John Chapin of Vanu Discusses Software Defined Radio
Vanu’s work comes from the SpectrumWare Radio project at MIT. SpectrumWare was the leading academic SDR project in the 1990s. It used off the shelf computer platforms such as the Intel X86 rather than DSPs as the Software Radio platform. Vanu was founded in 1998 and today has nearly 10 years of research. Here’s the discussion we had:
The lowest layer in your architecture is the antenna. What do you think about MIMO? Vanu’s technology applies equally well to MIMO as to a single antennas but MIMO costs at least twice as much processing time, because there are two or more digital streams. As long as the processor has enough power, it doesn’t make a difference to our system.
What’s interesting is that you use the term “motherboard” in your radio. You’re basically applying a PC architecture to a radio. Is Vanu unique in this area? What makes Vanu unique is that it doesn’t have a DSP or FPGA in it. Those may be useful as waveform accelerators but so far Vanu doesn’t need them. Other companies center their software radio designs around FPGAs. In terms of using a computer-design style, most high-end radio designs such as those from Spectrum Signal Processing, Rockwell Collins, or Harris, are building boards that look like high-end computers but with accelerators on the back end of it. One difference between software radios and standard computers is that software radios require high-speed I/O which is something of a departure from standard PC architectures.
Your 2002 paper indicated the use of Linux. Do you still use Linux or do you see something better out there? Yes, Linux has many advantages for this area. We have shipped on several POSIX OSes and can port to other OSes if necessary.
What are the challenges with Linux? Linux is not a real-time OS but the benefits are so significant that it outweighs the drawbacks.
What are the major benefits of Linux? The OS is free and the software development tools are low-cost. On any new platform, Linux is usually the first OS available. Linux provides a huge advantage in diversity of hardware. There’s a wide-range of tools available. Some are free while some cost money.
You mention real-time being an issue. With the performance of today’s processors, do you still find this an issue? It is still an issue since the requirements of the waveforms are continuing to evolve and are much more demanding. By running only a few OS threads and doing resource allocation at user level, one can achieve real-time without having a real-time OS.
It appears you used buffering to overcome limitations in earlier systems. Do you still recommend buffering? There are latency requirements in some standards such as CDMA that prevents us from using a high level of buffering but it’s still used in some cases.
Do you see LabVIEW fitting into the application? LabVIEW would be useful for test systems. For Vanu, LabVIEW would be useful for VANU as a base station tester or a “gold standard” phone for testing base stations, or a device that pretends to be 20 to 40 phones. Base stations are hard to test because it’s manual and it involves a lot of physical hardware.
You use the TI DSP processor as an example of why waveform portability is important. National Instruments can cite the same example. In the early 1990’s NI built a DSP processor board. As the host processor of the PC moved up in speed, the DSP processor on a board became a bottleneck to the overall performance of the system.
Where are you going in the future? Vanu is a waveform company. We’re looking at the up and coming standards such as WiMax, 3G HSPA and others.