My dial-up modem fantasy

Hmmm...let me think. Since its not possible to exist, its a moot point in discussing any advantages or disadvantages

Let's say it was possible to exist. What would be the drawbacks?

I think there's no point. The limited data resolution available to both older, analog phone lines and contemporary, digital phone lines sets a hard, hard limit on the available bandwidth for *ANYTHING* going over a phone line, and contemporary modems are already very close to that theoreticallimit. So switching people over to a new and entirely incompatible modem technology would provide no significant benefit.

The laws of physics would have changed, so the universe as we know it would cease to exist. Other than that, seems OK to me.

I hate when that happens.

-- Grant

sorry, those of us who clone advanced life forms in our supersonic colonic colliders don't like to talk about such things...ooops gotta run, the Interociter is whizzing...

Is that a quantum interociter or an n-field multiversal interociter? I've noticed that the latter really kick up a racket when the neutron-flow polarity gets out of whack.

You're talking about bandwidth. A baud of 1-symbol-per-second uses only a small amount of bandwidth, regardless of the amount of bits-per-symbol. Right?

Wrong, but then trying to explain Shannon's Law to you would be a waste of bandwidth.

Aww...what the heck. I'll feed the troll.

Use of the word baud is outdated as it can't adequately reflect the additional data load achieved by advanced modulation techniques.

Technically speaking, 900 baud is the highest a standard "toll-quality" phone line will support. So how did they get 2400 baud modems to work? They didn't, they got 2400 bps modems to work. A 2400 bps modem is actually a 600 baud device; but using phase modulation, they got four different states per baud.

Do you really spend that much time on the john reading Popular Science?

I know that.

Baud and bits-per-baud are two different things.

The only way the bit-rate is the same as the baud is if you have 1 bit per symbol. The 2400 bps using a 600 baud is derived using 4 bits per symbol. 6X4=24.

So back to my question.

Wouldn't a baud of 1-symbol-per-second use only 1 Hz of bandwidth regardless of the amount of bits-per-symbol?

Back to my original answer, NO. You really need to revisit Communications Theory 101.

Depends on the modulation scheme, but basically yes. However, the number of states you can support depends on your noise floor. In order to get, say, 32768 bps over a 1 Hz channel, you would need a noise margin of:

log_2(1+S/N) = 32768

S/N = 2^32768-1 ~ 2^32768

log_10 S/N ~ log(2)/log(10) * 32768 ~ 9864

log_10 S/N ~ 9864 B = 98640 dB

Even if you're only limited by quantum noise, this still requires a trans-astronomical amount of power.

-hpa

In the real world, what is the maximum amount of bits-per-symbol that can be achieved using PSK without exceeding the dynamic range of a phone line?

Lets dispense with confusing references to bits, symbols, phase shift keying, etc. and get down to the nitty gritty....

Answer: About 48 Kbps data rate.

Well, given the fact that most telephone circuits are digitized to

The copper lines themselves are obviously capable of much higher capacity with a suitable change of equipment at the central office. This is the technical basis for ISDN and DSL.

-hpa

I am specifically interested in the bits-per-symbol only. I wasn't asking about the date rate -- i.e. the 48 Kbps you described. I am asking only about the bits-per-baud.

Assuming a baud of only 1-symbol-per-second, what is the maximum amount of bits-per-symbol that can be achieved using PSK without exceeding the dynamic range of a phone line?

I believe the maximum using any encoding should be in the neighborhood of 8 bits. That's assuming an SNR of around 24dB, which is...optimistic. In practice, a fairly conservative figure of 20dB is often seen, which yields

V.90 manages better than that, but it requires one end be connected directly to a digital trunk.