Digital Transmission

Noise 

Analog signals "attenuate" or dissapate as they transmit over longer distances.  Sound gets quieter.  Can be compensated by boosting the signal.

A relay is responsible for this, aka analog amplifier.

However you still have the problem of noise which distorts the sound.  The noise is also amplified.  You get more noise the further the signal travels, while the meaningful data remains the same.  Range of analog transmissions is therefore limited.

Analog Multiplexing 

Multiplexing for analog transmission is only possible in one way   FDM.  You transmit multiple signals on one line, on different frequencies.  Suffers more from noise.

Codec:Coding and Decoding (MU-255 / A-LAW)

Redundancy in the data (set amplitude corresponds to 1) makes it far more resistance to noise.  Also an analog relay isn't needed, because a digital relay can read in the digital bits and duplicate it to eliminate noise.

Nyquest's Theorem: Sample at least twice per second * frequency range of the wave you want to reproduce.  Then you can represent it digitally.

Analog phone range: 300Hz - 3300Hz. 

Allocate 0Hz-4000Hz to help prevent cross-talk.  At least 8000 samples per second.

Quantizing scale: Used to measure the samples, and convert it into a byte representing the amplitude.

DS-0

Digital Signal Level 0.  The equivilent of one phone line.  Maximum 8000 bytes/s, 64000 bits/s.  8000 lost to noise (56k modems).

DS-1 (T1)

Equivilent to 24 DS-0 lines.  1536000 bits/s.  Advertised at 1544000.  The remaining 8000 belong to a communications channel.

Each DS-0 channel has 8 bits every frame (192 bits) plus one more bit for framing (193). * 8000 is 1544000.

Analog Over Digital 

Switches need to tell each other signalling information.  You have 8000 bits/second, and you can make all odd framing bits a syncronization bit (4000/s).  This is sufficient.

We use even framing bits to convey signalling information.  Only need a few to convey signalling information, 4000 is enough.  But not quite.  More advanced signalling needs more bits.

Instead we can use framing bits to indicate which frames contain signialling information, but it is wasteful.

So we still keep the voice data, but drop the least significant bit of each byte and use those for signalling information (24 bits per frame, 4000 possible frames maximum).  9600 bits/second.  Much better.  Bandwidth still takes a little hit.

Superframe: 12 frames

Extended superframe: 24 frames.  Doubles the number of signalling bits.  Used almost exclusively on T1.

Framing bits can carry CRC information, end of frame information, etc.

DS-1 refers to the signal, T1 refers to the technology.  DS-1 does not have to deliver T1.

ISDN

ISDN has the same bitrate but a different signalling structure.  Can carry voice and data.

ISDN is clear channel transmission.  ISDN transmits signalling information over separate, predefined channels.

T1s are now cheaper than ISDN and simpler, and now a better choice.  ISDN bills per bit.

Some providers absorb these differences and offer ISDN as T1.  No difference to end user.

More T1: Data

Originally intended to carry voice, great at carrying data.  Does not need to carry as much signalling information. 

Data transmissions tend to contain lots of 0s, tends to confuse receiving equipment.  Need to keep syncronized.

Solution: B8ZS - Bipolar 8 Zero Subsitution.  We will never translate 8 zeros in a row.  Instead, if we need to, we send two bipolar violations (transmits - - and + +) to indicate 8 zeros.

T1 wire looks like two copper pairs, terminated on a smart jack.

DS-2

Aggregate of 3 or more DS-1s.

Exist for syncronization. 

DS-3 (T3)

45mbit/s

You can buy partial T3s (or T1s, but that's silly).

T3s are multiplexed into really big pipes, too high frequencies for copper.  Optical is needed (fiber optics).