec
  • GSM-EFR Speech Codec
  • GSM-FR Speech Codec
  • GSM-HR Speech Codec
  • Opus
  • SILK Speech Codec

Audio Voice, Sensor Fusion, and IoT Partners

Please see our listing of the specialty of each partner here.

Partner Software Offerings for Audio, Voice, Sensory Fusion and IoT

The Strongest Partner Ecosystem for Audio, Voice, Sensor Fusion and IoT

Cadence has the largest partner ecosystem for audio, voice, sensor fusion and IoT. Here's a list of our partners, by specialty.

Partner Application Home Entertainment Mobile/Tablet/PC Automotive IoT Engineering Services
Accusonus Speech enhancement check mark
AM3D Audio post-processing check mark check mark check mark
Arkamys Audio post-processing, active noise control check mark check mark check mark
Audyssey Audio post-processing check mark check mark check mark
Cogent Embedded Services check mark
Conexant Voice pre-processing check mark check mark check mark
Voice recognition check mark check mark check mark
Sensor fusion check mark
dbx-tv Audio post-processing check mark
Digital Wave Audio codec check mark
Dolby Audio codecs, post-processing check mark check mark
Embedded audio software development
Services
DTS Audio codecs, post-processing check mark check mark check mark
DVSI Voice codecs check mark
eInfochips Services check mark
Forte Media Voice pre-processing check mark
Fraunhofer Codecs check mark check mark check mark
Audio post-processing check mark check mark check mark
Hillcrest Labs  Sensor fusion, management check mark check mark
iBiquity Digital radio codec check mark check mark
Inband Software Services check mark
Services check mark
Ittiam Services check mark
Kronoton Audio post-processing check mark
Malaspina Labs Voice pre-processing check mark check mark
Maxim Integrated Audio post-processing check mark
Microsoft Codecs check mark check mark
MightyWorks Logo Audio, voice pre-processing check mark check mark
Movea Sensor fusion, management check mark check mark
Active noise control check mark
Nuance Voice activation, recognition check mark check mark
NXP Software Voice pre-processing check mark
Opalum Audio post-processing check mark check mark
Services check mark
PProduct Services check mark
QNX ANC, voice pre-processing check mark
QSound Voice pre-processing, audio post-processing check mark
Real Codecs check mark check mark
Retune Voice pre-processing check mark check mark check mark check mark
Rubidium Voice activation, recognition check mark check mark
Sensor fusion, management check mark check mark
Sensory Voice activation, recognition check mark check mark
Skype Voice codec check mark check mark
Sonic Emotion Audio post-processing check mark
Spirit DSP Voice codecs check mark
SPL Audio post-processing check mark
Tata Services check mark
Technicolor MP3 check mark check mark
Audio post-processing, voice pre-processing check mark check mark

Design Tools Speed Your Adoption and Integration Processes

The Industry's Most Powerful and Complete Design Environment

Port C to HiFi

Port Software Quickly in C

A key advantage of the HiFi DSPs is that they use a simple programming model. Software developers can port audio, voice, and enhancement software codecs completely in C, while maintaining or surpassing the performance of assembly on other DSP and CPU architectures. This minimizes the time required to port and maintain special or proprietary audio software.

 

Processor design process

For Processor Designers

We deliver patented, proven tools that automate the process of generating your HiFi DSP along with matching software tools. These tools have been proven in hundreds of designs and the results are guaranteed correct by construction.

View the complete set of tools for processor designers.

Softwre development process

For Software Developers

The Xtensa® Software Developer's Toolkit provides a comprehensive collection of code generation and analysis tools that speed the software development process. The Cadence Tensilica Eclipse-based Xtensa Xplorer Integrated Development Environment (IDE) serves as the cockpit for the entire development experience.

View the complete set of tools for software developers.

Articles

EEJournal: Watching AI Evolve. A Common Trajectory?

SemiWiki: Why Did Ambiq Micro Select HiFi-5 DSP IP for Next Generation MCU?

EETimes: New Architectures Bringing AI to the Edge

SemiEngineering: AI Begins To Shape Chip Design

CDRinfo: New Cadence Tensilica HiFi 5 DSP Is Optimized for AI Speech and Audio Processing

Defining Electronics: Cadence Introduces the Tensilica HiFi 5 DSP, the First DSP Optimized for AI Speech and Audio Processing

Electronics Weekly: Neural network IP aimed at SoCs processing complex speech

Tech Design Forum: Cadence adds deep-learning support to audio DSP

eeNews Embedded: Cadence Tensilica HiFi 5 DSP for audio and voice processing

Radio-Electronics: First DSP optimized for AI speech and audio processing

Softei: HiFi DSP boosts audio and neural network processing

ETN Finland: Your next device controls your voice

Electroniques France: Cadence doubles the power of the DSP Tensilica HiFi

Electronique Mag France: Cadence Announces Tensilica HiFi 5

Electronic Specifier France: DSP with optimization for AI audio and audio processing

Electronic Specifier Germany: DSP with optimization for AI voice and audio processing

Elettronica In Italy: Cadence presents the Tensilica HiFi 5 DSP, the first solution based on Artificial Intelligence

Elektronik Tidningen Sweden: Optimized for speech recognition and sound

EDA Express Japan: Cadence Announces Tensilica HiFi 5 DSP for Neural Network Audio Applications

Nikkei Japan: Cadence announces DSP Tensilica HiFi 5 DSP for AI speech and audio processing