CEVA-TeakLite-4: A Multifunctional DSP Architecture for High-Performance, Low-Power Audio/Voice/Sensing and Wireless Communication Applications
The CEVA-TeakLite Family of DSP cores is designed to address the needs of high volume, cost-sensitive markets. Founded on a classic memory-based architecture, the CEVA-TeakLite family combines small die size, high code density, and high processing power.
The fourth generation of the CEVA-TeakLite family, the CEVA-TeakLite-4 is a low-power, native 32-bit, variable 10-stage pipeline, fixed-point DSP architecture framework. The CEVA-TeakLite-4 is a fully synthesizable, process-independent design that allows the SoC designer to select the optimal implementation in terms of silicon area, power consumption, and operating frequency.
CEVA-TeakLite-4 V2 architecture further enhances the performance and efficiency of the CEVA-TeakLite-4 family by delivering 30% code size reduction, 20% power consumption reduction and new instructions and mechanisms for audio/voice/wireless standards implementation.
CEVA-TeakLite-4 Target Applications
Thanks to its scalability, the CEVA-TeakLite-4 DSP family is efficiently handling most demanding audio, voice, sensing and wireless connectivity use-cases, whether targeting ultra-low-power always-on functions, advanced multi-microphone voice processing, or high-performance multichannel audio processing and post-processing. CEVA-TeakLite-4 cores are adopted by mobile, home, and automotive chip vendors, ranging from the smallest, lowest-power audio CODEC and voice activation chips, to Baseband, Application Processors and connectivity chips, to high-end audio such as wireless speakers, digital televisions (DTVs), set-top boxes (STBs), game consoles, and more.
Always-on User Interface (UI)
Hands-free activation is becoming a 'must have' feature in a multitude of devices, primarily smartphones and wearables, and solutions have to operate efficiently and reliably to meet demanding consumer requirements. Be it voice control, gesture control or face unlock, a natural user interface (UI) is expected to be always-on in order to be totally hands-free and therefore must consume very little power. The CEVA-TeakLite-4 offers multi-functional DSP capabilities, making it ideal for implementing always-on solutions that often include intensive signal processing algorithms such as those used for voice recognition and face detection. Furthermore, latest generation power scaling technology of the CEVA-TeakLite-4 V2 enables it to run always-on UI with ultra-low power consumption by shutting down unneeded hardware on a cycle-by-cycle basis. Being CEVA’s smallest DSP core, it also benefits from very low leakage power, which is especially important when implementing always-on functionality, such as that of a voice trigger chip for example.
Sensing is a key feature of smartphones, tablets and IoT devices. Combined processing of sensory data from disparate sources, also known as Sensor Fusion, has proven to provide valuable contextual awareness for a host of user applications. Multiple MEMS sensors such as 9-Axis motion sensors, barometer, thermometer and more call for a high-precision sensor fusion DSP to accurately deduct motion and environmental data, and eliminate false readings. The CEVA-TeakLite-4 boasts a highly optimized arithmetic ISA as well as a rich control ISA that are efficiently utilized by the compiler as it handles Sensor Fusion code, which often mixes arithmetic and control.
Audio and Voice
Mobile devices like smartphones and tablets are required to handle increasing audio and voice processing requirements, such as multi-microphone beam-forming and noise reduction, wideband voice processing like VoLTE, and long playback time audio, Similarly, automotive infotainments and home-entertainment systems like digital televisions (DTVs), Smart TVs, and Set-Top Boxes (STBs) must support multiple audio and voice standards, advanced audio post-processing, and far-talk voice communication and command. With its dedicated audio and voice ISA, a huge library of certified audio/voice codecs, functions and partner offerings, the CEVA-TeakLite-4 is the most popular audio DSP family.
Combined with the CEVA-Bluetooth hardware and software, the CEVA-TeakLite-4 enables a wide range of wireless applications without requiring an additional CPU. A Smartwatch solution, for example, can use a single Bluetooth-enabled CEVA-TeakLite-4 to run always-on voice activation and voice commands, sensor fusion functionality, audio/voice processing and dual mode Low Energy Bluetooth (also known as Bluetooth Smart Ready) capable of both BLE such as that of Bluetooth Smart, and wireless audio. Given the highly integrated nature of IoT solutions, such a capacity cannot be overestimated. On top of Bluetooth, CEVAnet Partner Program members offer additional wireless technologies for the CEVA-TeakLite-4 that can either be used separately or in a combo solution. For example, a wireless speaker solution combining both a WiFi speaker and a Bluetooth speaker can have both Bluetooth and WiFi audio using a single Bluetooth-enabled CEVA-TeakLite-4 integrated with a CEVAnet WiFi solution.
CEVA-TeakLite 4 Architecture-compliant Cores
The CEVA-TL410, CEVA-TL411, CEVA-TL420, and CEVA-TL421 DSP cores are based on, and compliant to, the high-performance, low-power CEVA-TeakLite-4 DSP architecture.
With a primary target of standalone audio DSP chips used to implement audio CODECs, voice activation, and noise-reduction chips, the ultra-low-power CEVA-TL410 audio DSP core offers the smallest die size with its single 32x32-bit MAC, dual 16x16-bit MACs, and direct memory interface. If higher performance is required, the CEVA-TL411 audio DSP core provides dual 32x32-bit MACs and quad 16x16-bit MACs.
Alternatively, for CPU-centric SoCs, such as the application processors and main SoCs used in smartphones, digital televisions (DTVs), set-top boxes (STBs), and game consoles, the CEVA-TL420 audio DSP augments the features of the CEVA-TL410 with data and instruction cache controllers and a master/slave AXI system interface (the CEVA-TL421 augments the CEVA-TL411 with the same high-end capabilities). All CEVA-TeakLite-4 architecture-compliant cores are fully compatible with each other and with previous generation CEVA-TeakLite family cores.
An audio ISA (instruction set architecture) providing dedicated audio instructions is present in all CEVA-TeakLite-4-based cores. Also, all members of the family include an integrated, second-generation Power Scaling Unit (PSU 2.0) for smart power management.
The CEVA-TeakLite-4 architecture is scalable to support stand-alone audio/voice tasks such as filters, voice pre-processing, audio post-processing, and noise reduction; also for mobile applications such as off-loading the main CPU by performing multi-channel audio decode, transcoding, voice pre-processing, and audio post-processing under the Android operating systems, using the Android Multimedia Framework. The CEVA-TeakLite-4 architecture framework also supports user differentiation to allow expansion for handling proprietary algorithm acceleration and future use-cases.
Smart and Connected development platform
The CEVA 'smart & connected' development platform is the ultimate tool for algorithm and system designers who wish to develop and prototype TeakLite-4 based designs using real time, full speed silicon. Built around a TeakLite-4 DSP chip that also incorporates an integrated set of peripherals and fabricated at the IoT-centric SMIC 55nmLP process, the TeakLite-4 silicon is running at 500MHz. This provides ample processing power for multiple functions running concurrently using the included DSP RTOS, and enables prototyping of real life use-cases. Such use-cases may combine always-on sensing, local intelligence, and wireless connectivity, all supported by the smart and connected platform. With low power signal processing being a key element in smart IoT device differentiation, the platform also provides real-time power measurement and display that allow developers to optimize and power-tune their DSP firmware.
The CEVA smart and connected development platform includes the following elements:
- DSP development chip: 500MHz CEVA-TeakLite-4 and subsystem of integrated peripherals (TDM, DMA, I2C, I2S, ICU, Timers, GPIOs)
- Host CPU: Linux running on-board ARM Cortex-A9 for complete CPU + DSP system prototyping
- Connectivity: multiple wireless technologies available from CEVA and its partners, including Bluetooth, Wi-Fi, ZigBee and GNSS
- Peripherals and interfaces: including on-board digital (MEMS) microphones, I2C I/F for sensors, audio CODEC with digital and analog audio in/outs, USB, UART, PCIe, and Ethernet ports, user configurable FPGA, GPIOs, DDR memory, SD card, and LCD display
- Arduino connectors and drivers: allows Arduino Shields to be connected to the board, leveraging a massive ecosystem of Arduino-related products
- RTOS and DSP libraries: multi-tasking RTOS and a DSP library to jumpstart system design
CEVA-TeakLite 4 Deliverables
The CEVA-TL410, CEVA-TL411, CEVA-TL420, and CEVA-TL421 DSP cores are supported by a wide range of deliverables, which significantly reduce risk and time-to-market. These deliverables include a complete implementation along with associated hardware and software development tools and verification and simulation environments. CEVA-TeakLite-4-based designs can also be implemented in an FPGA for prototyping and system integration.
The CEVA-TeakLite-4 is also backed up by a wealth of software and algorithms. To further reduce the cost, complexity, and risk in bringing products to market, CEVA has established an ecosystem of partners who provide application software, reference designs, complementary IP, design services, and complete solutions based on CEVA’s DSP cores and Platforms and Solutions. Visit our CEVAnet Partners page for more information.
|Small size and ultra-low-power
||CEVA-TeakLite-4 DSP cores can be used for applications that are highly sensitive to die-area and power consumption|
||CEVA-TeakLite-4 DSP cores support the toughest audio and voice use cases|
|Native 32-bit, Harvard/SIMD architecture DSP with multiple options:
||Supported by a broad range of fully-certified HD-Audio and voice codecs|
|Easy software development
||Smooth C-level software development and easy integration into target SoC reduces risk and time-to-market|
|...and many more – Download the CEVA-TeakLite-4 Product Brief|
Based on a native 32-bit architecture, different members of the CEVA-TeakLite-4 family can perform one, two, or four 32x32-bit Multiply-Accumulate (MAC) operations and two or four 16x16-bit MAC operations in a single cycle. The CEVA-TeakLite-4 also offers:
- Enhanced bit-manipulation capabilities for stream processing
- Up to 5 operations executed in parallel
- Dedicated single-precision and double-precision FFT instructions
- Up to 4GW program memory and 4GW data memory (16-bit words)
- L1 program memory (TCM or cache)
- L1 data memory (TCM or 2-way, set-associative, hardware-configurable cache)
- For more information, please contact firstname.lastname@example.org
The CEVA-TeakLite-4 architecture supports an advanced set of digital signal processing instructions as well as general-purpose microprocessor instructions. The CEVA-TeakLite-4's instruction set and programming model are designed for the straightforward generation of compact and efficient code.
The integrated second generation Power Scaling Unit (PSU 2.0) provides advanced power management including support for clock and voltage scaling.
All members of the CEVA-TeakLite-4 family are compatible with each other and are backward compatible with their predecessors, including the widely adopted CEVA-TeakLite, CEVA-TeakLite-II, CEVA-TeakLite-III, and CEVA-Teak DSP cores. This allows new designs to leverage existing applications and a large installed base of software; it also makes it easy for existing designs to be migrated to a higher performance core.
Codecs available directly from CEVA include:
|Vocoders||Audio decoders||Audio encoders||Post-processing|
|G.723||MP3||MP3||Dolby Mobile 3+|
|G.729.1||MPEG4 AAC-LC||MPEG4 AAC-LC||Dolby ProLogic IIx|
|G.711||HE-AAC V1||HE-AAC V1||Dolby Volume|
|G.722||HE AAC V2 7.1||SBC||DTS Extended Surround (ES)|
|AMR-NB||WMA10||Dolby Digital encoder (DDCE)||DTS Neo:6|
|Dolby Digital Plus|
|Dolby Digital (AC3)|
|DTS Master Audio|
|DTS High Resolution|
|DTS Digital Surround|
Many other codecs and post-processing functions are available from our CEVAnet partners.
TeakLite Announcements21 Feb 2017 Waves Audio and CEVA Partner for Far-Field Voice Pick Up and Psychoacoustic Sound Enhancement
24 Feb 2016 CEVA Powers DSP Group's Next-Generation Ultra-Low-Power Always-On Voice Processor
23 Feb 2016 CEVA and Spreadtrum Expand Long-Term Partnership for LTE SoCs Targeting Mid and High End Smartphones
- Products Overview
- DSP Cores
- Connectivity Platforms
- Development Environment