Overview

The CEVA-TeakLite-4 is the most advanced member of the CEVA-TeakLite DSP family, designed for efficient audio signal processing in high-volume, cost-sensitive applications. Founded on an efficient memory-based architecture, the CEVA-TeakLite family combines small die size and high code density with high levels of processing power.

Based on a native 32-bit architecture, different members of the CEVA-TeakLite family can perform one, two, or four 32×32-bit Multiply-Accumulate (MAC) operations, as well as two or four 16×16-bit MAC operations in a single cycle. These scalable options enable a precise balance to optimally address different target applications, like baseband audio and voice, media codecs, voice-activated interfaces and more.

The CEVA-TeakLite-4 architecture supports both an advanced set of digital signal processing instructions as well as general-purpose microprocessor instructions. The instruction set and programming model are designed for straightforward generation of compact and efficient code and the integrated second-generation Power Scaling Unit (PSU 2.0) minimizes power consumption via advanced power management (including support for clock and voltage scaling).

All members of the CEVA-TeakLite family are compatible with each other and are backward compatible with their predecessors, including the widely adopted CEVA-TeakLite, CEVA-TeakLite-II, and CEVA-TeakLite-III DSP cores. This enables new audio DSP chip designs to leverage existing applications and a large installed base of software and simplifies the migration of existing designs to a higher performance DSP core.

Benefits

The integrated, high-performance, ultra-low-power CEVA-TeakLite-4 family enables rapid design of DSP-based solutions for audio, voice, and speech applications.

Fully featured and scalable family of DSPs with four compatible DSP cores
Advanced fourth-generation architecture with straightforward ISA and programming model
A large and rich set of compatible software modules

Main Features

  • Enhanced bit-manipulation capabilities for efficient stream processing
  • Up to 5 operations executed in parallel
  • Dedicated single-precision and double-precision FFT instructions
  • Up to 4 GW program memory and 4 GW data memory (16-bit words)
  • L1 program memory (TCM or 4-way cache)
  • L1 data memory (TCM or 2-way, set-associative, hardware-configurable cache)
  • Optional floating-point unit

Block Diagram

TeakLite4 Block Diagram

The chip that enables "OK, Google"

Talking with CEVA at MWC 2017

Gary Sims stopped by the Ceva stand at Mobile World Congress 2017 to talk about voice, DSPs and audio over Bluetooth 5.