Computing vs Communications

Engineers who design computers must meet many simultaneous and conflicting design goals, such as high CPU speed at low cost and low energy consumption. These engineers are knowledgeable of the various necessary technologies, from computer architecture to semiconductor manufacturing, and know the tradeoffs. While their task is extremely complex, it also is deterministic, in the sense that the assumptions and operating environment of the computer system are only minimally affected by random phenomena. While there is no agreed upon definition of an optimum computer, various system features can be maximized, subject to various constraints as illustrated above.

However, the design framework is profoundly different for communication systems engineers. For them, random processes, such as thermal noise in circuitry and signal multipath phenomena in wireless systems such as the iPhone 4 dominate every aspect of system design. These engineers must use probability theory as the paradigm in which they design. Their core task is to design a system with acceptable bit error probability, subject to cost and technical system constraints similar to the computer constraints, such as low energy consumption, high transmission bit rate and physical size of the chip set.

The mental processes and ways of seeing the world are very different for  computer designers and their managers compared to engineers and managers of communications/wireless systems. Indeed, one of the most annoying and bizarre phenomena for the users of mobile phones is to get a huge, random variation in call quality while both parties are stationary. They, along with computer engineers, don’t appreciate how channels can produce signal fades that reduce received signal power to 1% of the original power level without any motion at all. Making this effect subjectively worse is to note how often the “bars” don’t correspond in any obvious way to the user’s call quality. This effect results from idiosyncratic methods (often required by “marketing”) vendors use to generate bars on their displays.

Apple

Boiled down to the nub of meaning, Jobs:

1. Admitted and explained how the outside antenna could be affected by users touching it across an impedance gap, thereby changing the effective size and impedance. So far so good.
2. Promised all owners either a full refund upon return of the iPhone 4, or a free case for the phone, designed to solve the antenna problem. Again, good business decision.
3. Claimed all Smart Phones have problems with connection quality, and showed some video of how phones made by Samsung, Research in Motion and HTC  behave similarly to the iPhone as a result of hand position. Apparently, this was an attempt to minimize the impact of his design problem by effectively saying “Hey, everybody has antenna and connection problems, not just us.”

However, his pitch was misleading. Point 3 is where he went off the rails, giving us a chance to elaborate on an important culture clash in the electronics world. Jobs ignored the core point that just a change in “signal bars” when the hand changes grip says NOTHING about the quality of the call. Did he know this fact? Did he care? We will ignore this particular speculation and place his actions in a larger context.

We claim the antenna problems of Apple were enabled by a compute culture which, according to EETimes, knew of the antenna problems but simply forged ahead anyway.

It is highly unlikely managers in a communications culture would make this same decision. In fact, wireless system engineers fully appreciate the particular tradeoffs of antenna outside vs inside the phone case. But it was a compute-centric culture within which this disastrous decision was made. Coincidence? We think not.

Intel provides another Example of Communications in a Compute Culture

The clash of these cultures is further exemplified by Intel’s failure to profit in the wireless systems arena, in spite of around $3B invested (Intel spent at least $1B to acquire DSP Communications in the late 1990s, and has since spent at least $1B promoting WiMax, not to mention its $1B investment into Clearwire, a WiMax service operator.) We claim this failure by Intel drives from its corporate DNA which is compute-centric, and thus doesn’t have a workable intuition and value system about the wireless world. In this regard, Intel’s successful compute-centric business ironically caused the huge success of WiFi because of Intel’s huge CPU volume shipping with integrated early WiFi chips.

More Culture Wars

The iPhone, while enormously successful, succeeds because of the excellent compute culture of Apple, just as Intel’s compute culture drives its success. The majority of software and silicon in the iPhone, as in every Smart Phone, is compute centric. Very little of the processor capacity is used for actual cellular or WiFi communications!

Warning to Qualcomm

By exactly the logic of the culture war, Qualcomm will face grave danger as it tries to include more compute centric devices and systems into its own phone designs and product offerings. Qualcomm, for example, purchased a display making firm and now is innovating the nature of cell phone displays. Qualcomm also is trying to promote video content for its video technology. These and other functions are quite different from the work of Qualcomm’s intensely communications centric engineers.

Nokia Animation on Cognitive Radio.

Enjoy!

Our briefing on Cognitive Radio is here.

CR_SIG_2-25-10

Our major point is that to advance the effective use of limited spectrum, market and regulatory forces are more uncertain and dominant than technology, which is already being deployed within many standards.

The panelists and their presentations can be found here:

WCA Cognitive Radio presentations from panelists

Defined by Joseph Mitola, the original coiner of the term as, “a really smart radio that would be self-aware, RF-aware, user-aware, and that would include language technology and machine vision along with a lot of high-fidelity knowledge of the radio environment,” cognitive radio has become a catch-all bucket for all sorts of radio intelligence. Got transmit power control? Definitely CR. Have adaptive frequency hopping? You’re in! Can your radio operate in more than one band? You’re a CR pioneer!

While all these ‘features’ are elements of what Mitola had in mind, the true vision goes way beyond these singular advancements into truly ‘intelligent’ radios that can operate in and avoid interference in any band based on a priori knowledge to predict what bands are open, where interference may come from and what the user’s application may be. Much of the intelligence for this will reside at a higher level in software, but the capability to act upon and take advantage of that intelligence depends on advancing the state of the art in radio design. That’s where ISSCC comes in.

Even with the various interpretations of the term Cognitive Radio, it is a very lively product development activity. CR is real today and is most readily understood as a variety of techniques used to install adaption and goal seeking into transceivers. These advances will clearly use bandwidth more effectively.

At the layman level, the San Jose Mercury News has an article about the exploding demand for bandwidth and how the FCC’s efforts to provide more spectrum can collide with other FCC initiatives. For example, spectrum can be more effectively allocated by various pricing mechanisms, but these in turn affect the proposed policy of “net neutrality”, in which telecom operators are forbidden to distinguish types of traffic for purposes of network management.

Recently, the market research firm iSuppli released a “tear down” analysis” of several smart phones. Their teardown analysis of the actual phones reveals the costs in the end product’s bill of materials. You can read an article about this study in The Economist’s article: “The lowdown on teardowns” . The results are shown below in Table 1. In Table 2 we simply show the costs normalized to each firm’s total Cost.

Table 1: Comparison of Component Costs in Four Smart Phones

Table 2: Costs Normalized to each Firms’s Total Cost.

We take these quite low costs as signs current technology and design teams are competent to build any CR system desired over in the near future. Of course, we still don’t have clarity on the regulatory regime or markets…..such details!

The most likely outcome, in our humble opinion, is that WiMAx will be a de facto monopoly in wireless broadband, until LTE finally deploys. Then, given the usual insatiable demand for bandwidth, both WiMax and LTE will co-exist. We seriously doubt that either of these deep technologies, repleat with financially heavy duty backers, will fail.

See IEEE Spectrum for more.

The ubiquitous mobile phone is emerging as one of the most important sources of data about people. Every time someone uses a mobile phone, information can be collected, stored and analyzed, including the phone’s time stamped geo-location, characterisitcs of the user’s voice patterns deduced by signal processing on-phone or at operator facilties, imagery directed by the user, medical emissions and physical activity. This large array of possible applications is enabled by a variety of measurements taken by the sensors on the phone. The  data is processed on phone and in the network. By using measurements from many cell phone users, “crowd sourcing”, or distributed computation for applications, can be accomplished.

The Nexus One platform sports these sensors: a 5-megapixel camera with a Xenon flash, a light sensor, proximity sensor, an accelerometer and GPS receiver.

Here is a list of some generic sensor types and possible applications:

Download “Prediction Markets”

See Lloyd Nirenberg’s presentation to the Wireless Communications Alliance on 20 SEPT 05. He shows how new research about “Spectrum Games” will enable economic concerns to guide allocations and enable unlicensed bands to increase their capacity. New wireless businesses will emerge from CR and SDR technologies combined with Spectrum Games.

Download “Let the Spectrum Games Begin“