We subconsciously think of Wolfson as small, perhaps because your chips are small. But do you think of yourselves that way?
Yes, we still see ourselves as a small company in Edinburgh, but at the same time our customers are some of the biggest companies in the world, so we think we're punching above our weight.
You recently signed a multi-year agreement with Samsung to be its "primary audio partner," apparently leaving your Apple days far behind. But Samsung only accounted for 30 percent of your sales last quarter. What's the rationale behind emphasizing one customer like that?
We already had a long relationship with Samsung, and we're in a lot of their devices, including their phones, tablets and cameras. But these days we enable more features on our audio architecture, and those features require much closer working than the normal supplier-customer relationship. Samsung wants access to our system-level IP and the design freedom to use it. And in return we're allowed to work with them on their future-generation products.
You're in plenty of Samsung smartphones -- but not in the flagship Galaxy S III or Galaxy S 4 for sale in the United States, which both use Qualcomm application processors and audio. How will you grow your presence in the US and other LTE markets?
Samsung has always used different people and different platforms, because of its size. Qualcomm specifically has had a very strong footprint in the US because they bundle their audio hub with their main processor and it has traditionally been difficult for manufacturers to break that bundle because of problems with software integration.
"We're now competing like-for-like."
Qualcomm is a rarity in that respect. Samsung's Exynos, NVIDIA, Renesas, Marvell and the others don't work that way, so it's been easier for their processors to work with our chip.
Things are changing though. This is the first generation in which Qualcomm separated out its own audio chip, which means we're now competing like-for-like. If Samsung had wanted to use Qualcomm's audio in its Exynos phones, it could have -- but it chose to go with us. There's also the new Sharp AQUOS phone that uses a Qualcomm application processor alongside our audio chip. So customers who stuck with the bundled option for the first generation can now look further afield to differentiate their product.
You say that phone manufacturers go to Wolfson because they're looking to differentiate. But if everyone goes with Wolfson, what happens to that argument?
The power of our next-gen device, the WM5110, is that it's a platform just like a smartphone is a platform. It has four DSP cores with a lot of processing power, which means it can run different apps, different software, and we're just making sure that those apps are getting access to the best power efficiency and the best audio quality. We're also giving them the tools to write software for the platform.
A lot of your reputation comes from expensive hi-fi products. You claim to bring that across to mobile, but to what extent is that just marketing?
Yes, we have high-end chips for hi-fi, but we're able to record and measure what those signals look like. We architect our mobile products from the same intellectual property and we're able to use software to make the signals sound the same -- so we're able to get pretty close to the hi-fi products.
You'll always make a trade-off with power and signal-to-noise ratio, but we're never going to compromise on audio quality. Our philosophy is never to destroy the signal before it gets to the speaker, and that's what we're trying to do.
"Smaller transistors are better for digital processing, but worse for analog."
Are you going against the grain in terms of forcing manufacturers to make room for a separate chip, rather than integrating audio into the main system-on-chip?
That trend has actually reversed in the past couple of years in relation to audio. So much is going into that single chip now, including graphics and everything else, that it means that in order to get enough processing power you have to use smaller transistors. Smaller transistors are better for digital processing, but worse for analog. You get to the point where you just can't do good audio on small transistors, so our strategy is to make a separate chip to do audio. We used to use 180nm transistors, then we moved two nodes down to 65nm and we think our IP will work down to 40nm, but we can't go south of that. [For ref: many high-end mobile processors these days are 28nm. See our Primed article on transistor size for more.]
Aside from that, our processor doesn't share memory, doesn't interrupt the main processor and doesn't need to wake it up -- so we can do things like voice activation without using the application processor and draining the battery.
One area we have seen you in is feedback circuitry that allows for better audio from speakers -- the kind of thing promoted by HTC and NXP. Can you improve mobile speaker audio too?
Feedback circuits are something we're planning to deliver in the WM5110. We're trying to put all of that intelligence [i.e., real-time monitoring of audio output] into the main audio hub, which can reduce the overall price of the system.
How expensive will the WM5110 be? Will it affect the price of devices that have it?
It's going to be significantly more expensive than our current products. But in terms of billable materials, what we try to do is eat up other components, such as speaker protection and noise cancelling, so that the costs balance out. With a strong enough audio processor, adding those features, and even adding studio master sound, is just a matter of software.
Will Wolfson always be an audio company?
Audio will always be our foundation. But we can also use voice commands to change the way people interface with their technology, and we can also help with other real-world signals.
We're very good at taking analog information, getting it into the digital domain, offering low-power manipulation of those signals and then getting them back out into the real world again. There's going to be lots of different uses for that.