The Myth of Interference

[shanek]

I was debating whether to post this here or over in P&CE. Since it's heavily technical, and I don't understand it all, I thought I'd post it here in Science so that greater minds than mine (of which there are many, especially on this forum) can help me out.

This is a new Salon article about a guy who basically says that bandwidth over radio waves is unlimited, and it's our outdated technology that's the limiting factor. He calls for the ending of government ownership of radio waves in favor of "smart" radio receivers that can somehow distinguish one signal from another.

Any idea how this is supposed to work? Is it a good idea, or just woo-woo stuff?

http://www.salon.com/tech/feature/2003/03/12/spectrum/index.html?x

Some selected quotes:

"Interference is a metaphor that paints an old limitation of technology as a fact of nature." So says David P. Reed, electrical engineer, computer scientist, and one of the architects of the Internet. If he's right, then spectrum isn't a resource to be divvied up like gold or parceled out like land. It's not even a set of pipes with their capacity limited by how wide they are or an aerial highway with white lines to maintain order.

Spectrum is more like the colors of the rainbow, including the ones our eyes can't discern. Says Reed: "There's no scarcity of spectrum any more than there's a scarcity of the color green. We could instantly hook up to the Internet everyone who can pick up a radio signal, and they could pump through as many bits as they could ever want. We'd go from an economy of digital scarcity to an economy of digital abundance."

The problem isn't with the radio waves. It's with the receivers: "Interference cannot be defined as a meaningful concept until a receiver tries to separate the signal. It's the processing that gets confused, and the confusion is highly specific to the particular detector," Reed says. Interference isn't a fact of nature. It's an artifact of particular technologies. This should be obvious to anyone who has upgraded a radio receiver and discovered that the interference has gone away: The signal hasn't changed, so it has to be the processing of the signal that's improved. The interference was in the eye of the beholder all along. Or, as Reed says, "Interference is what we call the information that a particular receiver is unable to separate."

That's what the telephone companies do: They add Caller I.D., and now their network is more valuable. We know it's more valuable because they charge us more for it. But the end-to-end argument says that adding services decreases the value of a communications network, for it makes decisions ahead of time about what people might want to do with the network. Instead, Reed and his colleagues argued, keep the network unoptimized for specific services so that it's optimized for enabling innovation by the network's users (the "ends").

That deep architectural principle is at the core of the Internet's value: Anyone with a good idea can implement a service and offer it over the network instead of having to propose it to the "owners" of the network and waiting for them to implement it. If the phone network were like the Internet, we wouldn't have had to wait 10 years to get caller I.D.; it would have been put together in one morning, implemented in the afternoon, and braced for competitive offerings by dinnertime.

But surely there must be some limit. "Actually, there isn't. Information isn't like a physical thing that has to have an outer limit even if we don't yet know what that limit is. Besides advances in compression, there's some astounding research that suggests that the informational capacity of systems can actually increase with the number of users." Reed is referring to work by researchers in the radio networking field, such as Tim Shepard and Greg Wornell of MIT, David Tse of UC-Berkeley, Jerry Foschini of Bell Labs, and many others, as well as work being carried out at MIT's Media Lab. If this research fulfills its promise, it's just one more way in which the metaphor of spectrum-as-resource fails and misdirects policy.

[Stimpson J. Cat]

I don't know if he is a loony, or if his idea is just being misrepresented, but radio waves most definitely do interfere with each other.

He is correct in stating that photons don't interact with each other, but the interference is not due to an interaction. It is due to the fact that photons of the same frequency are indistinguishable from each other.

Bottom line, for any given frequency, if you combine different signals of that frequency, it is not possible to resolve the result back into the original signals. The information simply isn't there anymore.

The theoretical limits to bandwidth are pretty easily defined. The maximum frequency that your signal can be varied at is one-half of the frequency of the carrier signal. The maximum information that can be conveyed in one period of your signal depends on the background noise.

You can, in principle, make the frequency band of your signal as narrow as you like, but the narrower it is, the lower your signal frequency must be, compared to the carrier frequency. This means that the total information that can be transmitted by a certain frequency range is proportional to the frequency range, and inversely proportional to the signal-to-noise ratio in that range.

From the point of view of the receiver, the best you can do is try to minimize the noise within the receiver. But even if you take it to the theoretical limit of zero, there is still a certain amount of background noise at any frequency range.

Dr. Stupid

[Aoidoi]

I'm hardly an expert on RF theory, though I do have a bit of practical experience in cellular systems.

In cell phone systems there are basically 3 (current) ways of passing information over RF and dealing with interference. The first is by allocating a particular range of frequencies to a user. The second is by allocating a particular slice of time to a user (1 out of 8 in GSM, IIRC). The third is by assigning each user one of a set of orthogonal codes which he then uses to decode messages being sent.

The first 2 techniques are inherently inefficient and require strict access rights to prevent users from interfering with others. ie if you're using somebody else's frequency generally neither can user can function well. There has to be an authority that assigns timeslots and frequencies, otherwise the system just doesn't work. Using those technologies I don't think you could get close to what he's arguing.

In option 3 (CDMA) there is a central authority that assigns codes to users but everyone shares the whole frequency range (for the most part... I'm not going to get too far into details here). The code determines things like throughput, and the whole system is constrained by interference, meaning that if one guy is blasting RF he can hurt everybody else. Meaning that without a central authority assigning codes and performing power control the system is a mess.

I'd assume the OP guy is talking about some other form of technology, but damned if I can figure out what it'd work like. RF has different properties at different frequencies, I don't think you could (or would want to) broadcast on the whole range with something like CDMA... that'd be a freakin mess, shortwave frequencies would be bouncing off the ionosphere all around the world and much of the spectrum would only reach the horizon, for instance.

I'm not sure what the claim he's making about unlimited information transfer is... given the entirety of the RF spectrum I suppose it's sort of possible, but the technical and physics limitations seem... daunting. His reference to compression is somewhat irrelevant, as compression can be used regardless of transfer medium, the only issue being that RF tends to be less reliable than, say, fiber optics and therefore generally has more error correction because you simply cannot know the RF conditions beforehand.

I'm always reluctant to say that anything is technologically impossible, so I'll just say that there would have to be some serious breakthroughs to make what he's arguing for practical.

[garys_2k]

Perhaps he means some sort of packet-switched system, requiring two-way communications. In any case, to call it "unlimited" is overstating things to the extreme.

[scotth]

For him to say it is strictly a receiver issue implies he already knows how to get around the problem.

If he doesn't have the receiver solution, he doesn't have anything but alot of hot air.

If he does, why is he talking about it instead of selling it?

It is my opinion that he has no idea what he is talking about on this one. His comment, "This should be obvious to anyone who has upgraded a radio receiver and discovered that the interference has gone away: The signal hasn't changed, so it has to be the processing of the signal that's improved. The interference was in the eye of the beholder all along." was much less than apt. These better recievers of which he speaks were designed completely within the bounds of existing theory. The places in which to make improvements were entirely driven by it.

An antenna receiving signals from two transmitters is nothing more or less than a mixer. Without prior knowledge of what one of the signals is, I can't imagine how they could be seperated. In fact, I am pretty sure I can prove that it is impossible. It seems pretty obvious to me that that for any point in a signal that is a composite of two or more signals, there are an infinite number of possible combinations to make the composite signal. How is a receiver to get what the original signals were? Actually, all I have to prove is that there is more than one solution for any given composite signal. I can visualize that easy enough. It should be trivial to prove on paper.

That make sense to you, Dr Stupid?

[Stimpson J. Cat]

scotth,

An antenna receiving signals from two transmitters is nothing more or less than a mixer. Without prior knowledge of what one of the signals is, I can't imagine how they could be separated. In fact, I am pretty sure I can prove that it is impossible. It seems pretty obvious to me that that for any point in a signal that is a composite of two or more signals, there are an infinite number of possible combinations to make the composite signal. How is a receiver to get what the original signals were? Actually, all I have to prove is that there is more than one solution for any given composite signal. I can visualize that easy enough. It should be trivial to prove on paper.

That make sense to you, Dr Stupid?

You are absolutely correct. And it is pretty trivial to prove. If you have two signals,

x1 = c1 * sin (w * t + a1)

and

x2 = c2 * sin (w * t + a2),

where c1 and c2 are the amplitudes, w is the frequency, and a1 and a2 are the phase shifts, then the sum of these two signals is

x1 + x2 = c3 * sin (w * t + a3),

where c3 and a3 depend on c1, c2, a1, and a2. If I alter any of the above four parameters, I can alway adjust the other three in such a way as to get the same values of c3 and a3.

In particular, I can just set a1 and a2 both equal to a3. In that case, we have c3 = c1 + c2. Alternatively, I could set a1 = a3 - pi / 4, and a2 = a3 + pi /4. In this case, we have c3 = sqrt(c1^2 + c2^2). And so on.

It is impossible to resolve what the original two signals were from the composite signal. That is what is meant by interference, in this context.

Dr. Stupid

[BillyJoe]

Hear, hear. :cool:

[scotth]

I knew it would be trivial to put down on paper (if I got someone else to do my homework for me).:D

[garys_2k]

After re-reading the article I think I see what he was getting at, but it still seems to be ridiculous to speak of bandwidth as unlimited.

If you take all the boradcast technolgies we now have, AM, FM SSB, spread spectrum versions of these, digital, and use a mix of them you could get more bandwidth than we now have, but that's about it.

If you read about the "gnu radio" linked from the article you can see that it still relies on an analog RF front end, but uses a "software IF strip" to pick out signals with all sorts of modulation schemes. But that reliance on old technology on the front end shows that this is far from unlimited bandwidth.

[Agammamon]

Or he could mean that the EM spectrum if infinitely divisible and that newer technologies can distinguish between and broadcast signals at tighter bandwidths, of course there is still a lower limit on how tight neighboring frequencies can be squeezed. In the US the FCC places limits on how close in frequency 2 stations can broadcast, this is done to limit interference; but the FCC has the same problem as most other governmental regulatory agencies, namely a lack of responsiveness. The FCC is using guidelines set up for tech that is several decades old.

[scotth]

Originally posted by Agammamon
Or he could mean that the EM spectrum if infinitely divisible and that newer technologies can distinguish between and broadcast signals at tighter bandwidths, of course there is still a lower limit on how tight neighboring frequencies can be squeezed. In the US the FCC places limits on how close in frequency 2 stations can broadcast, this is done to limit interference; but the FCC has the same problem as most other governmental regulatory agencies, namely a lack of responsiveness. The FCC is using guidelines set up for tech that is several decades old.

This gets you no where.

When information is encoded on a carrier signal (modulated) it MUST spread its freqency spectrum.

The only way to infinitely divide a spectrum up would be to put no data on each carrier wave.

As soon as data is encode, the carrier is no longer monochromatic. Carrier frequencies must be spaced at least as far a part as the maximum frequency of the encoded data.

[Mark]

Don't know about the technical aspects, but since the FCC has already effectively turned control of the airwaves over to global corporations, the fellow's original point is moot anyway.

[shecky]

The article treats interference both as a technological issue and a metaphor. My guess is that you should ignore the interference as a metaphor, as I think it's confusing enough to obscure the whole point of the article.

What he's saying is the current practice of reserving huge chunks of spectrum to AM/FM radio/broadcast Television is outmoded and the the radio spectrum should be treated more like the internet, that is, largely unregulated, where consumers eventually decide which hardware will "set the standard" for broadcasting. The presumption is that digital broadcasting techniques, spread spectrum and such, are much more efficient uses of radio spectrum, allowing much more information to be broadcast than traditional analog signals.

It seems that , ideally, he advocates access to the airwaves by everyone, not just the big broadcasting corporations as is the current paradigm. This doesn't mean anyone can run a TV station accessable by everyone. Rather, this means you can run a station accessable by anyone with the equipment capable of receiving your station.

Technologically, I think it's a bit of a assumption to make, that the radio spectrum is almost a infinite resource. Seems like interference can still be a problem if the sheer volume of digital traffic is too high. You just wouldn't call it "interference" anymore. You'd probably call it a "slow connection" or something. Kinda like if you were to download download twenty mp3s at a time over a modem connection, you'd likely experience a overall traffic slowdown and that streaming video would probably get pretty jerky.

It seems another assumption is that competing broadcast protocols wouldn't interfere with one another. There's no way of knowing this before hand. Certainly, every radio and TV used now would be not only obsolete but probably unusable if the airwaves were filled with enough traffic of this sort. And who knows how it would work out in the future.

The analogy that the market will decide (such as in the computer world) lags a bit. This isn't quite like users deciding to abandon the AppleII when better technology comes along. This is more like if all the houses on your block shared one modem line. It's possible to run several communication protocols simultaneously over that line. The unregulated part simply means you could run IP, peer to peer, AOL, whatever you like. But you must still use the same modem line.

While I kinda like the idea of the radio spectrum as more of a free for all, I'm pretty cynical about such a scheme actually turning into something I like. I suspect it'll be like cable TV, where you have 100+ channels of crap to watch.

[scotth]

Shecky,

Why does it sound like you know even less about what can be done with RF than the guy who's rediculous pronouncements were originally quoted?

[shecky]

Originally posted by scotth
Shecky,

Why does it sound like you know even less about what can be done with RF than the guy who's rediculous pronouncements were originally quoted?

Specifics?

[scotth]

Originally posted by shecky


Specifics?

Seems like interference can still be a problem if the sheer volume of digital traffic is too high.

If more than one signal is broadcast on the same frequency in the same location, it is 100% likely that at most one of the signals would be receivable, and then only if it completely swamps the others because it is of much greater power.

You just wouldn't call it "interference" anymore. You'd probably call it a "slow connection" or something. Kinda like if you were to download download twenty mp3s at a time over a modem connection, you'd likely experience a overall traffic slowdown and that streaming video would probably get pretty jerky. It would be interference, precisely. It wouldn't be a slow connection, it would a completely unfunctioning one.

It seems another assumption is that competing broadcast protocols wouldn't interfere with one another. There's no way of knowing this before hand. Certainly, every radio and TV used now would be not only obsolete but probably unusable if the airwaves were filled with enough traffic of this sort. And who knows how it would work out in the future.


Protocols are completely and utterly irrelevant. During interference, the original carrier signal cannot be determined. You must accurately receive the carrier before you can even think about extracting data from it.

The analogy that the market will decide (such as in the computer world) lags a bit. This isn't quite like users deciding to abandon the AppleII when better technology comes along. This is more like if all the houses on your block shared one modem line. It's possible to run several communication protocols simultaneously over that line. The unregulated part simply means you could run IP, peer to peer, AOL, whatever you like. But you must still use the same modem line.


Again, complete irrelevant. What he is proposing is much more like trying to connect two modems to the same phone line and having them both "online" at the same time. The comparison isn't perfect, but it much closer than what you are saying.

[shanek]

Originally posted by Mark
Don't know about the technical aspects, but since the FCC has already effectively turned control of the airwaves over to global corporations, the fellow's original point is moot anyway.

Oh? The FCC won't let Dish Network and DirecTV voluntarily share their spectrum space with each other because it's "collusion." They wouldn't let the two companies merge because one company with that much spectrum would be a "monopoly." And they won't give the two companies more spectrum because that would be "anticompetitive."

Seems to me like the government is still very much in control.

[shanek]

Originally posted by scotth
Again, complete irrelevant. What he is proposing is much more like trying to connect two modems to the same phone line and having them both "online" at the same time. The comparison isn't perfect, but it much closer than what you are saying.

Well, drawing from my knowledge of computer networks, I do know that an almost infinite number of computers can theorietically be transmitting along one line simultaneously, but there needs to be a common method of media access—contention, token passing, etc. And everything on the line has to agree with it. But that assumes that there's nothing else on the line interfering with the signal.

From what I gathered reading the article, it seems like he was saying there was some sort of magic number that would allow you to encode data onto a frequency range, and despite the level of interference anyone else who knew the magic number could strip off everything else and end up with the original transmission. But I don't know enough about the physics of it to know if that's even remotely feasible.

[Mark]

Seems to me like the government is still very much in control.

I would say the government is slightly in control. When one company is allowed to own hundreds of broadcasting stations, it creates the horrible, bland, state of broadcasting today. The FCC will, of course, go ballistic if someone says something they don't approve of...that's REALLY dangerous. :rolleyes:

[scotth]

Originally posted by shanek


Well, drawing from my knowledge of computer networks, I do know that an almost infinite number of computers can theorietically be transmitting along one line simultaneously, but there needs to be a common method of media access—contention, token passing, etc. And everything on the line has to agree with it. But that assumes that there's nothing else on the line interfering with the signal.


Your knowledge implies that you realize the point of the contention management is so that no two computers "talk" at the same time.

[shanek]

Originally posted by scotth


Your knowledge implies that you realize the point of the contention management is so that no two computers "talk" at the same time.

Of course. That results in a collision and the computers can't sort out what's what, although they can determine who took part in the collision and needs to retransmit.

From what this guy is saying, the technology exists or will soon exist such that the computers won't have to worry about collisions. But every single media access method I'm aware of depends on either detecing or avoiding collisions.

[shecky]

Originally posted by scotth




If more than one signal is broadcast on the same frequency in the same location, it is 100% likely that at most one of the signals would be receivable, and then only if it completely swamps the others because it is of much greater power.

This is true for traditional analog signals, which is one reason the rf spectrum is so rigidly divided. Not necessarily true for digital broadcasting techniques, where a transmitter may not need, say, 50kHz of relatively clear contiguous spectrum to broadcast clearly. Instead, the transmitter may still need 50kHz of bandwidth, but chops up the broadcast signal over a much broader range, perhaps 50MHz, and it's up to the "smart" receiver to figure out where all those little bits and pieces are and reassemble them into meaningful data.

Originally posted by scotth
It would be interference, precisely. It wouldn't be a slow connection, it would a completely unfunctioning one.

The analogy to a "slow connection" is that interference wouldn't be manifested by snow or ghost images (for example of a video image), but probably more like a jerky image, like what you might encounter watching streaming video over a phone connection on your PC. If your computer is on a network with lots of traffic, the connection doesn't necessarily stop functioning. It just functions poorly.

Originally posted by scotth

Protocols are completely and utterly irrelevant. During interference, the original carrier signal cannot be determined. You must accurately receive the carrier before you can even think about extracting data from it.

Not true. Protocols and broadcast modes are very relevant. The FM protocol made it much easier to broadcast static-free audio over fifty years ago (at the expense of rf bandwidth), no doubt current and future digital modes and protocols will allow even more efficient and higher fidelity signals to share the same spectrum.


Originally posted by scotth

Again, complete irrelevant. What he is proposing is much more like trying to connect two modems to the same phone line and having them both "online" at the same time. The comparison isn't perfect, but it much closer than what you are saying.

You seem to misunderstand what I said. This sounds like what I said when I wrote "This is more like if all the houses on your block shared one modem line." Seems we're not in disagreement about this very much.

This is where I find fault with the original article. It seems to advocate a "free-for-all" rf spectrum based on the assumption that the rf spectrum is a near infinite resource. On the internet, speed bottlenecks are relatively easily fixed by simply adding more wires, fibers, etc to accomodate the increased traffic. With RF, you simply cannot add more spectrum when you run out. Instead, you have to make better use of what you've got.

[scotth]

Originally posted by shanek


Of course. That results in a collision and the computers can't sort out what's what, although they can determine who took part in the collision and needs to retransmit.

From what this guy is saying, the technology exists or will soon exist such that the computers won't have to worry about collisions. But every single media access method I'm aware of depends on either detecing or avoiding collisions.

What he was saying is that if the data signal is carried by RF there is no such thing as collision. Or that receivers can be built that can read two more colliding signals at once and sort them out.

By his comments about todays receivers, he clearly doesn't know how they work. His comments about the not running our of the color green indicates he doesn't understand EM radiation and how it can be used to carry data at all.

By the formula provided by stimpy, it is clearly impossible to do what he implies unless he significantly overturns the laws of physics as they relate to light (EM in general).

This is not likely.

[scotth]

Originally posted by shecky

This is true for traditional analog signals, which is one reason the rf spectrum is so rigidly divided. Not necessarily true for digital broadcasting techniques, where a transmitter may not need, say, 50kHz of relatively clear contiguous spectrum to broadcast clearly. Instead, the transmitter may still need 50kHz of bandwidth, but chops up the broadcast signal over a much broader range, perhaps 50MHz, and it's up to the "smart" receiver to figure out where all those little bits and pieces are and reassemble them into meaningful data.
If 50kHz of bandwidth is needed, at minimum 50kHz of spectrum must be used. Breaking it up into pieces does not make it less than 50kHz.

The analogy to a "slow connection" is that interference wouldn't be manifested by snow or ghost images (for example of a video image), but probably more like a jerky image, like what you might encounter watching streaming video over a phone connection on your PC. If your computer is on a network with lots of traffic, the connection doesn't necessarily stop functioning. It just functions poorly.

You miss the point, if two (or more) RF tranmissions overlap in frequency, that data is unrecoverable, forever, period. He is implying otherwise. That is the only way to make the spectrum a limitless resource.

Not true. Protocols and broadcast modes are very relevant. The FM protocol made it much easier to broadcast static-free audio over fifty years ago (at the expense of rf bandwidth), no doubt current and future digital modes and protocols will allow even more efficient and higher fidelity signals to share the same spectrum.

If two signals overlap in frequency, you will never get far enough along to figure out what the encoding protocol (modulation) is. You will have lost the carrier from which to try to decode.


You seem to misunderstand what I said. This sounds like what I said when I wrote "This is more like if all the houses on your block shared one modem line." Seems we're not in disagreement about this very much.

This is where I find fault with the original article. It seems to advocate a "free-for-all" rf spectrum based on the assumption that the rf spectrum is a near infinite resource. On the internet, speed bottlenecks are relatively easily fixed by simply adding more wires, fibers, etc to accomodate the increased traffic. With RF, you simply cannot add more spectrum when you run out. Instead, you have to make better use of what you've got.

I would probably agree with your analysis of the desirability of using the spectrum as he suggests. Rather, I would agree if I didn't think it mute point. There is no reason to think it could be used that way except his word.

Why should we think he is on to something in a field he is not familiar with, that is directly at odds with all the experts in that field?

[Solitaire]

Originally posted by shanek
This is a new Salon article about a guy who basically says that bandwidth over radio waves is unlimited, and it's our outdated technology that's the limiting factor. He calls for the ending of government ownership of radio waves in favor of "smart" radio receivers that can somehow distinguish one signal from another.

Any idea how this is supposed to work? Is it a good idea, or just woo-woo stuff?

It's a bit junky.

One possibility. Divide the radio band into a series of spread specturm
channels. Have each smart radio know the other smart radios around it and
act as node in a network. When the user wants a resource from a distant
smart radio that's out of range, a series of smart radios along the path act
as a series of relays passing the signal from one point to the other. In a way
you can reduce the power requirements, but it will still cause problems for
standard radios.

The problem of capacity. (http://www.cs.ucl.ac.uk/staff/S.Bhatti/D51-notes/node6.html
)

Anther and very expensive possibililty, one can use a phased array apply
a bit of optical interferometery to cancil out all the signals you don't want
to tune into the one you do. It's used in astronomy to look for dim light
sources next to bright ones.

[shecky]

Originally posted by scotth
If 50kHz of bandwidth is needed, at minimum 50kHz of spectrum must be used. Breaking it up into pieces does not make it less than 50kHz.

Which is why I specified 50 kHz of contiguous spectrum. A digital broadcast may spread this bit of badnwidth over a much wider range.

Originally posted by scotth
You miss the point, if two (or more) RF tranmissions overlap in frequency, that data is unrecoverable, forever, period. He is implying otherwise. That is the only way to make the spectrum a limitless resource.

You're still thinking in terms of analog signals. Broadcast modes such as spread spectrum don't really work this way. Here's some background on this:
http://www.sss-mag.com/ss.html
http://www.kmj.com/proxim/pxhist.html

Originally posted by scotth

If two signals overlap in frequency, you will never get far enough along to figure out what the encoding protocol (modulation) is. You will have lost the carrier from which to try to decode.

Once again, this is more true with analog signals than digital. But because digital modes promise more efficiency doesn't mean it's unlimited efficiency. Eventually if there's enough traffic, performance will degrade. This is where the article looks a bit woo woo.

Originally posted by scotth

I would probably agree with your analysis of the desirability of using the spectrum as he suggests. Rather, I would agree if I didn't think it mute point. There is no reason to think it could be used that way except his word.

I'd agree with this. Just because such a paradigm shift would have the potential to become a great egalitarian endeavor doesn't mean it will. I tend to think it would simply mean large corporations would squeeze out the independent guys, and we'd still end up with more crap media. Not to mention the tons of electronic equipment that would become obsolete as quickly as PCs, every radio and TV would basically be glorified computers, etc. Sometimes a dumb appliance is the perfect tool for a job. And nothing to stop the next M$oft from deciding it's new technology is deserving enough to hog all the spectrum and consumers will just have to adjust, etc.

Originally posted by scotth

Why should we think he is on to something in a field he is not familiar with, that is directly at odds with all the experts in that field?

This may be the biggest problem with the article. This guy comes off as a fervent evangelist making a claim of "interference is a myth" only to totally obfuscate the issue of deregulation.

[scotth]

Originally posted by shecky


Which is why I specified 50 kHz of contiguous spectrum. A digital broadcast may spread this bit of badnwidth over a much wider range.

You're still thinking in terms of analog signals. Broadcast modes such as spread spectrum don't really work this way. Here's some background on this:
http://www.sss-mag.com/ss.html
http://www.kmj.com/proxim/pxhist.html

Once again, this is more true with analog signals than digital. But because digital modes promise more efficiency doesn't mean it's unlimited efficiency. Eventually if there's enough traffic, performance will degrade. This is where the article looks a bit woo woo.


I'll just go in order instead of interlacing my answers.... I am getting lazy.

This still doesn't get around using 50kHz of spectrum.

Why do you assume I am thinking in analog terms. I am quite familiar with digital signal encoding. The point is that if you cannot accurate receive the carrier freqency, analog or digital encoding is meaningless. There would be nothing to decode.

Ignoring what could be done with the technology if it existed... He specifically says that it is possible to develop better receiver technology that would get around interference problems. Digital and analog receivers are identical until you get to the demodulator sections. If two carrier waves are stepping on each other, the demod section will be fed unusable garbage.

[scotth]

Originally posted by John Lockard

It's a bit junky.

One possibility. Divide the radio band into a series of spread specturm
channels. Have each smart radio know the other smart radios around it and
act as node in a network. When the user wants a resource from a distant
smart radio that's out of range, a series of smart radios along the path act
as a series of relays passing the signal from one point to the other. In a way
you can reduce the power requirements, but it will still cause problems for
standard radios.

The problem of capacity. (http://www.cs.ucl.ac.uk/staff/S.Bhatti/D51-notes/node6.html
)

Anther and very expensive possibililty, one can use a phased array apply
a bit of optical interferometery to cancil out all the signals you don't want
to tune into the one you do. It's used in astronomy to look for dim light
sources next to bright ones.

Well, at least the phased array idea would get us somewhere. But it is not a receiver technology. It is a way of making the antenna only sensitive to signal coming from a very specific angular direction. A parabolic antenna would be similarly useful.

This is a quite established old school way of working and shouldn't be confused with support for the original assertions.

btw... phase array has been in use in RF comm for a long time. Two of the radar for which I am a tech on use phase array antennas.

[MRC_Hans]

OK, I finally got around to reading the whole article. This guy's point is that interference does not happen till you treat the received signal in a way that mixes the incoming signal. Which does essentially not happen till you introduce an unlinear element. And basically, he is right.

Where he goes wrong is when he jumps to the conclusion that this opens unlimited possibilities of communication. We have the classic "conservative government/gready bit corporations" strawman explanation on why this has not been utilized. This is obvious crap, as the potential money to be earned in this would surely have made it reality long ago, had it been feasible-.

Now for the technicalities:

In the present wireless scenario, we mainly divide signals in the frequency domain. That is, we allocate different frequency (wavelength) slots to each signal, and this puts a limit to how many signals we can allow in a given area; once the useful spectrum is allocated in an area, we cant allow more, for fear of interference.

The claim is that interference does not happen except in the receiving equipment. This claim is basically true, although not very useful. What the article suggests is that we use the amplitude domain. I'll try to explain this:

Suppose we have two transmitters using the same frequency in the same area. In the air, those signals can coexist with no interference: The wave systems pass through each other without any effect. If we want to receive them, we put up an antenna. The electromagnetic waves will cause an electric current to run in the antenna. This signal is a combination of the two signals, and theoretically, the signals are still perfectly intact. Using sufficiently advanced signal processing we should be able to detect each individual signal by evaluating the content.

However, receiving two signals take up twice the amplitude of one signal, and here is the crunch: As long as all signals arrive with equal amplitudes, we might be able to separate tens, perhaps hundreds of signals, provided we use sufficiently advanced signal processing. But signals do NOT arrive at equal amplitudes. A close-by transmitter will be represented by a signal that is thousands of times stronger than the signal from a distant source.

Forget about digital; we are in solid analog here: In order to preserve the integrity of different signals with the same frequence, our analog circuitry must be able to handle the summed amplitudes of all received signals linearly, and in state of the art analog systems, the whole available amplitude domain is already used up by the dynamic range neccessary to handle signals from transmitters of different distance and power. A modern receiver can handle signals with a difference in power of about six orders of magnitude, and even this is accomplished only by automatically adjusting the gain of the circuitry. Thus the receiver can only, at best, handle perhaps two or three transmitters of equal amplitude.

So the idea in the article, which comes from a no doubt very competent digital engineer, will not work in the analog world. And, as it happens, the real world is largely analog.

Hans

[garys_2k]

Originally posted by MRC_Hans
Suppose we have two transmitters using the same frequency in the same area. In the air, those signals can coexist with no interference: The wave systems pass through each other without any effect. If we want to receive them, we put up an antenna. The electromagnetic waves will cause an electric current to run in the antenna. This signal is a combination of the two signals, and theoretically, the signals are still perfectly intact. Using sufficiently advanced signal processing we should be able to detect each individual signal by evaluating the content.

No, I don't think so.

Say you are receiving two transmissions on the same frequency, one from station A, the other from B. A is transmitting a 1 kHz tone and B is transmitting a 1.5 kHz tone. These signals mix in your antenna.

How would any receiver "know" if the 1 kHz tone was supposed to be assigned to A or B? I submit that there is no totally unambiguous way for a it to do so.

[shanek]

Originally posted by shecky
This is true for traditional analog signals, which is one reason the rf spectrum is so rigidly divided. Not necessarily true for digital broadcasting techniques, where a transmitter may not need, say, 50kHz of relatively clear contiguous spectrum to broadcast clearly. Instead, the transmitter may still need 50kHz of bandwidth, but chops up the broadcast signal over a much broader range, perhaps 50MHz, and it's up to the "smart" receiver to figure out where all those little bits and pieces are and reassemble them into meaningful data.

But you would still have the problem of interference. If something interfered with part of the signal, that part of the digital stream wouldn't make it through. In wireless protocols this is usually taken care of by broadcasting the data 2 or more times in case part of it gets interfered with. But, of course, that cuts your throughput in half or less.

Maybe he's figuring that smart transmitters and receivers would be able to detect unused frequency ranges and transmit in the gaps, and that there'd be no way that everyone in a single area could use the entirety of the radio spectrum.

[shanek]

Originally posted by MRC_Hans
However, receiving two signals take up twice the amplitude of one signal, and here is the crunch: As long as all signals arrive with equal amplitudes, we might be able to separate tens, perhaps hundreds of signals, provided we use sufficiently advanced signal processing. But signals do NOT arrive at equal amplitudes. A close-by transmitter will be represented by a signal that is thousands of times stronger than the signal from a distant source.

I dunno...it seems to me if it were that simple there'd already be a form of multiplexing that takes advantage of it. And since the multiplexing would occur from the same transmitter, you wouldn't have the problem of amplitude loss because they'd all go down together.

[MRC_Hans]

Originally posted by garys_2k

No, I don't think so.

Say you are receiving two transmissions on the same frequency, one from station A, the other from B. A is transmitting a 1 kHz tone and B is transmitting a 1.5 kHz tone. These signals mix in your antenna.

How would any receiver "know" if the 1 kHz tone was supposed to be assigned to A or B? I submit that there is no totally unambiguous way for a it to do so.

First of all we have to be a little careful about the word "mix". Electronic signals only mix if they interact. The two signals will both be present at the antenna. The only way to take them apart will be by evaluating the information in them. But that is in principle the same thing that happens on a digital net: When your computer receives a packet, it has no way of knowing where it came from, till it decodes it and read the "from" info.

As of the two tones you mention, the only thing that might tell us which transmitter was which would be if we knew which one transmitted which signal. But then, those signals dont have much information in them.

Hans

[MRC_Hans]

Originally posted by shanek


I dunno...it seems to me if it were that simple there'd already be a form of multiplexing that takes advantage of it. And since the multiplexing would occur from the same transmitter, you wouldn't have the problem of amplitude loss because they'd all go down together.

Oh, but there is. Old-fashioned telex lines used sub-carrier multiplex; each telex channel had two frequencies, one for 1 and another for 0, then they were all shucked into the same cable and sorted aout at the receiving end, up to a dozen of them. But, thats just a few signals, and very narrow-band (telex is 50 bits/sec). This fellow talks "unlimited capacity", and he DOES talk about many transmitters sharing the same frequency channel.

Hans

[shanek]

Originally posted by MRC_Hans
First of all we have to be a little careful about the word "mix". Electronic signals only mix if they interact. The two signals will both be present at the antenna. The only way to take them apart will be by evaluating the information in them. But that is in principle the same thing that happens on a digital net: When your computer receives a packet, it has no way of knowing where it came from, till it decodes it and read the "from" info.

There's a problem with this logic. According to the OSI model, examining the identification of the sender is a logical function of the Data-Link Layer (Media Access Control (MAC) sublayer in the IEEE model). The interference of the signal we're talking about takes place on the Physical layer. If the Physical layer isn't getting the right ones and zeroes out of the signal, then nothing at the MAC sublayer can fix it.

So, let's say we've defined a 1KHz tone as a "1" and a 1.5KHz tone as a "0" on the Physical layer. It could tell that there was both a 1 and a 0 on the frequency, but it would be unable to distinguish who was sending the 1 and who was sending the 0 because it hasn't gotten there yet. It needs a whole packet of 1s and 0s to send up to the Data-Link layer before it can tell who sent what.

There would need to be some way of encoding the 1s and 0s on the signals while still preserving the above distinction.

[shanek]

Originally posted by MRC_Hans
Oh, but there is. Old-fashioned telex lines used sub-carrier multiplex; each telex channel had two frequencies, one for 1 and another for 0, then they were all shucked into the same cable and sorted aout at the receiving end, up to a dozen of them.

But that's purely point-to-point, right?

[garys_2k]

Originally posted by MRC_Hans


First of all we have to be a little careful about the word "mix". Electronic signals only mix if they interact. The two signals will both be present at the antenna. The only way to take them apart will be by evaluating the information in them.
The signal do mix in the antenna, their electric waveforms sum and form sideband signals. Just being in the same conductor will cause them to interact. I submit that there is no way to unambiguously separate them, ever, once they're mixed like that.

[Walter Wayne]

Originally posted by garys_2k

The signal do mix in the antenna, their electric waveforms sum and form sideband signals. Just being in the same conductor will cause them to interact. I submit that there is no way to unambiguously separate them, ever, once they're mixed like that. The linearity of metals and those signal levels is high. 99 point bunch of 9s percent high. They will not form sidebands.

Having said that, distinguishing between the signals would be difficult. Using an array could be used to resolve where the signals are coming from, thus if there is significant angular seperation of source A and source B one could differentiate. As mentioned before distinguishing between signals is usually done within the signal. Since tones aren't much of a signal (no information) why would one want to distinguish anyways.

Note that one can overlap signal in frequency and time in analog, however doing so from seperate sources would be difficult. One would also be limited in the number of overlapped signals so this would not give one infinite through put.

Walt

[MRC_Hans]

Originally posted by garys_2k

The signal do mix in the antenna, their electric waveforms sum and form sideband signals. Just being in the same conductor will cause them to interact. I submit that there is no way to unambiguously separate them, ever, once they're mixed like that. No, they won't. There are, of course, practical limitations, but in principle any number of signals can share the same conductor without interacting. Only when a non-liear component is added, do they interact. And only in that case side-bands are formed.

The problem, apart from the signal processing problem, is amplitude range.

When you recieve a distant station, it will generate, say, 10 microvolts of antenna signal. To process that signal, you need to amplify it to get a signal in the order of 1 volt. This requires 100,000 times amplification. Thats no problem, but the signal from a closeby transmitter may easily be 0.1 volts. With the same amplification, you get an output of 10,000 volts :eek:
. The conventional reciever takes care of the problem by AGC, Automatic Gain Control: When it receives the powerfull signal, it simply lowers the gain accordingly (in this case to 10).

But the "interference myth" receiver cannot afford to do this, because the weak signal would be lost. So we would need some signal processing circuitry that could handle different signal strengths within 5-6 orders of magnitude simultaneously, and thats way beyond anything we can do at present. I hesitate to call anything impossible, but ---


Shanek:

But that's purely point-to-point, right? Right. and wires only. I can't imagine how it could function multipoint and wireless.

Hans

[Agammamon]

Originally posted by scotth


This gets you no where.

When information is encoded on a carrier signal (modulated) it MUST spread its freqency spectrum.

The only way to infinitely divide a spectrum up would be to put no data on each carrier wave.

As soon as data is encode, the carrier is no longer monochromatic. Carrier frequencies must be spaced at least as far a part as the maximum frequency of the encoded data.

And as the receiver's ability to discriminate between 2 frequencies increases, you can use a narrower and narrower band. In a (obviously hypothetical) perfect discriminator, you can divide you spectrum into arbitrarily small bands. If the receiver can tell the difference between say 100MHz and 100.000000. . . 1 MHz then the frequency spread used to encode data can be smaller and smaller.

[scotth]

Originally posted by Agammamon


And as the receiver's ability to discriminate between 2 frequencies increases, you can use a narrower and narrower band. In a (obviously hypothetical) perfect discriminator, you can divide you spectrum into arbitrarily small bands. If the receiver can tell the difference between say 100MHz and 100.000000. . . 1 MHz then the frequency spread used to encode data can be smaller and smaller.

Doesn't work that way.

The carrier is only its exact center frequency when no data is being encoded. At maximum, the modulation rate can only be as high as the channel spacing. If you only have a 1Hz channel spacing, at most you can carry a 1Hz data signal.

If two adjacent channels are gonna carry 100kHz of info, they must be AT LEAST 100kHz apart.

[MRC_Hans]

Right. Or to make it even more plain: You need at least as much bandwidth as the information speed you want to transfer. This is a physical law, and there is no way around it. If you want to transfer 1000 bits/sec, you need at least (with perfect equipment) 1000Hz bandwidth, etc.

Hans

[shanek]

Originally posted by MRC_Hans
Right. Or to make it even more plain: You need at least as much bandwidth as the information speed you want to transfer. This is a physical law, and there is no way around it. If you want to transfer 1000 bits/sec, you need at least (with perfect equipment) 1000Hz bandwidth, etc.

Actually, that hasn't been the case in years. There are methods of encoding more than one bit on a single oscillation.

[MRC_Hans]

Shanek:
Nope. We have come very far in data compression, but that just means we need to transfer less data for a given information. But I think you are talking about multi-level encoding. That is essentially what the guy in the article is talking about. But it still requires bandwith according to the amount of data transmitted. Multilevel encoding has its virtues where bandwith in the transmission channel is not the limiting factor, like in some optical links.

Hans

[scotth]

Originally posted by shanek


Actually, that hasn't been the case in years. There are methods of encoding more than one bit on a single oscillation.

This is quite false.

[garys_2k]

Originally posted by Walter Wayne
The linearity of metals and those signal levels is high. 99 point bunch of 9s percent high. They will not form sidebands.

OK, I'll stick to subjects I'm more familiar with, thanks! :)
Having said that, distinguishing between the signals would be difficult. Using an array could be used to resolve where the signals are coming from, thus if there is significant angular seperation of source A and source B one could differentiate. As mentioned before distinguishing between signals is usually done within the signal. Since tones aren't much of a signal (no information) why would one want to distinguish anyways.

Well, one tone could be a zero and the other a one, so telling them apart could be the most important thing to do wrt recovering information.

Note that one can overlap signal in frequency and time in analog, however doing so from seperate sources would be difficult. One would also be limited in the number of overlapped signals so this would not give one infinite through put.

Walt
I agree. Saying we have "unlimited" bandwidth seems just wrong.

[shanek]

Originally posted by MRC_Hans
But I think you are talking about multi-level encoding.

Yes, I am.

But it still requires bandwith according to the amount of data transmitted.

Of course, but it's no longer a 1:1 ratio.

Multilevel encoding has its virtues where bandwith in the transmission channel is not the limiting factor, like in some optical links.

It's even used in some cases where it is, like dial-up lines. We wouldn't have 56Kbps modems without it.

[shanek]

Originally posted by scotth


This is quite false.

No, it isn't. You can use multilevel signaling. You could, for example, combine amplitude-shift keying with phase-shift keying, and this would put two bits of data in each oscillation. There are other methods. It's done all the time.

[Edited to fix strange wording and an unfortunate typo]

[Walter Wayne]

If one wants to look at digital data transmission,in perfect receivers (with ideal filters), the limit is a function of noise and accepted bit error rate.

For example. Lets assume your receiver can differentiate 1 mV levels up to +/-1 V, an analog signal could take anyone of 2000 value (remember that a digital signal is essentially an interpretation of an analog signal). Which is almost 11 bits/symbol. However, keeping the incoming noise to this "perfect reciever" low enough that bit error rate wasn't ridiculous isn't possible on a realistic channel.

Thats before taken into account non-idealities of the reciever of course. But in a mathematical sense, I don't know of any limit on the number of bits described by a single symbol.

Walt

[BobM]

No, it isn't. You can use multilevel signaling. You could, for example, combine amplitude-shift keying with phase-**** keying, and this would put two bits of data in each oscillation. There are other methods. It's done all the time.

[Edited to fix strange wording]Err.. that edit may have fixed the wording but it failed to fix some of the strange words.

[jj]

The guy said But surely there must be some limit. "Actually, there isn't. Information isn't like a physical thing that has to have an outer limit even if we don't yet know what that limit is.

Err, yes, there is a limit. It's known. It's called the "Shannon Bound".

There is exactly and indeed a limit to the amount of information that can be sent in the channel. Up to that limit, with unlimited time to decode the information, you can eventually get the entire message without corruption, using appropriate methods.

Above that limit, the message is lost.

This mathematical fact is shown by digital systems. I remember one I worked on in the 1990's, where at -100some dB, where the first decimal place in the attenuation was .5 dB, the system would show 1 error per hour. at .6dB, it was one error per minute. At .7 dB, the signal was unrecoverable, entirely.

This is somewhat intuitively boggling. Wozencraft and Jacob's classical book in Communications is a good place to start. Shannon's paper isn't bad, either.

It doesn't matter if you use FEC, multilevel, etc, etc. The Shannon bound, to the present, for very good mathematical reasons, has stood firm.

[jj]

Originally posted by Agammamon


And as the receiver's ability to discriminate between 2 frequencies increases, you can use a narrower and narrower band. In a (obviously hypothetical) perfect discriminator, you can divide you spectrum into arbitrarily small bands. If the receiver can tell the difference between say 100MHz and 100.000000. . . 1 MHz then the frequency spread used to encode data can be smaller and smaller.

But that doesn't help. If you have very narrow bands, you can send very little information.

The integral across bandwidth of log2(snr) is the bound. It's sorta like the speed of light or the Heisenburg Uncertainty.

[jj]

Originally posted by MRC_Hans
Right. Or to make it even more plain: You need at least as much bandwidth as the information speed you want to transfer. This is a physical law, and there is no way around it. If you want to transfer 1000 bits/sec, you need at least (with perfect equipment) 1000Hz bandwidth, etc.

Hans

NO!

If you have a constant SNR in your channel, then you can send, per Hz, log2(snr) bits/second.

If log2(snr) is 10, you can send 1000bits/second in 100Hz bandwidth.

If log2(snr) is .1, you will need 10kHz bandwidth.

It's called the "Channel Capacity Theorem" or "Shannon Bound".

Originally posted by scotth


This is quite false.

No, actually he's right. Things like Quadrature Amplitude Modulation let you do this and get fairly close to the maximum predicted by Shannon's Law.

edited to add:

Oops, someone beat me to it.

[jj]

Originally posted by MRC_Hans
Shanek:
Nope. We have come very far in data compression, but that just means we need to transfer less data for a given information. But I think you are talking about multi-level encoding. That is essentially what the guy in the article is talking about. But it still requires bandwith according to the amount of data transmitted. Multilevel encoding has its virtues where bandwith in the transmission channel is not the limiting factor, like in some optical links.

Hans

Ok, look.

I know just a WEE bit about data compression, and a WEE bit about modems and transmission systems.

Multilevel encoding is OLD OLD OLD. PCM is such a form, it can be reversed, you know, to turn bits into a PCM signal. It can work either way, just for instance. PCM is way old. Multilevel modems are not new, please.

log2(snr) is the bits/hz you can get in a given channel.

If you have a really good SNR (say 2^20), you can send 20 bits/symbol, and get 20 bits/Hz.

If your SNR is really really bad, you get nada.

[jj]

Originally posted by scotth


This is quite false.

It's dead on true.

If I have a channel that can resolve 256 different levels, then I can send 8 bits on each symbol.

DOH, dude!

Imagine I am using both phase modulation and amplitude modulation in the following way. I define four different phase angles; 0, 90, 180 and 270, and four different amplitude levels.

This gives me 16 combinations, each of which can be transmitted in a single "baud". Now if we simply assign the binary numbers 0000 thru 1111 to these 16 combinations, bingo - we are sending 4 bits with each pulse.

[scotth]

Originally posted by jj


It's dead on true.

If I have a channel that can resolve 256 different levels, then I can send 8 bits on each symbol.

DOH, dude!

The more bits you encode per symbol, the less often you can switch symbols.

The faster the symbols switch, the broader the spectrum becomes.

[shanek]

Originally posted by BobM
Err.. that edit may have fixed the wording but it failed to fix some of the strange words.

Oops...thanks for catching that.

[jj]

Originally posted by scotth


The more bits you encode per symbol, the less often you can switch symbols.
WRONG, unless you must increase the symbol time to remain under the Shannon Bound or under your equalizer's ability to cope with the channel. In general, you do not have to slow symbols down UNTIL you meet one problem or the other.

Look up "Viterbi", for instance, and "channel equalization".



The faster the symbols switch, the broader the spectrum becomes.

RIGHT.

Look up "equalizer" in a communications text.

The fact is that the better the SNR, the more bits/second/Hz you can send. You do that by, in some way or other, using more "levels". QAM, PQAM, etc, are all ways to do so. The trick is to use a constellation that has maximal minimum distance, if you want best error performance. There are entire books written on how to do that.

All of them do NOT argue with the idea that better SNR means more bits.

If your assertion that we MUST slow down symbol rate when adding levels was true in all cases, then the Shannon Bound would be terribly, horribly wrong. It's not.

And all of them handle the issue of equalization due to limited bandwidth. Most of them talk about Viterbi decoding, using the knowledge of previous state to untangle intersymbol issues.

Adding levels to a signal DOES NOT INCREASE THE BANDWIDTH. You may have to equalize to get the settling time right, but that's something different. You may have to use a Viterbi decoder to untangle things, but that's the way of the world nowdays.

Now, there are modems that deliberately use long symbols, say things like ODFM and the like. There are arguments for and against, and like anything else, it all depends on the application. They present another way to approach the Shannon Bound, and there is nothing wrong with that.

Your denial, on the other hand, of the Shannon Bound, is something you really, really need to reconsider. I'm not speculating, here, I've done real-life work using the lot.

[scotth]

Originally posted by jj

Your denial, on the other hand, of the Shannon Bound, is something you really, really need to reconsider. I'm not speculating, here, I've done real-life work using the lot.

I didn't deny the shannon bound. Don't know where you got that one.

I'll hit some of the other items in a bit.

[jj]

Originally posted by scotth


I didn't deny the shannon bound. Don't know where you got that one.

I'll hit some of the other items in a bit.

When you assert that you MUST slow down symbol rate if you go to more levels, you are refuting the Shannon bound.

Viterbi decoding, channel equalizers, etc, are all methods that make it possible to NOT slow down the symbol rate. All are commonly used in the present day.

I can't even count how many kinds of Viterbi decoders, modifications, limitations, etc, I've seen, because I don't remember them all.

[MRC_Hans]

I think the crunch here is that those systems are for cable transmission (any kind of cable).

The restraints on bandwidth in a cable system are different from radio transmission systems.

In a cable system, the available BW is determined by several factors, like dielectric characteristics, reflections, losses, dispersion, resonances, transparancy windows (for optical conductors), etc. All of these do not impose any sharp cut-off, instead what we basically see is an increasingly poor S/N as frequency rises. This gives room for various encoding protocols that make it possible to cram more information through the channel that a bare BW figure might suggest.

The original question, however, was wireless transmissions. Here, the situation is much different. The medium (the "ether") has essentially unlimited bandwidth (I know the atmosphere has some opaque windows below and above the visual spectrum, but thats beside the point here). The problem is instead that if a transmission channel creates side-bands outside its allocated BW, it will create interference problems.

The side-band consideration puts very definite restrictions on usable BW. The fastest information rate within such a channel is a sine-wave of the same frequency as the channel width (assuming we use single-sideband transmission, SSB). You can encode this wave with a binary signal, leaving out periods. In essence you can have a bit per period, so with SSB, you should just be able to transmit BW bits/sec. But thats it. Phase shifting, extra levels, or what else you might think of, will generate additional side-bands, sending your signal out of bounds.

So in radio transmission, it is not a question of getting your side-bands through, but that you are not allowed to generate them at all.

The Shannon equation is rather rusty with me, I must admit, but the overall conclusion seems to make sense in a cable system; sidebands wille be attenuated for each order of overtone and it then becomes a question of when they sink into the noise floor.

Hans

[garys_2k]

So it seems that everyone is in agreement that:

- Bandwidth is not unlimited, in the literal sense, as the original article stated, and

- Interference in the harware layer, ie amplifiers and antennas, is not a "myth," but will continue to be a problem (as interference contributes to degraded S/n ratios).

Is that right?

[MRC_Hans]

Aye!
(except for the antenna bit, but nevermind)

Hans

[jj]

Originally posted by MRC_Hans
I think the crunch here is that those systems are for cable transmission (any kind of cable).

The restraints on bandwidth in a cable system are different from radio transmission systems.

The problem is instead that if a transmission channel creates side-bands outside its allocated BW, it will create interference problems.



Err, the Shannon Bound holds for both. While the problems in cables vs. atmospheric transmission are different, the basics are quite the same.

Neither is anything like infinite.

[MRC_Hans]

I must read up on the theory. I am very curious to see how you can transmit a level change without creating a sideband.

Hans

[jj]

Originally posted by MRC_Hans
I must read up on the theory. I am very curious to see how you can transmit a level change without creating a sideband.

Hans

You can't.

But the RATE of level changes determines the sideband bandwidth.

You can have multiple levels, and still have the same BANDWIDTH. The number of different levels does not determine the bandwidth, the rate of change in symbols does.

Hmm. How technical do you want to get here? I can accomodate you on this one. :D

[MRC_Hans]

As technical as you will ;)

So we agree that the rate of level changes determines BW.

No matter how many levels you have to choose from, you can only send one level at a time. Obviously, giving different levels different values will help you assemble a given value faster; basically, it relieves you of transmitting zeroes.

Hans

[jj]

Originally posted by MRC_Hans
As technical as you will ;)

So we agree that the rate of level changes determines BW.

No matter how many levels you have to choose from, you can only send one level at a time. Obviously, giving different levels different values will help you assemble a given value faster; basically, it relieves you of transmitting zeroes.

Hans

Well, you seem to be getting it. Now, this is NOT, repeat NOT the best kind of modem, but it's easy to explain, so let's do it this way:

Let us send PCM samples at a "symbol rate". According to the Shannon/Nyquest theorem, we only need the same bandwidth as the symbol rate (although we will have to deal with intersymbol interference as a result, that is doable). So, let's say the signal to noise ratio is 16.0001:1, and the noise is constant, two-valued, at + or - a constant. (no noise like this exists, this is an example).

If we send a 16 level signal, we can JUST avoid having it changed by this noise. So we can send 16 levels. Into this 16 levels, we can encode 4 bits. So we can send 4 bits/symbol.

We can send as many symbols as we have bandwidth. So, the bit rate of this hypothetical, idealized for explaination's sake channel is 4 bits/Hz.

Did that sink in?

[MRC_Hans]

Did that sink in? No need to be condescending, I have worked with electronics almost as long as you, although I'm more of a generalist.

In an earlier post, you said:
If log2(snr) is 10, you can send 1000bits/second in 100Hz bandwidth.

I'll go with you on the 4 bits/Hz rate as practical (not bad going at that), but for each bit you want to add, you'll need to double the number of levels, as far as I can see. I'd like to see you get an open eye plot on much more than 16 levels in any practical channel.

But it does shed light on how they send 128Kb to my house through an ancient phone cable :)

Hans

[jj]

Originally posted by MRC_Hans
No need to be condescending, I have worked with electronics almost as long as you, although I'm more of a generalist.

In an earlier post, you said:


I'll go with you on the 4 bits/Hz rate as practical (not bad going at that), but for each bit you want to add, you'll need to double the number of levels, as far as I can see. I'd like to see you get an open eye plot on much more than 16 levels in any practical channel.

But it does shed light on how they send 128Kb to my house through an ancient phone cable :)

Hans

Well, in fact a 56k modem gets very close to 7 bits/Hz. Don't forget that the bandwidth is not 4kHz, it's 3.2kHz, so when you figure out how much of the actual "trained to" mulaw (or alaw) levels they can actually learn, they do awfully well.

I wasn't trying to be condescending, I was trying to find out if I was headed the right direction to explain this.

Now, ISDN (your 2B example above) uses a much wider bandwidth. By measuring the dispersion of the cable and equalizing it, they can use much more than a 4kHz or so bandwidth, subject to some interwire interference and some length limits beyond which nonlinearities in dispersion begin to accumulate to the point they are hard to eliminate.

DSL is even a more extreme example, of course, with even tighter loop length limits.

As to 10 bits/Hz, there are a FEW channels for which one might accomplish that, but you're very right, it's hard to get.

And, like I said, one other thing to remember is that the way I presented was an EXAMPLE. It is not the best way to implement a modem. Using QAM-like things, creating a "constellation" of levels that are as far apart as can be created for the necessary number of levels is how to go about it. You could sorta think of it as PCM on both the sine and cosine parts of the symbol, but that again is an approximation that isn't really valid.

And, of course when going through mulaw or alaw, the levels aren't uniform, since the transmission mechanism itself isn;'t uniform.

[jj]

Originally posted by garys_2k
So it seems that everyone is in agreement that:

- Bandwidth is not unlimited, in the literal sense, as the original article stated, and


Quite true.



- Interference in the harware layer, ie amplifiers and antennas, is not a "myth," but will continue to be a problem (as interference contributes to degraded S/n ratios).

Is that right?

Noise is a fundamental outcome of QM. It sets a bound on information transmission. It's not the only bound, as interference, other sources of noise, etc, all also bound information transfer.

The "unlimited" claim is simply bogus.