Wireless baseband inflection point – SDR as a technological breakthrough

After reading Robert Cravotta’s article on Recognizing technology’s inflections it brought to mind just how many changes our industry is going through. It appears that Moore’s Law can be applied to many different fields in the embedded processor market, and perhaps that the original two year timeline for each generation has become even shorter.

One specific market my company participates in that is going through very dramatic changes these days is wireless baseband. We see an ever growing demand for bandwidth intensive content on mobile devices. Smartphones are definitely a key driver of this growth and according to the latest market study by Futuresource Consulting, by 2013, more than one billion people will own a smartphone.

Beyond Smartphones, this very dynamic space is being driven by a plethora of new and evolving consumer electronic devices demanding high bandwidth data delivery. Take the recently introduced Apple iPad for example. This product will certainly draw a lot of attention in 2010. If the iPhone was the key driver of 3G services in the US during the last two years, the iPad will further escalate the growing consumer demand for ‘on-the-go’ content. The iPad adds to a mobile market where there are so many other new and evolving products demanding mobile broadband: netbooks, portable media players, gaming consoles, eBook readers, cameras, and camcorders. The latest AT&T numbers show that it added 2.7 million new wireless customers in the fourth quarter of 2009. But this is not the big news - out of the new 2.7 new users 1 million are non-phone devices, with devices like the Kindle, Nook, and Sony Reader.

With such strong market opportunities, both on the terminal and infrastructure sides, baseband chip vendors are racing to implement next generation solutions to address this inevitable increased demand for broadband applications. In the US, for example, Verizon plans to deploy LTE networks that will reach 100 million potential customers this year. LTE markets will get even hotter in 2011, when analysts anticipate the release of the first LTE handsets and a sharp uptick in LTE base station spending. Growth will continue to build through 2013, when analysts predict that there will be over 72 million LTE subscribers.

All this opportunity is certainly not without its challenges. The level of system complexity brought about by this demand for higher data rates is exceptional. LTE Cat-4 needs to support a downlink data rate of 150Mbps. Just as a reference, advanced HSPA networks today support typical user download rate of either 3.6 Mbps or 7.2 Mbps. In addition, as we move towards 4G, multiple wireless air interfaces must be supported on a single device in order to meet different baseband evolutions and standards in various geographical regions. HSPA+ is a good example as it continues evolving in all the latest 3GPP releases. HSPA+ is built on top of the original WCDMA networks and will enable data rates to reach 42 Mbps on a 5MHz bandwidth, without requiring the same level of investment by the operators as LTE. Another consideration is geographical, such as China where the Chinese LTE equivalent version named TD-LTE, is based on TDD rather the commonly used FDD mode. And then there is WiMAX, LTE-Advanced and other standards that need to be taken into consideration.

So the road towards 4G involves multiple paths that require baseband chip vendors to enable higher flexibility than before. The cost of development and the multiple evolving standards increase the risk of the traditional hardware-based design approach. Hence, this higher demand for flexibility, naturally leads processor architects to a new strategy and technical approach. The concept of implementing multiple standards in software is addressed by an approach named Software Define Radio (SDR).

SDR means a single programmable platform can be used to support multiple air interfaces without the need for chip replacements. However, this brings many new challenges both in terms of architecture capabilities and software development. We need to look into new types of architectures as conventional DSPs and CPUs are incapable of meeting both the processing requirements and power and cost constrains of 4G chips. Robert’s article mentioned VLIW (very-long-instruction-word) processors like the C64x that have to support up to eight operations in parallel. When we talk about 4G, we need to think an even higher level of computation capabilities. In this regard we can mention the CEVA-XC that supports up to 64 operations in a cycle. The CEVA-XC is architected around a new type of emerging technology called a vector processor. It basically further extends the VLIW concept to much larger scales of ILP (Instruction-level parallelism).

Now, providing such extreme processing capabilities may become a huge burden on software development if not equipped with a very advanced development environment and an efficient compiler that enables software developers to develop their software in C and the compiler will be responsible to efficiently utilizing the processor resources.

So the wireless baseband arena will definitely continue to be very interesting in the years to come. We can expect new markets with new devices, new technologies to address many different challenges and a lot of innovation in the processor technologies and chip design to implement the new technologies.

Eyal Bergman, Director of Product Marketing for CEVA