The whole point of Cognitive Radio is very simple: it's a radio that can program itself.
A corollary of that point is: it's a radio that can and will adapt *by itself* to functioning in response to situations and requirements that *you don't and can't in principle know yet*. So a cognitive radio has to be, first and foremost, a general purpose computer in the technical sense. You might say, well, I can trim all this down into firmware once I've got a sufficient picture of the application domain. Unfortunately, no, you can't, precisely because of the corollary. There's no way around the fact that, if you want a Cognitive Radio, there *has* to be some general-purpose computing power inside beyond the pure "radio" functions, and that's true *no matter how much* you take out of the "computer" side and graft it onto the "radio" side. It's a theorem. Two theorems, actually. One of them is due to Jorma Rissanen concerning the asymptotic nature of certain binary sequences, the other being the Halting Problem. To give up on that is to have something other than a Cognitive Radio. Now, whether a Cognitive Radio is what you want (or need) is a wholly separate question. But in the end that's what determines whether you want a "computer" or something else inside your box. 73 Frank AB2KT Jim Lux wrote: > At 08:31 AM 1/4/2007, Lyle Johnson wrote: >>> ... It makes no sense to run anything but Linux on >>> the embedded controllers and then have them be a server. >> Um, I respectfully disagree. It may make no sense to run Vista / XP / >> Windows-flavor-of-the-era on an embedded controller, but if, for >> example, one wants to embed a DSP inside an SDR, it doesn't necessarily >> need to run Linux, and the DSP does not have to be Intel or AMD silicon >> with their many watts of power dissipation. >> >> It just needs to do be able to do the signal processing and have a way >> to get data in and out, get commands in and get status out. SDX and >> Suitsat2 (aka Odyssey) are examples of this. > > > The advantage of Linux on a PC style platform is cost, and that's > what drove the original SDR1000 design. You could get the box to do > the dsp (nonoptimized as it might be) anywhere. > > You can get a complete PC in Mini-ITX form factor with enough crunch > to run dttsp for under a couple hundred bucks. You'd be hard pressed > to find another DSP solution available off the shelf for that kind of > cost. Sure, there's a pile of DSPs out there and a pile of FPGAs, > but even in eval board form, they tend to be more expensive than a Mini-ITX. > > The other issue is development tools. There are a plethora of free > (or nearly free) tools for developing on the PC platform (Linux or > Windows) but not so many for other platforms. I do a fair amount of > work on developing software radios on the Xilinx Virtex II platform, > and the tool chain is NOT cheap for those parts, especially if you > want simulators, etc. > > So, you trade off the use of commodity hardware and software for > lower initial investment. Sort of like the Beowulf high performance > computing concept. > > Jim, W6RMK > > > > _______________________________________________ > FlexRadio mailing list > FlexRadio@flex-radio.biz > http://mail.flex-radio.biz/mailman/listinfo/flexradio_flex-radio.biz > Archive Link: http://www.mail-archive.com/flexradio%40flex-radio.biz/ > FlexRadio Homepage: http://www.flex-radio.com/ > > FlexRadio Knowledge Base: http://kb.flex-radio.com/ > _______________________________________________ FlexRadio mailing list FlexRadio@flex-radio.biz http://mail.flex-radio.biz/mailman/listinfo/flexradio_flex-radio.biz Archive Link: http://www.mail-archive.com/flexradio%40flex-radio.biz/ FlexRadio Homepage: http://www.flex-radio.com/ FlexRadio Knowledge Base: http://kb.flex-radio.com/
