How can digital be more spectrum efficient than analog ?

glen herrmannsfeldt wrote in news:E7CdnVV8Ytd_mNrYnZ2dnUVZ snipped-for-privacy@comcast.com:

[ snip ]

Not speaking from practical experience, but in theory if the dam's connection to the power grid is "good enough", they have an "infinite bus" situation. In that case, the other generators and synchronous loads on the grid force the generator to stay synchronous in the short term. Over the longer term, if total load and total generation are out of balance, frequency varies but the whole grid varies and the generators stay synchronous, but at a slightly off frequency. (Of course if the power flows exceed the grid's capacity, lines fail and usually a blackout ensues.)

I was taught that system operators try hard to hold frequency within very tight tolerances, and especially to ensure that each day they have exactly 60*60*60*24 cycles (at least in the US, and excepting the old 25 Hz industrial and railroad systems) so all the old motor-driven electric clocks keep good time.

--
 Rich Wurth / Rumson, NJ

Sure, there's nothing inherent to the expression that makes it analog.

I don't believe he was. Perhaps we should let him speak for himself?

True.

I can't think of an analog system that uses an analog modulation scheme that takes advantage of both SNR and bandwidth to reach "capacity." Can you name such a scheme?

This is standard terminology. See, e.g., [proakiscomm] p.2, [sklar] p.1, etc.

@BOOK{proakiscomm, title = "{Digital Communications}", author = "John~G.~Proakis", publisher = "McGraw-Hill", edition = "fourth", year = "2001"} @BOOK{sklar, title = "{Digital Communications}", author = "{Bernard~Sklar}", publisher = "Prentice Hall P T R", edition = "second", year = "2001"}

--
%  Randy Yates                  % "My Shangri-la has gone away, fading like 
%% Fuquay-Varina, NC            %  the Beatles on 'Hey Jude'" 
%%% 919-577-9882                %  
%%%%            % 'Shangri-La', *A New World Record*, ELO
http://home.earthlink.net/~yatescr

I should have said, "... that makes it digital."

-- % Randy Yates % "Though you ride on the wheels of tomorrow, %% Fuquay-Varina, NC % you still wander the fields of your %%% 919-577-9882 % sorrow." %%%% % '21st Century Man', *Time*, ELO

And your statement above was nonsense.

He did speak for himself: "Assume you have an analog channel of only a few Hz in bandwidth, but it was noiseless and you had a billion-bit resolution DAC/ADC pair."

You statement is nonsense.

So nothing you've said so far makes any sense at all...

Every single one of them. They all have finite bandwidth, therefore capacity is determined by SNR just as much as by bandwidth.

So you *are* using standard terminology, not your own. In that case I can't see any validity to your point above. First, nothing that was said confuses the two, as you claimed. Second the "spirit" of the original question is not violated by the use of source coding to increase apparent throughput and/or reliability of a digital communications system, after all the OP

*specifically* mentioned it, though not by name...

--
Floyd L. Davidson            
Ukpeagvik (Barrow, Alaska)                         floyd@apaflo.com

Very interesting analogy. I'd not thought of that before.

Eric Jacobsen Minister of Algorithms, Intel Corp. My opinions may not be Intel's opinions.

There's a DAC in the modulator and an ADC in the demod for all DSL "modems". They both have analog filtering, amplification, and mixing (generally) between the DAC and ADC, and a cable or channel with "analog" characteristics.

Pretty much all of the devices that you claim use "digital signalling" do this.

You need to clarify what you mean.

Eric Jacobsen Minister of Algorithms, Intel Corp. My opinions may not be Intel's opinions.

I have mallets and hammers in various weights. "Impedance match" is a good way to explain why.

Jerry

--
        "The rights of the best of men are secured only as the
        rights of the vilest and most abhorrent are protected."
            - Chief Justice Charles Evans Hughes, 1927
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯

One of Shannon's main points, the one that led to the entire field of study that we now call "information theory", is that each channel has a "capacity" above which error-free information transmission is _always_ possible. It is not possible to achieve "capacity", or even approach it very closely, without "forward-error-correction bits".

The presence of "forward-error-correction bits", (aka, parity bits, overhead, redundancy, whatever you want to call them), reduce the raw information rate from an uncoded system using the same channel.

So I don't see any errors in John's post.

I do, however, see quite a few in yours.

Can you provide an example of a digitial satellite modem that doesn't use "forward-error-correction bits"? Can you name a satellite modem that has performance approaching "channel capacity" as Shannon described it? If so, do you know how it achieves that performance, i.e., what's different between it and other modems that don't do quite as well?

FWIW, I'm quite well aware of the earth station equipment used for satellite communications. I used to design satellite modems for a living, including their error correction systems (at another company).

Sigh.

I've recognized all of your incorrect statements. They're plentiful and hard to miss.

As I mentioned, I've spent quite a few years designing such systems, improving them, taking new techniques for modulation and error correction into standards bodies, doing industrial research and sponsoring research in universities on just these topics.

I used to help educate the guys at CIA/NSA/NRO/other non-existent TLA organizations on the theory behind just this stuff for satellite earth-station equipment that we were providing them at the time (this was at another company before I joined Intel).

And what are your qualifications, exactly?

:)

Eric Jacobsen Minister of Algorithms, Intel Corp. My opinions may not be Intel's opinions.

Here's a hint: If you really think you have something to add, then do so. In comp.dsp you're just going to look like other trolls that show up here from time to time if you continue to resort to such unproductive denegration.

I don't believe there are any analog systems that "reach" capacity. Can show how any existing analog system reaches capacity as described by Shannon?

To the original point, when trying to achieve capacity in a channel the channel coding is crucial. Source coding, if it's used at all, is usually done separately from channel coding for a large number of reasons. The source coder is almost always far separate, physically and architecturally, from the channel coder. The channel coder typically knows nothing about the existence or type of source coder, so the ability to carry "information" in the channel is independent of that.

BTW, just so you know, Randy isn't exactly a digitial comm noob, either.

Eric Jacobsen Minister of Algorithms, Intel Corp. My opinions may not be Intel's opinions.

*BELOW* that capacity error free operation is possible. It is *not* _always_ possible, even then.

Regardless, you are confusing two different issues.

That "capacity" is not the same as the threshold for virtually error free operation above a minimum SNR for each different type of digital encoding.

If what you have just said were true, what does that say about this statement:

that is entirely irrelevant. To achieve this magical feat, it must do so at a reduced *information* rate (forward-error-correction bits, retransmissions, etc. don't count as information).

You now say FEC is required to achieve rates as high as the maximum capacity, as part of your defense of a statement that says it reduces the information rate...

The conflict is because you have confused the SNR threshold for digital systems to operate essentially error free with the maximum capacity that Shannon theorized.

So?

It is used in order to maintain low a BER when the SNR is not high enough.

The question is not if it reduces throughput, but whether it is required for low error rates when the SNR is high enough.

Then you would have pointed them out, which you haven't. You are dishonest and continue to post cheap insults. Shame on you.

It can be configured off on virtually all of them. That is not the point though. The reason it is used is because the modems are operated at times with an SNR that is low enough to generate errors. The *target* design point for such a link is barely above the SNR threshold. Which is to say that if FEC is disabled, the channel will (barely) meet specifications most of the time. Of course most of the time is not good enough, and therefore FEC is enabled.

Irrelevant to the discussion.

How do you explain the above confusion then?

Not an insignificant point. In particular given the confusion you posted above.

If so you would have pointed out a few by now.

What accounts for your current confusion?

Three decades working with satellite communications systems.

--
Floyd L. Davidson            
Ukpeagvik (Barrow, Alaska)                         floyd@apaflo.com

Good point. Given there are so many variations, we need to be specific.

A v.90 modem using digital signaling, not analog. It of course has filtering, amplification, etc. etc.

--
Floyd L. Davidson            
Ukpeagvik (Barrow, Alaska)                         floyd@apaflo.com

When defined as an SNR or Eb/No (which is common), but of which I wasn't explicit, error-free transmission is _always_ possible at or above capacity.

Could you explain what you mean? Because it sounds like you're describing channel capacity.

No, there's no inconsistency. For any modulation type and code rate one can compute channel capacity. Channel capacity cannot be achieved without channel coding.

So it helps explain why John's statement was not erroneous as you'd claimed.

What do "low" and "high enough" mean with respect to BER and SNR? Achieving capacity means truly error-free, not "low" BER, and using FEC is _always_ more power efficient than not using it. If you're ever using a digital satellite modem with the FEC turned off you're wasting channel resources, regardless of the target BER or SNR.

It is sometimes possible to achieve a desired BER without coding by cranking the power way up (to make the SNR higher), but it is not efficient. It is always more efficient to use FEC, so, as I just mentioned, regardless of your BER target or SNR, if you have the FEC turned off you're wasting channel resources, either bandwidth or power or both.

More troll behavior, besides your logic being flawed. I already explained why I hadn't gone back and addressed all your errors. Feel free to review the thread if you've forgotten.

That's generally for test purposes, as it is not efficient to turn the FEC off during use. So I take it you can't name a digital satellite modem that doesn't use (i.e., include) "forward-error-correction bits"?

Can you explain "target design point for such a link" and "SNR threshold"? And what "specifications" are you trying to meet? It matters quite a bit.

And, as I've already mentioned, if you ever turn off the FEC you're wasting resources, regardless of your definitions of any of those terms.

How so? Here's your chance to show you know what you're talking about.

Oh, now I see... :rolls eyes:

Easy. I'm not the one who's confused.

In what capacity? Operator? Technician? Installer? Dust them off occassionally?

Eric Jacobsen Minister of Algorithms, Intel Corp. My opinions may not be Intel's opinions.

Mr. Davidson,

I considered responding to your comments one-by-one but decided not to because I don't think you are interested in the truth, just in glorifying yourself.

If your intention changes and you decide to pursue knowledge rather than glory, and you can formulate your responses in a respectful manner (Hint: you can disagree with a person without insulting them), then I may consider resuming a conversation with you.

--
%  Randy Yates                  % "...the answer lies within your soul
%% Fuquay-Varina, NC            %       'cause no one knows which side
%%% 919-577-9882                %                   the coin will fall."
%%%%            %  'Big Wheels', *Out of the Blue*, ELO
http://home.earthlink.net/~yatescr

That's not really true. Typically the channel coder is designed in a very source coder aware manner. At least in the really high volume systems, like cellular. The most crucial bits coming out of the source coder (e.g. the MSBs of the voice energy) get a lot more channel coding protection than the less interesting bits.

Steve

Yes, a system that carries only voice can sometimes do that. Not all cellular systems do that (I'm taking your word that at least one does), but systems that carry generic payloads (e.g., mixes of voice, video, data, whatever), don't have that luxury. A voice-only cellular system is unique in that the source codec is in close proximity to the channel codec and they can be matched as you've described, but that's a fairly unique (although practical) case.

As the cellular systems move to VoIP, voice+data, voice+video+data, then joint source-channel coding starts to get pretty intractable.

Cheers,

Eric Eric Jacobsen Minister of Algorithms, Intel Corp. My opinions may not be Intel's opinions.

(snip on digital vs. analog)

If it isn't baseband, I would say it has to be modulated. If you call ethernet phase modulated instead of manchester coding is the ethernet controller then a modem?

I would probably believe that an NRZ signal isn't modulated. Most others are in some way removed from a direct representation of the digital signal.

(snip)

DSL isn't baseband or it would interfere with the voice phone signal. I think you will find some modulation going on.

(snip)

-- glen

(snip)

I wasn't thinking about frequency, but if the poles are not exactly the same then each sine peak won't be exactly the same amplitude. The result, then, could be a 1.2Hz component in the spectrum. Though a 100 pole generator won't have a 3.6Hz component, the next should be 6Hz.

-- glen

Eric,

It occurs to me that you're having a chat with Eliza .. or the near equivalent.

Fred

I assume what you are trying to say is you can't put DC through a transformer, but things like Ethernet and T1/E1 all require isolation. They cannot, therefore, be truly baseband.

I'd say manchester coding would classify as modulation. If you scramble to introduce plenty of timing transitions, and minimise DC, is that a form of modulation? Some of these signals aren't transmitted as conventional 2-level digital signals. They turn the data into a ternery code, and round all the corners to minimise the generated wideband noise. They are trying to follow all the rules of a well behaved analogue signal on the wire.

Not all DSL works in conjunction with voice, so that is not a comprehensive argument. However, pretty much all DSL in use today seems to use some form of OFDM.

Regards, Steve