Audience earSmart Not in Upcoming iPhone, Suggests New SoC?

Since the iPhone 4, Apple has been including Audience earSmart voice processors in devices to improve both receive and transmit audio for handsets. The iPhone 4 started the trend with a discrete Audience voice processor, and with A5 Apple made the unique move of integrating that IP into their own silicon, something we have yet to see be done with any of the other handset vendors. Since then I've seen Audience in a host of handsets for suppressing noise, reducing echo, and cleaning up audio in both near (phone on your head) and far field (speakerphone and hands free) modes. Getting clean audio is important not just for understanding voice calls, but also maximizing any potential silent gaps in normal speech which in turn enables a significant power savings on voice calls. In addition recent emphasis on handset voice recognition furthers the need for cleaned voice audio.

Recording demonstrations of what the transmit audio sounds like has been a regular feature in our smartphone reviews, and I've included a few comparisons above just as a refresher. 

Yesterday Audience provided news to investors that although it had developed and licensed a new voice processor for use by a particular major OEM (which was noted as Apple in a follow up conference call), it does not believe it will see inclusion in the OEM's next upcoming product. 

Pursuant to a statement of work under the MDSA, amended in March 2012, Audience developed and licensed a new generation of processor IP for use in the OEM's devices. However, the OEM is not obligated to use Audience's processor IP.

 

Audience now believes that it is unlikely that the OEM will enable Audience's processor IP in its next generation mobile phone. Audience is not aware of any intended changes by this OEM to its use of Audience's processors or processor IP in prior generations of the OEM's mobile phones.

In the follow up conference call, Audience notes that it had met all its deliverables for a new IP block designed to Apple's specifications, finished tape out, but is confident that it is not implemented in the next upcoming product due to a lack of final testing or confirmation.

There are a number of other players in the noise rejection game, and most are solutions from the SoC vendors themselves. For example, both Qualcomm (Fluence) and Texas Instruments have and license their own noise reduction technology usually alongside purchase of any SoC and PMIC combo, though these implementations either use blind source detection with an array of microphones or a beamforming technique. In addition audio codec players such as Wolfson also have solutions for handset makers.

Whether Apple has gone with an in-house solution of its own or licensed one of these other blocks is unknown. A number of carriers have voice quality standards which much be met to get approval or subsidy, I've been told by a number of players that AT&T's voice quality standards are some of the highest out there, so it will be interesting to see whether some other IP has been selected or if an Apple-proprietary solution is included. I look forward to running our noise rejection tests on any upcoming device and seeing what differences emerge. Interestingly enough, Apple recently shipped its own two-microphone beamforming system in the Retina MacBook Pro. 

Also of particular interest is what implications this has on making an accurate prediction of what SoC will be in the noted upcoming iDevice. Audience's earSmart IP was previously integrated into the A5 SoC (S5L8940) directly, and to the best of my knowledge was inherited into the 32nm A5R2 (S5L8942) revision. As a result it isn't a logical jump at all to suppose that we'll see a new SoC entirely for this upcoming iPhone launch. Commenter ltcommanderdata points out as well that another possibility is that the IP block is simply not enabled in favor of another solution or for cost reasons.

Source: Audience