More on Flexible SOC Design from Today’s GSA Luncheon

Less than two hours after writing my latest blog entry—Why Products Don’t Make it to Market: EDN’s Mind of the Engineer, Part 2—about the need for more flexible system design, I found myself listening to representatives from RIM (the company behind the Blackberry) and TiVo (the company behind—well TiVo) talk about the same subject as keynote speakers at today’s GSA Emerging Opportunities luncheon.

Jim Denny, VP of Product Marketing at TiVo, spoke first. TiVo has about four million subscribers and has partnerships with content providers (Amazon, Netflix, Rhapsody, YouTube, Yahoo!, NBC, CBS, etc.) and content carriers/distributors (Comcast, Cox, DirecTV). The company also has international markets for its DVRs. Although I’d mostly associate TiVo with cable- and satellite-delivered video content, Denny clearly has his eye on the high-growth opportunity in Web-delivered video content as well. According to emarketeer.com, the number of online video viewers will grow from about 138 million in 2007 to 190 million viewers in 2012, which will make that viewing audience about 60% as large as the number of expected TV viewers—projected to be 302 million viewers by 2012. A large, attractive audience to serve. Lots of consumers.

Denny stated that there are three drivers for serving these viewers:

1. A lot of content needs to be free
2. There must be enormous breadth of content
3. Content must be available on demand and purchasable a la carte

He noted that there’s been a huge increase in the amount of choice for online video over the last year and a half. While content providers exhibited a lot of reluctance to offer online video two years ago, they have been learning from the experience of the music industry and the rampant success of Apple’s iPhone. They’d like to see similar success and consumer uptake for online video. Although most video from online sources is viewed with a PC today, Denny clearly contemplates a future where TiVo delivers some portion of that video from the Internet to TVs. (I was amused to hear the term “over the top” used to describe services delivered over set-top boxes. That’s a new one on me.)

So what does TiVo want in the way of ICs to construct its future set-top boxes? The company knows that it will need flexible SOCs that support multiple media formats (H.264, VC-1, Flash, and variants), multiple security schemes (Microsoft DRM, DTCP/IP, AES), and multiple authoring formats (Flash and JavaScript/HTML). The need to support multiple video formats should be obvious to anyone who’s watched video using a PC. (Although he didn’t say it, the same is true for audio formats.)

There are many video formats although the number of important formats seems to narrow as we get into HD video. The need to support multiple digital-rights-management (DRM) schemes stems from the DRM being the “keys to the kingdom,” which every content provider desires to control. The need for multiple authoring formats allows developers to “use the right support toll for the right job.” You need a flexible SOC design to handle all of these formats.

Flexible SOC design means that the chip simply must be more programmable so that it can handle whatever the end market, the content providers, the carriers, or TiVo itself throw at it. In addition, there’s a strong desire for simpler system-development models. Right now, noted Denny, “it’s hard to move from one chip to the next, even with the same vendor.” He wants to see an easier way to port a system design from one chip set to the next, “at least within the same family of chip sets.” Otherwise, “it’s just as easy to move to another vendor as it is to move to the current vendor’s next chip set.” That’s a warning and a reflection of the reality of customer loyalty in the 21^st century. Success with this product generation doesn’t ensure success with the next generation unless you stay at the top of the game, give the customer a clear and relatively easy migration path, and—most important—build flexibility into your product to allow your customer to deal with the unknown future.

Paul Kempf, Vice President of Silicon at Research in Motion (RIM), is in a very different business but has very similar needs. RIM’s Blackberry family is a leader in the smart phone market. The original Blackberry was a data-only device. It added voice, color, multimedia, WiFi, and GPS over the years to evolve into the portable multimedia-capable mobile communications terminal that it is today. An early focus on data forced RIM’s design teams to pay close attention to data compression (because early comm channels had very little data bandwidth) and encryption (for business security). Coincidentally, these two elements are the focus of today’s smart phones and were also two of the elements Denny discussed with respect to TiVo’s chip needs.

The smart phone’s feature set is long and growing:

• Voice communications with speakerphone
• High-resolution color LCD with touch screen
• Wireless email, organizer, browser, SMS/MMs
• 3.2-Mpixel camera
• GPS with maps
• Bluetooth
• Media player and polyphonic ring tones

Putting all these features and more into an easy-to-use smart phone isn’t easy and not all designs are as successful as others. However, Kempf directly attributed some of the success to “feature creep,” which has made the smart phones more attractive to consumers. So expect more creep.

Expect displays to go from today’s ¼ and ½-VGA sizes to VGA resolution and higher. Expect multimedia functions to become core functions not add-ons as they are today. For example, look inside today’s smart phones and you’ll see two audio subsystems, one for voice and the other for low-power music playback. “That kind of thing has to stop” said Kempf. Camera quality [resolution and color depth] is also increasing, which will require more in-phone image processing and video handling. “We need hardware accelerators” said Kempf, “it can’t be done in software [running on a fast, general-purpose CPU].

Expect all-new and unexpected features to appear as well. One that Kempf mentioned that completely took me by surprise is the use of MEMS accelerometers and gyros to stabilize a displayed image while the user is bumping along. “Now that’s cool!” I thought. Consequently, you should expect the smart phone manufacturers to want to pack more features into the underlying silicon while reducing size and power consumption.

To date, Kempf noted, applications-processor and communications-processor development have been split on separate time lines. The advantage of this design approach is that the two parallel tracks can proceed uncoupled and can exploit developments as they happen without delaying development on the other time line. The disadvantage is that separate applications- and communications-processor chips require two packages and multiple memory subsystems, which increases BOM costs and makes the phone tend to be bigger. However, said Kempf, it’s easier to combine the two systems onto one chip at 40nm with little resulting compromise. You can see where he’s headed.