The Myth of Radio Spectrum Interference 603
Selanit writes "Just came across a fascinating article on Salon about a technologist who claims that there is no such thing as "interference" in the radio spectrum. He argues that interference is a symptom of inadequate equipment, not a fact of nature, and that with improved transceivers we could open the spectrum up to high-quality broadcasts by anyone. Reference is made to the GNU Radio Project. Neat stuff." We've posted other stories about this. I wonder if the "color" meme will catch on.
Limited Quantities (Score:3, Informative)
The FCC is currently forcing the switch to digital communications all over, which is shrinking the required spectrum. I'm sure when other technologies mature, they will make use of those as well to further free-up the spectrum.
Wha? (Score:4, Informative)
It's not that using a radio frequency somehow "depletes" a resource -- it means that if you put a green object in a green room with green lights, after a point you won't be able to see the object any more, kind of like how camouflage works. The problem is when you have a lot of signaling broadcasting in an area, the noise level can increase to the point that no single signal can be resolved. The classic example is how it's very difficult to understand a particular conversation in a noisy room. And that's why you have to generally parcel out radio spectrum and define limits on how it can be used (signal strength, bandwidth characteristics, noise levels, coverage patterns, etc)
That guy's nutty analogy makes me think he's a leftover of the dotcom era -- when eyeballs was more important than revenue and other silly things. Admittedly, I should read the whole article, but the first few paragraphs made me feel like I'm talking to a crazy guy on the bus.
bOINGbOING transcript from the spectrum conference (Score:2, Informative)
Re:shrinking the required spectrum.... (Score:5, Informative)
A major part of communications theory is issues dealing with bit-error rates, and interference. It is a reality. Now we can move to things like "spread spectrum" but even this is no panacea. Fact - for a given bit errror rate, bandwidth, and communications path conditions - there are a finite number of spread spectrum transmitters than can coexist in the same band before the bit-error rate is exceeded!
How do I know? Well I've been a ham for 25 years giving me practical experience, and I'm a EE as well.
The Stanford Spectrum Conference... (Score:5, Informative)
The idea that Spectrum doesn't need to be regulated is quite old, and it seems more and more likely to be valid. In any case, the idea that it needs to be controlled by government interests is less and less likely.
-R
Re:The article is crap (Score:5, Informative)
David Reed is many things, but crackpot is not one of them. He was a professor of computer science at MIT, then chief scientist at Software Arts during its VisiCalc days, and then the chief scientist at Lotus during its 1-2-3 days. But he is probably best known as a coauthor of the paper that got the Internet's architecture right: "End-to-End Arguments in System Design."
thank you for reading the article.
Take a college physics class 'tard (Score:1, Informative)
Re:Wha? (Score:1, Informative)
As interference minimization is really nothing new and he gave no cutting edge examples (the only example he mentioned was frequency hopping!) the article was simply high on hyperbolae (there is no such thing as interference) and low on any actual information.
He's also right in some ways..... (Score:3, Informative)
Re:complete bunk (Score:1, Informative)
Re:Wha? (Score:3, Informative)
Your point is also a good one, in that from an engineering point of view as the signals get closer together in the spectrum the ability to distinguish one signal from another is reduced.
However his answer to this is that the current method of spectrum allocation does a terrible job at utilizing the available spectrum partly because the transceviers we use for radio and television broadcast for example are relatively stupid and inefficient compared to what we could be doing, partly because of how the historical licensing stucture grew to be fixed ownership of particular frequencies and the space around them to allow dumb recievers to utilise them.
His idea is to try to promote the reduction of frequency requirements to the least restrictive set of rules to allow a reciever to recieve a broadcast from a broadcaster. One example given is through the use of smarter SDRs (software defined radios) to make more efficient use of the available spectrum.
Re:complete bunk (Score:2, Informative)
Re:complete bunk (Score:5, Informative)
Clearly there is no such thing as limitless bandwidth; Shannon's theory tells us there is maximum amount of information that can be transmitted over any one channel, and simple physics tells us that there are a limited number of channels, no matter how you slice it.
He's correct, on a technicality.... (Score:4, Informative)
The jist of the article is that RF waves do not "interfere" with each other. By this he means that two RF waves will not affect each other as they pass by each other in space. This is correct. The two waves will simply pass through each other. The problem is when you try to receive the signal.
When you receive a signal you get ALL the radio waves from the entire spectrum (not quite this simple, but it will do). Then the signal is amplified and the spectrum you don't want is filtered off. The problem is that if your antenna is receiving two RF waves in the same spectrum they will be superimposed.
What he's trying to say is that an intellegent receiver will be able to pick out one of these waves while rejecting the other. Much like when you pick out one conversation in a noisy room. Much easier said than done.
There are currently some schemes to do this, such as CDMA phones which work on a spread spectrum. Each of them transmit and receive on the same spectrum at the same time using what are called "codes" (Code Division Multiple Access). However there is still a capacity issue. When too many phones come into the same area, the noise floor comes up and nobody can receive information. To prevent this the cellular phone comany will limit the number of active cell phones in a given cell and drop any new calls over the limit.
There are more advanced methods, but as many people in this field know, the signal processing that your brain does to pick out only one conversation is mind blowing.
To sum up, he's technically correct. His use of the word "interference" is confusing to say the least. RF engineers talk about interference as the superposition of singnals as you receive them. He talks about interference as the interaction of signals in space.
Someone hand this guy a physics book, stat! (Score:5, Informative)
There are some very commonplace phenomena, such as the colors on a soap bubble or oil slick, which are the manifestation of interference of light. There are more fundamental experiments that can be done with lasers or radio waves to demonstrate interference.
Actually, if you do the experiment, there is a specific pinhole size at which you get the best image. Make the pinhole any smaller and the image starts getting blurrier because of diffraction effects which, loosely speaking, are due to the photons interfering with each other.
From his misunderstandings of the nature of light so far, it's impossible for him to have any real understanding of the quantum nature of light. He wouldn't know Schrodinger's equation if it walked up to him and smacked him upside the head, seeing as how Schrodinger's equation is a wave equation and predicts all sorts of interference phenomena.
The most fundamental problem is that he admits the notion of frequency, which is intrinsicly tied to the wave nature of light and radio. If he admits the wave nature of light, then he also has to admit interference of light as a natural phenomenon and not as a detection artifact, at which point all of his theories crumble.
Re:Wha? (Score:2, Informative)
CDMA systems showed us that it is possible to transmit two signals at the same time and the same frequency and distinguish them at the receiver; a task which at first might seem impossible. However, Shannon's theory still imposes limits on the maximum possible transmission rate.
What's new today is that by using multiple antennas it is actually possible to go beyond the limits Shannon established for point-to-point communication! This is not snake oil; it is well established, refereed research. In fact, it is already demonstrated technology [google.com]!
I still think it is a long, long way from these ideas to an unregulated spectrum.
Re:Wha? (Score:5, Informative)
Interference as we know it is the inability to derive meaning from information about the local radio environment. That's what happens when two people broadcast on the same frequency -- your receiver can't figure out which information to care about, because all it knows is "stuff on this frequency is important information" and we keep more than one person from broadcasting on more than one frequency by convention.
Where he seems to be going is treating the endpoints of radio communication more like endpoints in a network. Something analogous to modulation of a carrier frequency (in terms of complexity) is voltage modulation of wires in CAT5 cable. But network interfaces lay the notion of connections between two endpoints over something a good deal more abstract than that. They abstract the modulations into a binary stream, decode the binary into discrete data structures, interrogate the data structures to get meta-information about the data, demux the data (or defrag the packet, or reassemble the stream) based on the meta-information, and so on.
What he seems to be proposing is that radio receivers and transmitters do the same thing that network interfaces and protocol stacks do -- make the actual dance of bits considerably more complicated (to allow for things like error-correction when traditional "interference" is a problem, and to add more meta-information), then apply layered abstractions on top of it to get at the actual data.
Spread-spectrum communication does this already -- two SS messages can be sent to two SS receivers in the same range of frequency, because the two transmitters won't usually be broadcasting on the same frequency, and redundancy can be built into the transmission protocol so that when collisions occur, information isn't lost.
The article overpromises -- if I understand, this mode of communication is no better or worse than what we enjoy by using the OSI model to structure network communications. Even if the information space is "theoretically infinite" (which I doubt), we have to get increasingly more creative in how we utilize the space. In the networking world, however, we can talk at gigabit speeds over the same physical media that only supported 10mbps 10 years ago. We accept that wireless networking can find ways to squeeze increased "bandwidth" out of what is, in reality, a fixed width of spectrum allocated by the FCC.
What Reed seems to be agitating for is that the FCC and others get out of the way entirely, architecting a basic framework for the exchange of information and letting the transmitters/receivers figure out the rest of the details -- essentially the same thing he advocated for the Internet. I don't think it's a crackpot idea at all, though the style of the article masks that pretty well.
Yes, Claude Shannon says "he's full of shit." (Score:3, Informative)
In short:
YOU CAN'T TRANSMIT AN ARBITRARILY LARGE AMOUNT OF DATA/SECOND ON A FINITE AMOUNT OF BANDWIDTH. No matter how good your equipment, or how clever your signaling patterns, you will never be able to increase your data rate above the amount determined by Shannon's equations.
The flaw in Reed's reasoning is that we're talking about subdivisions of frequency, and the amount of data that can be transmitted in a given wavelength band has an absolute upper limit. It's Shannon's rule about bandwidth. So yes, Reed can go around giving everybody a gnat's ball hair width of radio frequency to push their data, but each riny segment will only be able to transmit a piddle of bits per second.
This is like people who don't know Calculus, but who think they've disproved Special Relativity with a thought experiment. Anybody who's sat through a class on it, or read a book, will laugh and laugh, while everybody who hasn't had the benefit of learning will probably be suckered.
Not entirely (Score:3, Informative)
What I found astounding... (Score:5, Informative)
There must be some other explanation, but it seems like Dr. Reed is making a freshman-physics terminology mistake. When a physicist says that two waves "interfere", he/she doesn't mean that one wave knocks out the other or that they undergo some linked dance. The linearity of Maxwell's equations indeed does show that each wave "passes through" the other without reducing or amplifying it.
Nonetheless, they interfere -- because "interference" is the interaction of the waves at a given point in space, where the amplitudes add algebraically. Consider a given location x at a given time t. If at that moment wave A has ampitude 5 and wave B has amplitude -2, then a receiver will measure a disturbance of amplitude 3. It doesn't -- and can't -- know that there are two waves, because there is only one signal. If the content in wave A is uncorrelated with the content in wave B (for example, two different radio stations playing different songs), then their addition will be essentially random -- and hence sound like noise (because it is noise).
Dr. Reed's proposal doesn't really speak to this. He wants smarter receivers that can track a signal and so distinguish wave A from wave B. The technology is not here, not cheap, and certainly not universal. The system we have was not foisted on us by some big government conspiracy and it's not maintained by the pressures of a cartel. It's here because interference is a fact and that "overcoming" it -- which is really more like shuffling past it -- is expensive and unproven.
And you would still have to deal with the transition from legacy to newfangled
Re:More than that... (Score:3, Informative)
You are still left with a limited piece of the spectrum and in this piece you are still going to run out of space (either in the frequency domain or in the code word domain). Shannon's law still applies for the signal/noise ration.
Jeroen
Re:complete bunk (Score:4, Informative)
The answer is none; you can't change the signal at all, so you can't send information. Once you start changing the signal, (i.e. change the amplitude) you are actually adding in more frequencies - this is Fourier 101.
To send information, you have to use a band of frequencies. The wider the band, the more information per channel, but the fewer channels. So there is a limited amount of information that can be sent.
Re:He's also right in some ways..... (Score:2, Informative)
Only a pure plane wave that is unchanging in time has exactly zero width in frequency space. As soon as you have a wave that lasts for a finite time (say, to transmit morse code or whatever) then you get a a definite width in frequency space, and suddenly the number of channels you can use is restricted.
In fact, the width of the channel is exactly proportional to the amount of information you can send down it, so using multiple channels provides no advantage (from an information-theoretic point of view) to using a single channel, since the total available capacity is a constant (for a fixed maximum frequency).
The practical limits on frequency are quite limiting. It would be a bad move, for example, to make a radio operating at gamma ray wavelengths. They are quite hazardous. And it only gets worse as the frequency is increased., up to the point that your transmitter will ionize the air around it.
Re:Someone hand this guy a physics book, stat! (Score:3, Informative)
It would have been much better if Reed had used the term 'interact' rather than 'interfere'. All waves interfere, as you point out.
The important point is that photons do not interact with each other (well, they actually do but the cross-section is so small that it is of no practical relevance). So, you can shine a laser at something, and the photons in the laser beam are essentially unaffected by passing through whatever background light in between the source and whatever you shine the laser at. This is a distinct effect from 'interference'.
And yes, just because something is non-interacting doesn't mean it doesn't occupy space. But it does mean that (in principle) an infinite number of photons can occupy the same space at the same time. So he is being very sloppy with his quantum mechanics, but its very hard to be precise when explaining these things to a magazine.
You are being no less sloppy with your statement that diffraction effects are "due to photons interfering with each other". You can do the same experiment with a single photon, and still get difraction. You probably already knew this, but I'm just making the point that its hard to explain quantum mechanics without being sloppy!
Interference is a *receiving* problem (Score:2, Informative)
problem, not a transmission problem. You also have to remember that
radio broadcasting predates the internet by almost 100 years. His main
focus seems to be to get needed spectrum for the expansion of the
internet into the wireless world. In the early days, the only way to
prevent interference was to separate the spectrum into pieces and assign
each user a specific piece. Up until the 1970's, there was no frequency
sharing between active users. This begin to change in the late 1970's
with the introduction of spread spectrum techniques. This is the
bandwagon that Reed seems to be jumping onto. However, there are
theoretical limits on how many users can share the same piece of
spectrum even using spread spectrum techniques - thus you still need a
spectrum policeman to decide who gets what.
Re:Reed is wrong (Score:2, Informative)
This sounds like the argument the phone company used to argue against allowing the Internet. Yes, computers cost a lot more than dumb phones, but people are willing to pay more for something that does more and especially for something that allows them to do more.
Phones haven't gone away, but allowing the internet has added greatly to our lives over the past ten years.
Re:Interesting thing about radio signals (Score:1, Informative)
It was probably the light traffic controller, a PLC probably, going through some sort of timing loop, waiting for the moment to change the light. Since the code is probably different when changing from red to green, than for green to yellow, the noise spectrum of the processor is different.
Re:complete bunk (Score:2, Informative)
The "few errors" you refer to are still interference. With a sensible frequency hopping pattern, the interference will spread out around users and be evenly spread in time, hopefully to the point where error correcting codes can catch it and compensate. But add more users and the error rate will pile up until your network falls apart, just as with non-hopping.
This effect is called "interference diversity" and is well studied in the literature.
Additionally, your throwaway line about "ask anyone who's signal you can see to choose a different color or time division on that particular color" would be enormously, insanely complex to implement. The amount of traffic necessary to keep this sort of scheme working would dwarf the useful traffic the network would handle; plus, this whilst it would improve things for a single user, it would likely make the next user over worse. It would not lead to a better network overall.
[Disclaimer: frequency hopping is my PhD thesis topic]
Re:This guy has no idea what he's talking about (Score:2, Informative)
Re:What I found astounding... (Score:2, Informative)
You are correct on the first point---the article does not really address the state-of-the-art of co-channel signal processing---how to process signals that overlap in frequency, time, and space. I like your explanation of interference.
However, regarding your other points, we do have methods to distinguish wave A from wave B. They are here. We have numerous signal processing techniques for processing co-channel RF/acoustic signals. It might sound like magic, but by imposing one or a few requirements on the transmitted signals---many of which are satisfied in practice---we can:
A few examples of the requirements we impose on the transmitted signals to do these sort of things: that they have constant modulus/envelope (e.g., the signal used in GSM phones), that they are statistically independent, that they are digital and use a finite alphabet of transmitted symbols (e.g., one symbol is transmitted for 1 and another for -1).
You are correct on another point: practical systems that incorporate these technologies are NOT cheap, but they do exist, in many different places. For example, some of these ideas are used in GSM transmitters and cell phones.
The article seems to imply that Dr. Reed is a big advocate of these kinds of technology as ONE means of dealing with interference, and they are important.
However, the article incorrectly refers to one of the GNU radio demonstrations as an example of co-channel signal processing. IN FACT, the code he links to takes TWO frequencies as arguments, not one as the author implies. The code just processes two different FM radio stations at the same time, NOT two FM radio stations at the same frequency.
Reed is only partly correct... (Score:4, Informative)
It's also true that two radio signals, each of a different frequency, will, when mixed together, produce an entirely different set of signals based on the sum and difference of the two frequencies.
This is the same principle that superheterodyne circuits (the type used in just about any kind of modern RF receiver) are dependent on. Example: You want to receive a signal on a carrier frequency of 146.5200 MHz, and your receiver has a 10.700 MHz IF.
OK, so the local oscillator (LO) in your receiver needs to produce a frequency of its own that will mix with the incoming 146.5200, and produce 10.7MHz as a result. That 10.7 signal will then be demodulated and turned back into audio.
Assuming you use low-side injection, your receiver's LO would need to generate a frequency of 135.8200MHz (this, by the way, is why scanning receivers are not permitted in commercial aircraft. 135.8200 is in the aircraft comm band), which is merely 146.5200MHz minus 10.700MHz.
Anyway... What I'm driving at is this; Think of a mountain top transmitter site that's got a ton of broadcast, public safety, amateur, and other kinds of transmitters on top of it, many of which are producing hundreds, if not thousands, of watts worth of RF.
There's going to be signal mixing. Lots of it. That means tons of the very "interference" that Reed doesn't seem to think exists.
The techniques mentioned in the article, BTW, including software-defined radios, are nothing new. They've been around for decades, and ham radio folk are already experimenting with them. For one example of a purely software-controlled radio, take a look at this radio kit from TAPR. [tapr.org]
73 de KC7GR
Re:complete bunk (Score:2, Informative)
basically what your saying is this:
noise = rand();
rx_sig= desired_sig+noise;
therefore:
desired_sig= rx_sig-rand();
how does that work? You can't know in advance what the noise spectra is.
Re:Someone hand this guy a physics book, stat! (Score:5, Informative)
The colors from a soap bubble are due to light interfering with light. Light is partially reflected from each surface of the soap film, and the reflected beams do interfere with each other and result in the colors that you see. That's about all the detail I want to go into describing it, but if I still don't believe me, it's probably described better and in more detail in either Hecht Optics, Born & Wolf Principles of Optics, or Lipson, Lipson & Tannhauser Optical Physics (in any of those books, look for the section on "multiple-beam interference").
It is true that when two beams of light cross paths in vacuum, if you were to observe them after they cross, you could not tell that they crossed. However, in the region in which they cross, they can interfere with each other. Again, any of the references I mentioned above will probably explain this much better than I can.
Even worse, once you physically manifest this signal by modulating it onto an electromagnetic carrier wave (like radio does), this communications spectrum is now superimposed on the physical spectrum of the electromagnetic wave. Now the signal is subject to the physical phenomenon of interference, which can further corrupt the signal if you don't allocate communications channels in the electromagnetic spectrum properly. And I think it's the allocation of commmunications channels which is what the article is trying to be about. However, that doesn't change the fact that Reed is dead wrong in the way he describes or interprets many of his physical examples, probably because he has a lot of background in computer science but not as much in physics.
Furthermore, Reed is wrong if he thinks that ultrawideband (UWB) or frequency hopping will increase the Shannon limit within a given range of the electromagnetic spectrum. Ultrawideband will interfere with other electromagnetic signals. It requires a lot of electromagnetic bandwidth, hence the name.
Frequency hopping can improve the efficiency of the spectrum allocation by moving communications channels to unused regions of the spectrum, but it does not create communication capacity where there is none. Furthermore, those channels have to be allocated in advance to prevent them from with other signals.
Reed is probably right that the electromagnetic spectrum is inefficiently utilized. But the many of the physical examples or explanations of physical phenomena that he presents are dead wrong, which was the point that I was trying to make in my original post.