Re: Rockwell's 56K is vapor
Roy (garlic@garlic.com)
Wed, 02 Oct 1996 11:45:32 -0700
Jeff Weisberg wrote:
>
> | By Shannon's law(?), a local caller (one A->D conversion) actually can
> | push at most 28K through the pipe (1/2 the sampling rate). Slightly higher
> | rates can be achieved if the analog device matches the sampling rate at
> | the Central Office.
>
> close....
>
> it's Nyquist's law that says that a signal bandlimited to B
> requires a sampling rate of 2B.
> our sampling rate is 8kHz. ergo, our signal is limited to (a
> theoretical maximum of) 4kHz.
>
> Shannon's law tells us that the Capacity of our channel
> (in bits/sec) is
> C = B*log2(1 + SNR)
>
> where our noise is (mostly) due to quantization (to 8 bits)
>
> --jeff
While all of this is very illuminating, the trick that is probably
being used is to use an amplitude modulated signal which the
same number of levels as the telco A-D conversion and which is exactly
time synchronized with the telco. None of this violates the
rules above. The amount of noise and distortion will lower the
data rate slightly (in accordance with the laws above). I suspect
the units will have error detect/correct capability so errors will
cause transparent retransmits just like the current modems do.
Roy