Slashdot Mirror
Ask Eric Blossom about Software-Defined Radio
Eric Blossom is an electrical engineer with a history of working with radio and communications security. He gave a presentation at the recent H2K2 conference about his work with GNU Radio, which is, bar none, the single most exciting software project in existence today. (Imagine computing devices that communicate seamlessly across the entire electromagnetic spectrum.) As usual, we'll forward some of the best questions to Eric and post his responses when we receive them.
[not a question for Eric Blossom unless highly moderated ;)]
Could actually someone explain what GNU radio does? Even after visiting the site, I stil have _no clue_ what it is and what's the goal.
Same for GNU Bayonne.
Both sound like cool projects but hard to understand what they do.
Would someone care to explain to the world what this project is (eg: with this, you can transmit RF waves using only a speaker), that'd be great.
Thx
This technology sounds like the kind of thing which could greatly add to the convergence of devices that clutter the electronic life. You could extend convergence not only as a Smartphone but have in one device (though perhaps not simultaneously):
1. Cell phone
2. Computing power (PDA)
3. FRS radio device
4. 802.11x network device
5. Police scanner
6. Television reciever
7. etc.
Have you been approached by police departments, FedEx, etc. to develop devices to allow their people to do more stuff in fewer packages?
If I'm interested in doing research in this field someday, and I'm currently a computer engineering major, what are some good electives that I might take? Aside from general programming necessities, what sort of signal processing courses are necessary to understand the underlying aspects of software-defined radio?
"I may be quite wrong." - Socrates
First, the old saw about "Don't say anything on the air that you wouldn't want the whole world to hear" has basically been discarded by the Cell Phone industry. They got Congress to put up legislation ensuring that the police need a warrent even to monitor your cordless phone. This is universally regarded as a dumb idea by most people who know what a radio is.
Second, they got certain modulation techniques declared off limits. It is illegal to demodulate an FM subcarrier without permission from the originator of the signal. Again, if I were monitoring SCA broadcasts from an FM station, how would they ever know?
Just as Open Source is driving copyright owners and licensing contract lawyers nuts, so too is Software Defined Radio. It basically brings all modulation techniques back within easy reach of the average Joe radio listener. No more of this nonsense of saying that demodulating police MDT transmissions is illegal.
My question for Eric Blossom is whether challenging the ECPA was one of the motivating factors behind this project. If not, are they the least bit interested in adhering to the ECPA's various prohibitions, or is that also irrelevant?
Nearly fifty percent of all graduates come from the bottom half of the class!
You won't need any external DSP resources, but the term "software-defined radio" is still a gross misnomer because you still need an RF front end to do anything really interesting.
At a minimum, this would consist of a mixer and 1st local-oscillator synthesizer, preferably with an RF preamp in front of it all depending on what frequencies you want to cover and what kind of antenna and feedline you're running.
The mixer and RF preamp can be off-the-shelf parts from Mini-Circuits or eBay.
The synthesizer is less convenient. I'm finishing an article on a "turnkey" octave-range VHF/UHF/microwave synthesizer design (http://www.qsl.net/ke5fx/synth.html that anyone with a decent soldering iron should be able to put together, but it's still not a trivial project.
A lot of software-defined radio efforts are targeted at relatively narrow bands like the 2.4 GHz ISM band, where a bandpass filter can theoretically be used to take advantage of ADC aliasing via undersampling. With a scheme like this, you can dispense with the mixer and synthesizer. I'm not optimistic about these kinds of ideas, though... to achieve competitive performance, every dollar you save on the mixer and 1st LO synthesizer will have to be spent on a super-high-quality ADC configuration. It seems like the sensitivity would be pretty crappy, not to mention the fact that plenty of interfering sources near other aliased frequencies would be likely to survive the trip through the front-end filter. (I will admit that I don't have any direct experience with this topology, though.... there are probably quite a few advantages that I'm glossing over.)
At any rate, though, by keeping the traditional front-end mixer and oscillator while letting software handle the IF processing, you can use an ordinary sound card in a direct-conversion scheme (see this month's QEX to demodulate just about anything in the entire spectrum -- FM, NBFM, AM, SSB, FSK, QAM, you name it. To me, that seems a lot more interesting than the "Antenna at the ADC" schemes that a lot of people are hyping.
Dahlmann tightly grips the knife, which he may have no idea how to use, and steps out into the plain.
Actually, you DON'T need to sample a 10.7 IF at double speed.
The Nyquist theorm states you must sample your signal at not less than twice the bandwidth of your signal. You can limit the 10.7 IF to 60 kHz bandwidth and then sample it at 120 kHz just fine - this is called subsampling.
In effect your analog to digital converter becomes an additional mixer in the processing chain. If you think about it, sampling is just multiplying the signal by a stream of Dirac impulses in the time domain, which corrisponds to convolution by a Dirac impulse in the frequency domain - just like an analog mixer.
HOWEVER - you HAVE to bandwidth limit your IF correctly - if you sample at 120kHz and you have stuff leaking through at a bandwidth of 70 kHz you are screwed - your signals will fold over in the frequency domain.
For those who are interested, look up the Intersil 50216 - it is a dedicated chip that does much of the signal processing for you.
What would be COOL would be for the FSF to sell PCI cards with a 50216, 50217, and an FPGA - then you could really do some cool stuff.
www.eFax.com are spammers
This project is to the radio what the general purpose computer is to the abacus.
If tits were wings it'd be flying around.
This discussion has an apalling lack of 'peers' who are RF qualified.
Software radio has been a fad for at least a decade, but in a much less ambitious sense. Mostly the focus has been on using embedded DSP in place of dedicated radio circuits - and the cost/power performance has moved in this direction (generally), particularly for smaller manufacturing volumes .
A 'universal' radio is substantially more difficult. Radio sensitivity is limited by the inherent noise of the receiver, the dynamic range of the receiver components, including the demodulator, and gain. These components are in 'conflict', that is, more gain can limit the noise of the receiver, but may limit dynamic range more.
There are also difficult architectural issues. A direct conversion radio (converts in one step to baseband) suffers from LO re-transmission (illegal if over certain limits), DC-offset and drift, noise and dynamic range. Single conversion to an Intermediate Frequeny (IF) simplifies these problems, but the choice of Local Oscillator (Synthesizer), including noise characteristics, switching frequency and bandwidth, combined with the necesary amplifier gain, broadband filters, mixer dynamic range, input power and noise, are complex.
Any attempt to transmit raises complex circuit and regulatory issues.
Radio requires a real investment in time to understand RF circuits, modulation theory, encoding and decoding techniques, protocols and regulations.
Jumping ahead to a Linux application that 'assumes' an always appropriate (wideband or universal) radio seems like an unfortunate waste of time.
The same code might be interesting, however, as a tutorial on modulation and protocols - if it's any good, but there are better sources.
In my experience, it's much harder to develop DSP software compared to developing hardware for the same task. Hardware design these days is a mature engineering discipline. Software development remains a high-risk black art, DSP doubly so. Also, for some signal processing tasks, the demand on processing power far outstrips the current processors. For something like a 3G modem you might need 20 Gigaflops to implement it in software. Even a plain well designed comms receiver from 20 years ago with a couple of crystal filters cannot be fully emulated with today's DSP technology. Complemented and enhanced - yes, but not emulated. How do you show that software-defined radio is up to more than just filling a very limited and specialised niche?
"Argue with idiots, and you become an idiot." -- Paul Graham
You can just use a soundcard if that's all you have. It will get you 44 kHz of bandwidth. Then you'd need a rf tuner in front of that.
On the high end, we use a $1k+ 20Ms/s card with a cable tuner.
It is widely believed, but not yet proven, that you can coax raw samples from a BT8x8 video capture card, which would be a great boon, as those are cheap.
No. I think you missed the point of the question, one I would like answered also. What equipment is really needed to receive radio signals? For example, the website shows looking at the FM band, but talks about the same high end $1k+ card that you mentioned that goes up to 20Ms/s samples. Clearly more equipment than this is needed to deal with the 107mhz FM band. You are apparently assuming we have something lying around that will comvert the FM band to a lower frequency. Many of use don't, unless you want to count an FM radio (which makes the need for a software radio less pressing). If we're talking about software modulation/demodulation let's just say so, if we're really talking about "Imagine computing devices that communicate seamlessly across the entire electromagnetic spectrum." then lets spell out all of the equipment that is needed to pick out signals that might be in the Ghz range.
I'm an American. I love this country and the freedoms that we used to have.
Tangential to this.. is there any talk amongst the GNU Radio folks on building a piece of hardware that complements this software project, or is supposed to work with whatever devices the user has on hand/will build?
I used to be someone else. Now I'm someone better.
Real life is underrated.