Novice Phone Switch Network Sampling Freq. Question

[[email protected]]

Hello all.

I have some questions about how the "typical" phone line switch network
works (being in the US).

This is what I understand:
When two people (two phones/nodes) talk using standard hard wired or
cordless phones it may appear to them that they are having a continous
conversation, but they really are not. The phone network is a switching
network where multiple people (at multiple nodes) will talk and there
voice signals will be sent over the phone lines for a given time and
then not while another two nodes are connected. This is a good example
of TDM (time division multiplexing).

I have two questions:
1) Since people can hear audio signals around the bandwidth of 200Hz -
20kHz and most people speak with an audio signal frequency between 1kHz
- 2kHz. What is the sampling frequency that the phone lines switch at?
Would it be 44.1Khz? (Somewhat satisfying Shannon's Sampling Theorem).

2) From question (1) it seems to me that if this sampling frequency is
static then that puts a limit as to how many people (nodes) can be
connected to the phone line network. If there are more nodes than the
sampling frequency can handle then the "switched off" times for nodes
will be too great and there will be noticeable distortion.
If the sampling frequency is not static and is dynamic, what control
system(s) "sees" how many nodes are trying to connect so that the
sampling frequency can be adjusted accordingly.
Thanks for your time. I again am a novice.

jacobdav

 

[Bob]

Hello all.

I have some questions about how the "typical" phone line switch network
works (being in the US).

This is what I understand:
When two people (two phones/nodes) talk using standard hard wired or
cordless phones it may appear to them that they are having a continous
conversation, but they really are not. The phone network is a switching
network where multiple people (at multiple nodes) will talk and there
voice signals will be sent over the phone lines for a given time and
then not while another two nodes are connected. This is a good example
of TDM (time division multiplexing).

I have two questions:
1) Since people can hear audio signals around the bandwidth of 200Hz -
20kHz and most people speak with an audio signal frequency between 1kHz
- 2kHz. What is the sampling frequency that the phone lines switch at?
Would it be 44.1Khz? (Somewhat satisfying Shannon's Sampling Theorem).

2) From question (1) it seems to me that if this sampling frequency is
static then that puts a limit as to how many people (nodes) can be
connected to the phone line network. If there are more nodes than the
sampling frequency can handle then the "switched off" times for nodes
will be too great and there will be noticeable distortion.
If the sampling frequency is not static and is dynamic, what control
system(s) "sees" how many nodes are trying to connect so that the
sampling frequency can be adjusted accordingly.
Thanks for your time. I again am a novice.

jacobdav

The public switched telephone network (PSTN) samples each telephone 8000
times per second. The associate equipment's high-pass and anti-aliasing
filters result in a net bandpass of 300Hz through 3400Hz. Each digital
sample is 8 bits, but the weight of each digital level is not linear. The
"step size" of each level gets closer together as the sampled analog signal
is smaller (in magnitude). In the USA this is called ulaw (pronounce mu
law). In Europe they use alaw. Both coding schemes' goal is to provide
constant signal-to-noise ratio -- as a function of signal level (i.e., as
the signal level gets lower then the quantization noise gets lower, too).

This 8ksps @ 8bps results in a 64Kbps voice channel. This is commonly
referred to as a DS0 channel. An OC192 fiber interface (roughly 10Gbps)
carries 129,024 DS0's. That's a lot of phone calls on two strands of glass.

The sampling frequency/sample width was a compromise to maximize number of
phone calls (for a given interface rate) and to reproduce acceptable voice
quality.

Bob