Modern technology gadgets such as wireless speakers should be inexpensive, small and energy-efficient, to ensure the longest possible battery operation. At the same time, a satisfied user demands excellent sound quality, which, depending on the situation, can make a variety of demands on the digital processing of the audio signal. Since electronic components available on the market can not meet all these requirements at the same time, specific components must be developed for a successful product. No new processors are invented, but existing processor cores are extended by application-specific functions.
For a digital loudspeaker processor, a complete System-on-Chip (SoC) design was developed for which an SPARCv8 open-source processor has been extended with an arithmetic coprocessor that allows the efficient implementation of audio algorithms. The design goals were to minimize resource consumption and the throughput of one multiply-accumulate (MAC) instruction per cycle.
To further reduce the execution time of a typical audio filtering application, the developed architecture allows direct access to the on-chip memory. Since time-consuming memory register instructions (load and store) are eliminated, the execution time can be shortened by a factor of 3 compared to a standard register-register processor.
Peripheral interfaces such as I2S make the system a complete hardware platform for digital audio processing in modern embedded devices.