Analog or Digital Radio? Not the Most Relevant Question
By Rudy Lauwereins, IMEC, Leuven, Belgium
For some engineers, the choice between analog and digital seems as vital as being religious or atheist. In a world with ever-more-performing electronic devices, sold at practically no cost, one will preach about “digital radios” while the other strongly favors advanced analog solutions.
Rudy Lauwereins, IMEC, Leuven, Belgium
By Rudy Lauwereins, IMEC, Leuven, Belgium
For some engineers, the choice between analog and digital seems as vital as being religious or atheist. In a world with ever-more-performing electronic devices, sold at practically no cost, one will preach about “digital radios” while the other strongly favors advanced analog solutions. As is often the case, the truth lies somewhere in the middle. Besides, “analog or digital?” is not the most relevant question in the entire discussion.
What is in fact a digital radio?
Theoretically speaking, it's a radio in which the analog-to-digital conversion (ADC) is performed directly after signal reception at the antenna. Practically, however, this would require an ADC capable of sampling at frequencies up to 12GHz, which is not the most straightforward condition to achieve. Moreover, the dynamic range of such a component would have to be enormous, covering strong incoming signals (close to the base station) as well as very weak ones (far from the base station). Let's consider for a moment that one would succeed in developing such component, it would undoubtedly consume amounts of power – and area – that are unacceptable for mobile devices.
Clearly, I was astonished when I read a recent announcement on a digital radio used in handhelds. Doom scenarios of IMEC's analog designers suddenly being expelled to a life of unemployment vividly appeared to me. But a look under the hood of the digital radio technology quickly reassured me. They where not doing more (or less) than what everybody is aiming for: finding the most scalable technology.
This immediately introduces the question engineers should really ask themselves in this discussion: “How does one design an analog front-end in such way that it will scale at the same pace as the digital baseband according to Moore's law?”
Does it make sense in shifting the ADC towards the antenna? Definitely, if it would result in a more power- and area-efficient solution at the system level. Definitely not, if your ADC would end up requesting more power and area than when leaving him on the more traditional spot. It's no coincidence that this option has not yet been implemented in IMEC's advanced radio solutions. Simply because the answer to the question always was “definitely not”. It's only now that we've found an entirely new ADC concept – requiring far less power than classical approaches – that we can think about moving the AD-conversion closer to the antenna.
A second approach towards “digitized” radios is the rethinking of the components themselves. The 'ancient' technology of switched capacitors formerly stranded at fairly low frequencies, but now 'easily' reaches RF. This immediately explains its sudden revival in for example filter technology. Also the new ADC concept developed at IMEC makes use of a comparable approach of switching of capacitors (mind the nuance).
But also in this context, the analog-digital discussion flourishes. A component that uses (active) switching in theory is digital. But when it makes use of continuous amplitudes and contains a lot of passives, should we then not call it analog? Especially seasoned marketers are experts in cunningly exploiting this kind of paradigms.
The 'marriage' (as opposed to 'religious' separation) between analog and digital is probably most visible in the third approach towards more scalable front-ends. At IMEC, we often refer to it as C3: calibration, compensation and control. Analog components have the tendency to produce non-linear distortions in the signal. Because of the incompatibility of this non-linearity with certain mathematical operations, they cause a lot of trouble during coding and decoding. But, since everything becomes smaller and the problem of non-linearity increases in scaled technologies, there is no other option to learn and live with it.
One possibility is to calibrate the components at-runtime. This can be done by introducing a known signal into the circuit and calculating the deformations at the output. Once the distortion is known, it can be compensated for in the original signal (called pre-distortion). On top of this, advanced analog technologies allow to minimize the imperfections by controlling certain parameters in the components themselves. This all seems quite logical in theory, but can be a true mind breaker to effectively implement it. It's only thanks to the relatively enormous computing power that has come available that these complex calculations can also be made in mobile applications.
To summarize, it's the capacity and performance of the digital circuits that mainly determine how far you can reduce the area occupied by analog components. All things considered, a true “digital radio” will not be on the market tomorrow unless, of course, as a marketing buzzword.
Click here for more information on the IMEC programs: IMEC
Rudy Lauwereins a PhD, is vice president nomadic embedded systems (NES), at IMEC (Inter-university Microelectronics Centre), Leuven, Belgium. He heads a team of researchers dedicated to developing the design methodologies, tools, and architectures that will find their way into wireless systems and multimedia products over the next three to ten years. Dr. Lauwereins is a frequent speaker and leader at industry and academic conferences, including ASP-DAC2004 (Asia South Pacific Design Automation Conference), DATE (Design, Automation and Test in Europe), and RSP (IEEE International Workshop on Rapid System Prototyping). He is a senior member of the IEEE and the recipient of numerous grants. Dr. Lauwereins earned both a Master’s degree and PhD summa cum laude in electrical engineering from Katholieke Universiteit Leuven.
Advertisement
