The Internet of Things (IoT) market is exploding, but not because of a plethora of new products is hitting the street. Mostly it’s because every company seems to be jumping on the IoT bandwagon by re-assessing the applicability of devices in their current product portfolios…and now bundling many of those existing devices under the IoT heading. That is to show that they are a major IoT (or Machine to Machine, M2M) player, too. Of course, M2M doesn’t require an Internet connection, but that doesn’t deter companies from labeling those devices as IoT, too. After all, a Morse Code Key and Sounder could also be labeled as M2M devices (and I once got up to a dozen words per minute on such a key).
The 50 billion (or a gazillion, if you like) unit IoT forecasts are meaningless, since they seem to count every interconnected device (wired or wireless) on the planet. Even devices that will never employ the Internet directly (like Bluetooth) are counted if they can be interconnected in some way. And home automation networks, like ZigBee and Z-wave multiply such interconnections and are also counted as IoT; after all, many alarm monitoring agencies are connected to each of those sub-networks by the Internet.
So, until we have more explicit definitions of what IoT (or M2M) devices are…or will be, market numbers aren’t to be relied on and we’ll have to live with the current hype.
Synopsys Introduces ARC DSP
The company that was once known as ARC International PLC (for Argonaut RISC Core) marketed general-purpose configurable microprocessor IP that could be optimized for speed, energy efficiency or code density. Although the basic architecture evolved into a popular configurable RISC/SIMD engine, the company never directly exploited the inherent DSP capabilities of the design.
Now owned by Synopsys, its ARC division has announced licensable ARC EM DSP Processors which combine high-efficiency control with ultra low-power DSP. The design employs a single-cycle 32×32 MAC and the ARCv2DSP ISA adds over 100 new DSP-focused instructions, including vector/SIMD, matrix, saturating and complex math. An optional floating point unit is also available. The design is aimed at energy-efficient voice/speech, audio and sensor data processing. Synopsys has long enabled clients with their own DSP implementations, but now it can be direct.
C-RAN Lives! But Mostly In Asia
In my last newsletter, I presumed that development of Cloud RAN (C-RAN) was dead. That assumption was largely based on China Mobile’s C-RAN website that was last updated in September 2012. I have since found that the website suffered a server fail-over where the backup server lost all of its information, too. So maybe we’ll see the website rising again.
But more importantly, my mention of C-RAN’s presumed death brought me emails and calls from a variety of sources, including chip vendors, IP consultants, and operators. Collectively, they showed me that C-RAN is alive and well, with successful trials and even commercial implementations in various forms in China, Korea and Japan.
There are two primary versions of C-RAN; the simplest of which is simply to stack the DSP baseband units (BBUs) associated with the fiber-connected remote radio heads (RRHs) from each base station in a central location (some refer to that as “front-haul” as opposed to traditional backhaul to each base station). That C-RAN approach is called “hotelling” of BBUs, and is a relatively straightforward implementation.
The second and more ambitious approach to C-RAN still connects the RRHs to a central location, but rather than dedicate BBUs for each base station’s fiber, The BBU function is pooled through real-time network virtualization. Traditional DSPs are probably not applicable for the BBU function…as traditional CPUs (read X86 or ARM) coupled with accelerators (currently as FPGAs) are employed. In short, no collection of DSP cores is necessary.
Korea Telecom’s Cloud Communications Center has already implemented the hotelling BBU approach while China Mobile is working on the more ambitious second approach. China Mobile is working with partners Alcatel-Lucent, Huawei, IBM, Intel and ZTE and has trials at 7 different locations. Although France Telecom and others in Europe have explored various approaches to Cloud centralization, most implementations of C-RAN currently seem to be confined to China, Korea and Japan. DoCoMo in Japan employs the basic hotelling BBU approach for C-RAN, but is also tied in to a number of small-cell base stations for better coverage.
Why Asia? First and foremost is the availability of fiber. 96% of China Mobile’s base stations are already connected by fiber, while microwave backhaul is dominant in the West. Second is the monthly operating cost of that fiber, which is far cheaper in Asia than in the West. Since I don’t see an abundance of cheap dark fiber coming soon in the U.S. or Europe, Asia will likely remain the focus of C-RAN.
As always, I encourage your feedback.
President & Principal Analyst