2003 Software Defined Radio Conference
Wave Issue 0345 01/09/04
The origins of the SDR date back 10 years with the US Department of Defense. Under the JTRS (Joint Tactical Radio System) a Software Communications Architecture (SCA) was developed which has become the basis for military radio procurements. However the potential of SDR goes well beyond the military. Early regulatory efforts, including those by the FCC, recognized the potential of this technology and now include regulation activities in Europe and Japan.
With wireless filling out the anytime anywhere matrix this technology is becoming all the more important. At the same time the radio technology has become very complex. The standard for GSM is 5,000 pages. The objective of SDR is to move the digital and software controlled sections of the radio as close to the antenna as possible. This has profound impacts on the architecture of the radio and the processing rates of the components. Some feel it requires fundamental rethinking of the architecture of how processing is done to accomplish the radio. Now some highlights.
University of Karlsruhe
Professor Friedrich Jondral gave an overview of SDR including some of the latest work at the university. Some of the interesting points included:
They did a spectrum utilization assessment at Lichtenau, Germany in 2001. Those spectrum plots were shown and was illustrated how little the spectrum bands are actually being used. The measurements went from 50 – 1GHz. This shows the potential for overlay uses of the spectrum if radios can operate on a non-interfering basis.
Research at his institute examined the concept of Licensed Users (LU) and Rental Users (RU). The assessment was done of a LU based on FDMA/TDMA and the RU being HIPERLAN/2. The intent was to show that these two very different technologies could co-exist.
In the LU/RU problem, to keep the impact on the radios small, there was no “out-of-band” coordination of the spectrum use. Thus, the effectiveness of this technique was very much dependent on interference detection and minimization.
One of his graduate students, Arnd-Ragnar Rheimeier, presented a concept for modular radios. He suggested that a SDR means that the A/D and D/As are as close to the antenna as possible and that the DSPs be done in software, when flexibility is desirable.
It was suggested that significant work is underway in the automobile industry with SDR. The realization has come that all cars will have GPS and radio technology of some type – even including satellite radio. Currently there are a number of projects to integrate the radios for both into one.
Mike Chartier, the Intel Ombudsman for “Radio Free Intel” reviewed the state of SDR regulatory actions. He made a number of points that included:
If SDR is an inflection point in technology it must also be met by an inflection point in the regulatory environment.
Key regulatory issues are spectrum management, device certification and competitive issues. This latter point asked the question if the radio interfaces should be Open.
Intel characterized the regulation in terms of generation. 2nd Generation SDR proceedings are already addressing “overlays.” That is, the deployment of agile and smart radios for innovative spectrum management techniques. One of the most interesting suggestions is the consideration of an Overlay in the existing TV bands. This is particularly valuable spectrum because of its relative long wavelength and excellent propagation characteristics compared to 2GHz and 5 GHz. A 2nd FCC SDR proceeding is planned on cognitive radios.
The CTO of Xilinx, Ivo Bolsens, looked at enabling the future of SDR. Key points included.
SDR is at the union of computational and wireless complexity in user devices. An example was given of a channelization problem which has 1GHz 8 bit samples, over 16 channels which in turn feeds 16 complex filters with 14 taps. This results in a compute requirement of 30GMAC/s and memory bandwidth of 400 GB/s. Another example is a smart antenna with 12 elements and 12bit converters. The I/O requirement is 30 GB/s.
From Xilinx’s perspective FPGAs are the best way to address these needs. One of the reasons is that FPGA have been in close sync with the semiconductor fabrication technology, especially the reduction in line width. This is, in part, due to the regular nature of the gate arrays and the value of high gate counts to fabricate essentially parallel and random structures. He stated that last month Xilinx announced a 1B transistor gate array.
A major issue is the cost cross over point between FPGA and ASICs. In an interesting graph Ivo claimed that FPGAs have a cost advantage over 90nm, 300mm based ASICs up to 200k to 300k units.
Xilinx is seeking to recast the role of the FPGA in this market by creating
a platform with their products. In order to do this there must not only
be the gate array but the tools to accomplish the design. Ivo claimed
that the combination of:
Dr. Jorge Pereira, Scientific Officer, European Commission DG XIII, Mobil & Personal Communications, provided a refreshing view of the role of that the technology could play. It is expected that SDR will eventually lead to “full spectrum sharing.” Given the long term leases of spectrum, which has been gained from auctions, it could be 10 – 20 years before the full scope of available spectrum can be used in a shared environment. There is another twist, those holding spectrum, which is not being used, may be forced to lease it. In the near term, actions are being taken by the FCC, German RegTP and UK RA to consider options for Dynamic Spectrum Allocation.
At the center of Jorge’s views was a broad
vision for 4G or the next generation of radio technology. It includes:
Dr. John Chapin, CTO of Vanu gave what he described as a software-centered view of SDR. He sees the industry in transition from hardware to software. History has shown when such transitions happen the challenge of implementation moves from hardware which becomes progressively easier to implement and with a rise in the software challenge and its complexity. In SDR this is particularly forbidding.
He cited the SDR hardware challenges which include:
Consistent with the view that software will rise in complexity as it
overtakes hardware he provided the following challenges:
The following factors only serve to multiply the cost of software development: need to support multiple waveforms and field upgradeability.
All the more difficult is that many of these radios are both mission critical and life critical. To meet these goals implies costs which are X those without such requirements.
His overall message was one of caution, from a company that is doing just what he described.
Here is a sample of papers presented at the conference.
FPGA Tradeoffs for Software Radio Applications
The Use of Ontologies for Self-Awareness of Communications Nodes
Field Trials of an All-Software GSM Basestation
Compact Multi-Band Handset Antennas for SDR Application
Open-Source Experimental B3G Networks Based on Software-Radio Technology
Biometrically Enhanced Software Defined Radios
One can only take a long range view of SDR. It is at its earliest stages. The technical challenges are very formidable and the regulatory issues daunting. The prospect for realizing anytime anywhere any bandwidth any mobility is significant.
Speculating on the future is easy, what is difficult, is the present. It was very clear that SDR is enormously complex. The scope of this was evident when it was suggested in a talk that SDR is the end of standards. That is, the radio implementation will be so complex that standards will not be possible. We regard this as a reflection of the immaturity of the technology and not a fatal statement on the futility of standards.
What will the wireless landscape look like when that time comes? Will there still be value in emulating today's motley assortment of wireless standards, or will there be enough convergence around UWB, 802.11n, and 3G such that one or two radios can do it all? In this case a highly reconfigurable radio would be less valuable in a consumer application?
We strongly doubt that the PC will ever move to the requirements of the date flow machine which is at the heart of SDR, especially at the front end of the radio. However, even at SDR we saw examples, from Vanu, that PCs can have a major role. Thus, it remains to be seen what role an attached computation device will have in SDR. This is an important question that remains to be addressed.