Thank you for your insight, I believe you make some valuable points.

To design a competitive solution, you need a dedicated hardware platform. In my opinion, SDR, in it's traditional sense, involves minimal amount of hardware (little more than an antenna and an ADC) and implements signal processing on a conventional general-purpose computing platform. Even those which use dedicated DSPs and FPGAs do not fall under this idea of an SDR system. Reconfigurability does not necessarily imply SDR and the use of dedicated signal processing modules defies the beauty of SDR.

It seems that more and more alternative solutions are emerging that bestow a system with the reconfigurability and flexibility of SDR, but provide improved performance, cost and size. My point, therefore, is that there is not place for SRD, in its pure form, in the mass market.

I agree with Johannes that this "pure" view is not very practical. SDR was always envisioned as a slow evolution from completely unconfigurable systems (i.e., hardware-defined systems) to reconfigurable systems (defined by software which may be uploaded to effect different personalities). For instance, see Mitola 99 "Software Radio Architecture: A Mathematical Perspective", the block diagram in that work is far more complex than what you are calling an SDR. I would suggest that at today's processing capabilities and expectations of extremely (or even moderately) high data rates from consumers the ultimate goal that you describe is not in reach. But this thing you speak of is not the one-and-only definition of SDR.

To me, the advantage of an SDR is its reconfigurability and its flexibility. Why else would you want one? So, here's my question to you. If you have a truly reconfigurable radio (say something even better than the prototyping platforms on the market today) that can be reprogrammed (by a host computer or its own "intelligence") to modify its carrier frequency, data rate, modulation format, etc... Albeit based on some "custom" but reprogrammable hardware. What is the difference whether we call it massively reconfigurable hardware-defined radio (MRHDR?) or software-defined radio (SDR)? Aren't the end goals of both the same? What more does the platform you describe offer us above and beyond this reconfigurability? I would argue that the amount of software and programmable hardware that goes into platforms currently available merits the name software-defined radio as it is the software that gives the radio its definition (operating parameters, signal characteristics, etc). Most of the SDR research that interests me would not be classified as SDR by your definition. I think this definition is too narrow. After all, I would suggest that our personal computers are software-defined even though they may have custom hardware attached (e.g., high performance video cards).

Hi Daryl,

Thanks for your comments, I agree, the definition of SDR as I had originally put it is quite narrow. In my view it represents the, perhaps unattainable, ideal case goal. Where the hardware defines only the limitations (in terms of bandwidth/centre-freq/digitizer resolution) of the receiver, while the software is unconstrained and can implement any modulation scheme / number of channels etc.

Your comments on the difference between massively reconfigurable hardware, and a completely software defined architecture are interesting. I do believe there is a difference and, I expect that recognizing it is important. To pick up on your point about video cards; they are a good analogy. Upon reaching the limits of what the general-purpose CPU could do, the video community migrated to a platform that, while multi-purpose, was designed with video processing in mind (interestingly, it has been used extensively in the CDMA community).

To answer your question on what the difference between SDR and massively reconfigurable hardware-defined radio (MRHDR) is: I think that it’s a question of how the field develops within the research and open-source community. This purist idea of SDR has provoked a lot of interest in the signal-processing and information theory filed, and is quite prevalent in the cognitive radio field, for example. A significant amount of open-source projects are currently ongoing which offer support for such an architecture. Having the least amount of hardware, SDR is the most portable and accessible. To achieve the same level of accessibility and portability with MRHDR, we would need widely adopted hardware platform.

This sort of dedicated platform is missing in the SDR arena. Is it the case that SDR on a purely general purpose commuting platform is failing and that some intermediate platform, dedicated to radio processing, is needed? Many products are available today which offer a flexible RF down-converter/digitizer, which offer raw IF samples across a standard interface (USB/Ethernet). Do we need to go one step further and port some of the basic RF/IF/Baseband operations to this hardware platform? Should the community stop focusing on getting more and more out of the standard CPU and work on assembling a standard set of SDR operations/instructions that could be implemented on a Radio equivalent of the GPU, some sort of general purpose RPU, that could not only host the hardware interface, but also the basic, general purpose SDR instructions?

Hmmm... Good points. I think that we (as a research community) are going to need some sort of additional processing power to develop competitive systems. I think that this will require additional skills above and beyond the standard Matlab or Python processing that has been prevalent (and is a good start). If we want to build real-time systems that are competitive in a commercial market, then we are going to need something more. This is the reason that I have been learning to take advantage of FPGAs over the past few years (and continue to do so). While I recognize that high-priced FPGAs are a cost disadvantage, I see no other clear way to make high-speed reconfigurable radio systems possible. I believe that the biggest hindrance in arriving at a well-developed platform is the lack of a clear definition and vision for SDR. SDR means different things to different people, and this is obvious at every conference I've attended. For example signal processing engineers tend to think of it a lot differently than networking engineers. What do you think would be the basic requirements for an RPU? I think it could be very helpful for the community, but I would also be a little worried about providing functional blocks to researchers that tie them to specific technologies (GSM, GNSS, GPS, WiFi, etc.). I feel that this might limit the creativity of the research community and to me it seems to go against the spirit of SDR.

I believe that SDR is already there but overtaken by other players so it is rather masked in the market. Idea of SDR has been adopted by the market. No-one could think about cognitive networks without flexibility of SDR. Also, all multi-technology devices on the market follow idea of SDR.

Initial idea we more than good but the push to the market was weak due to initial hardware availabity problems and good and strong market segment that could recognise itself in the field of SDR. It was simply left to individual activities. Hardware improvements, open standards and new radio technologies and spectrum sharing will keep SDR alive for years but I think that we shouldn't expect to have it widely recognised as SDR.

I think the main issue is lack of consumer demand specifically for a "do anything" device. People prefer their gadgets to have the long battery life which which specialised hardware can deliver.

However bedside clock radios added functions for DAB+ WiFi streaming media years ago.

Hardware is now 'manufactured' as a (soft) IP design, e.g. Imagination: IP CPU design -> Frontier Silicon: radio software and integration as System-on-Chip -> Cisco's Pure Digital: end-user devices.

From their comments in the press Frontier Silicon uses SDR to give it the flexibility to specialise products to meet an increasingly fragmented market.

Surely, I think there is a future for SDR, in military as well as civil domains.

What do I expect from an SDR platform?

Broad-band (few MHz to few GHz), low-noise figure, phase-stable, high computational powered with option of DSP/CPU/GPU and FPGA, expandable by coherently adding SDRs.

I don't expect configuration_on_the_fly, though, rather a fixed role each time it is programmed, to wit: being a wireless communication terminal, or a ranging radar, or a silent GSM/3G signal detector.

Current research efforts allow at least lab based demonstrators for all of the above and much more, for instance: beamforming, direction-finding etc. Down-side is that heavy processing is usually shifted on a high-performance laptop or a PC.

There are challenges w.r.t cost, integration, reducing the noise figure, and adding computational power on board. I think the push towards modular open system approach (MOSA) by military , might hasten the commercialization efforts for SDR.

I'd dare say that we might never be able to get one platform for all type of applications SDR, but instead a specific SDR for a specific area, such as: EW, wireless comm, or monitoring etc.

My two cents..

Yes, what IS the hold up.  Good comments all.

I'd like to see us push the SOA toward the "little more than an antenna and an ADC".  ADCs are getting close to achieving direct L-band digital sampling, eliminating all the nonlinear components and filters in the standard superhet conversion to IF.  This has obvious advantages for channelizing Mult-constellation receivers.  Plus it gives you access to the signal more directly. Signal processing typically done at high cost in analog hardware, such as signal conditioning, anti-jam, multipath mitigation etc. can be done in software, and more sophisticated algorithms can be brought to bear on the application.  Adaptive algorithms that combine functions, such as interference and scintillation detection, signal authentication,  integrity monitoring, receiver resource management and so on. Direct digital conversion with sufficient sampling rate, resolution and bandwidth eliminates the need to clean up the signal and opens the door to doing a lot more.  Your processor, algorithms and imagination set the limit.

Even with today's ADCs you could come close to direct sampling.  About 2 GHz with 12-14 bits is available commercially, though quite expensive.  Cost is the other reason I think you don't see more of a push - dual frequency ADC modules are available with RF front-end conversion built-in, including AGC and channel balance, for a fraction of the cost of the highest speed ADC-only.  You can't quite satisfy Nyquist at the carrier frequency, but you can more than cover the signal bandwidth. What you are left with is manageable if you are more careful about the ADC errors, effect on noise and correlation, etc.

"Build it and they will come."