[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Talking with Yahoo Messenger

John Hasler wrote:
Mark Allums wries:
In the US, a 56k dialup running at 53k (max allowed by law, and rarely
achieved in practice) gives you just about the necessary bandwidth for
voice over IP, *and nothing else*.

That's 56k downbound.  Upbound is 33k max.

Been upgraded to 44k.  dialup still sucks though.

Not a voice-over-IP expert, and yes, I *have* heard of Nyquist.

Good.  Now read Shannon.  The number of bits per second you can push
through a channel depends on the bandwidth _and_ the signal to noise

Read him. I am being specific about the approximate reproduction of a "waveform" from digital samples here, not about maximum channel capacity. Two samples per "wave" ensures that all the frequencies are there, but three bits makes a more "accurate" reproduction. At the highest frequencies, the human ear probably can't tell the difference, which is why the sample rate of a standard CD is "only" 48kHz. (44.1 after mastering)

What modulation does Skype use?  Some type of delta modulation would be
my guess for dialup, but I have no real idea.

None.  Skype surely uses some sort of compression, but the modem handles
the physical layer using complex multitone modulation schemes.

Again, I am being specific about the encoding of the voice signal into bits, not the type of signal modulation done on the physical channel. Which is a type of angle modulation, specifically a phase encoding whereby each transition encodes three bits.

Delta modulation is good for a DAC with a small sample size. Remember those DOS games where they managed to reproduce a digitized sound out of the PC "beep" speaker? That was a one-bit DAC, and they used delta modulation to make it work.

My problem is, I think, that I don't use the standard terminology very well.

Mark Allums

Reply to: