Dubbed WM5110, the chip has a total of 600Mips of digital audio processing capacity, and audio DACs with a signal to noise ratio (SNR) of 110dB.
An earlier product, the WM5100, achieved 75Mips and 102dB.
“That 8dB improvement has taken us almost a year to achieve,” said product line manager Duncan Macadie.
Power for headphone audio playback including DACs and amplifier is 3mA, compared with 4mA with the 5100.
The DSPs can be clocked (20MHz to 150MHz) and shut down separately to minimise power consumption for a particular task, and at full throttle “would be running a lot of audio enhancement algorithms and consuming several tens of milliwatts”, said Macadie.
By providing so much DSP, the firm is pitching at phone designers wanting to off-load audio processing from the main application processor.
“Voice control takes around 35mW on an application processor, we can do it for under 10mW,” claimed Wolfson product director Eddie Sinnott.
The other argument that the firm puts forward is that analogue functions become increasingly difficult as chip geometries are squeezed: so DACs and ADC built on 40, 28 or 20nm application processors are not going to perform well or produce good audio.
Wolfson will not say what geometry the new chip is on, except to say that it is “several process generations down from the 5100”.
On the 5110 are six ADCs that can handle six analogue microphones or up to eight digital microphones – three microphones are used in high-end phones for beam-forming to cut noise, and “four looks probable,” said Macadie.
There is a software development kit, plus Wolfson’s own suite of sound enhancing and noise reduction software, and the kit can also handle customer and third-party software.
Possible algorithms include: transmit-path and receive-path noise reduction (claimed to produce 32dB and 20dB noise reduction respectively), earpiece and stereo headphone ambient noise cancellation, acoustic echo cancellation, and wind noise suppression.
The chip is sampling now in a W-CSP package.