NXP adds DSP to turn up volume in smartphones
NXP is using a DSP and measurement of both output current and voltage to push 2.6Wrms through phone speakers.
The product, called TFA9887, is a single chip intended to replace standard audio amplifier ICs in phones.
Loudspeakers inside phones are already marvels of acoustic engineering, producing clear sound from remarkably small packages.
Two things limit the amount of power than can be driven into phone speakers: over-heating and over-excursion.
“These microspeakers are typically rated at 0.5W max,” Shawn Scarlett, director of mobile audio at NXP, told Electronics Weekly. “After that, heat melts the glue and resonant excursions crack and tear the speaker membrane.”
According to Scarlett, these issues are usually addressed by simple amplitude limits and filtering, or basic open loop techniques.
“Essentially, in development, phone makers turn the audio up till they have reliability problems, and then turn it back a bit,” he said. And to avoid resonance “right now, all phones on market have high-pass filter at around 1kHz”.
The open-loop techniques, said Scarlett, include calculating voice coil temperature from knowledge of its impedance, derived from sensing output current and voltage, and knowledge of its temperature/impedance characteristic.
This allows power to be limited when the coil gets hot.
Through knowledge of the speaker’s resonant frequency drift with temperature, it further allows a notch filter to be placed over the inferred resonant frequency – so bass music can be reproduced with the notch keeping the speaker safe from over-excursion.
However, this way of applying a notch requires that it be over-wide – wide enough to allow for initial manufacturing tolerance in the speaker, variations in thermal drift coefficient, and aging.
“Some guys focus on temperature to predict resonance shift, but if dirt clogs the speaker port, resonance changes,” said Scarlett. “Putting the phone in a poor case which blocks the speaker port can change resonance to around 2kHz.”
After three years of work, NXP claims that, from the output voltage and current, its DSP can remove all guess-work by calculating resonant frequency live, as well as carving an appropriate notch out of the incoming audio spectrum.
“A current-sampling stage built into amplifier allows us to really know what is going on in the speaker,” said Scarlett. “We look at impedance over the whole frequency range and see the resonance. We can use a notch with much tighter margins because we are actually measuring it.”
Because distance/volt force factor does not vary significantly between speakers, he said, speaker diaphragm movement can be estimated to calculate how much can be let through the notch.
“We are running a co-processor, compressing on speaker excursion not voltage, so we can look at when speaker becomes non-linear,” said Scarlett. “With us, you can add 6dB compared with state-of-the-art phone today and get twice the perceived volume. And there are no low frequency problems – you can add 10dB at 400Hz.”
So, what practical difference does that make?
“If you play Queen’s Another One Bites the Dust, the bass guitar is there, which is isn’t normally on a phone,” he said.
NXP claims its chip will work with any phone speaker, providing four speaker-related parameters including the thermal coefficient of the coil are saved to the chip.
As with many phone amplifiers, to save power the output stage’s voltage rail can be changed depending on the output amplitude required.
In this case, the rail is either direct battery volts, or 5V generated by an on-chip dc-dc converter – which is turned off when it is not needed.
“An advantage of integrating a DSP is that, if battery voltage starts to drop, it can calculate when clipping is going to happen and turn on the dc-dc converter.
With control over the dc-dc converter and knowledge of the speaker’s excursion, different operating modes are possible.
“We can go for maximum volume for the ring tone, then load a profile to extend bass as much as possible to watch movies,” said Scarlett. “We give the phone designer a set of tools to optimise for sound quality: for example, to set how much clipping they will allow.”
There are actually six parameters fed back into the local DSP.
“It is not looking at just I/V resonance. We sense six points which really allows us to look at sound pressure level and sound quality,” said Scarlett. “If you try to do that on a separate DSP, current and voltage would have to be provided full bandwidth real-time. If you tried to pass it all to the application processor, you would use up every bus you have.”
Audio is delivered to it over an I2S bus and control including volume and mute over I2C.
Digital quiescent current is 5mA with the DSP off, and 20mA with it on at full speed.