1) History: SONET is derived from old telephone company technology based on time division multiplexing³ (tdm). Upon the breakup of the AT&T monopoly in 1984 many local carriers, each having their own proprietary optical tdm systems, were forced to interface with multiple long distance carriers. This presented a challenge from which the SONET standard was born. SONET was paralleled by CCIT standards (G.707, G.708, G.709) in 1989. The CCIT recommendations are referred to under the blanket title Synchronous Digital Heirarchy (SDH) and differ from the SONET standard in minor ways.

2) Current Deployment: "Virtually all the long distance traffic in the United States, and much of it elsewhere now uses trunk running SONET in the physical layer."² Chipsets providing an interface between computers and SONET do exist and should decline in price, like all things, over time. These chipsets provide the basis for the creation of SONET network interface cards (NIC's) which allow for the direct connection of machines to a SONET ring via leased lines.

3) SONET Design Goals:

4) SONET Methods: SONET is a traditional TDM system devoting the entire bandwidth of the fiber to a single channel, which is divided into timeslots. In order to use this method a master clock is required, which is accurate to one part in 10⁹. As such, bits are sent out at precise intervals whether or not there is data to be sent. Framing will be discussed later.

5) SONET Components: SONET rings are composed of switches, multiplexers and repeaters, which are connected by fiber cable. SONET is often deployed in a dual ring topology. Note the self-healing characteristic of SONET and the automatic rerouting of data in the event of failure.

6) Framing: SONET Frames are 810 bytes long and sent every 125 m sec. This provides for 8000 frames per second, which is tied to the sampling rate of PCM channels, which is based on the Nyquist⁵ theorem. The frame provides 6264 bits for user data in a space called the Synchronous Payload Envelope (SPE).

7) Channel Composition: The composition of many small channels called tributaries is an important aspect of SONET. Small channels such as T1 or T3 carriers are converted to STS-1 channels, the basic SONET rate (51.84 Mbps). Three STS-1 channels can be multiplexed to onto one STS-3 carrier (155.52 Mbps). Three STS-3 channels can be multiplexed onto one STS-12 channel providing data rates around 622.08 Mbps. The optical equivalents of the STS channels also have names to complicate things but follow a system: STS-1 = OC-1, STS-3 = OC-3, STS-12 = OC-12 etc. Note that the multiplexing is done byte by byte in a round-robin fashion. When three STS-1 channels are multiplexed, one byte from chan. 1 is modulated, followed by one byte from chan. 2, followed by one byte from chan. 3 and then back to one. Also note that CCIT names are different from those provided here, this is only an issue when dealing with foreign carriers.

8) Concatenated Lines: By convention any OC line carrying data from a single source, non-multiplexed, is given the suffix c. An OC-3 line carrying data from a single source would be denoted OC-3c.

9) Physical Layer: The SONET physical layer is sub-divided into four sublayers, the lowest of which is the photonic sublayer. The photonic sublayer is where light and fiber properties are specified. The other three sublayers are concerned with the framing and transport over sections, lines and paths.

10) Language:

Multiple analog signals are terminated and digitized by a codec producing a 7 or 8 bit number. The sample frequency of a codec is 8000 Hz. Note that many incompatible PCM methods exist around the world, but this is normally handled by the carrier possibly increasing cost.

In time division multiplexing, a number of analog channels are sampled at a standard rate forming an analog stream, which is sent to a single codec for modulation onto the digital channel. This is done as opposed to the use of multiple codecs for financial as well as design reasons. Many carriers used in network engineering make use of tdm, including the T1 carrier.

There are two varieties of fiber optic cable commonly in use, single-mode and multi-mode fiber. Single-mode fiber is of higher quality and allows for longer runs between repeaters but also costs much more than multi-mode fiber. The main difference between the two is the reflection, or lack there of, of light at the silica/air boundary. Single-mode fiber reflects all light due to its diameter being a few microns where multi-mode fiber has some loss at the boundary due to its larger diameter.

The theories of Nyquist and Shannon are often looked to for determining digital signaling limitations. Digital signals are modulated onto a cable as a fundamental and a number of harmonics as described by Fourier, the number of harmonics you can get onto a cable is determined by the bandwidth of the cable. Note that the bandwidth of a cable is not the same that is referred to when we speak of "bandwidth" exactly, since a coaxial cable has bandwidth around 0000 but most coaxial networks run are said to have 10 Megabit bandwidth. It’s a matter of terminology. Now consider the number of Fourier harmonics modulated onto a cable, in order to achieve a perfect square wave, we need an infinite number of harmonics, which is not possible. Also the physical medium diminishes different harmonics at different rates producing distortion. The bandwidth of a cable is the highest frequency at which the cable is guaranteed to transmit harmonics without distortion.

The rate at which data is transmitted is often confusing. Two term's baud and bps exist which colloquially are interchanged but actually have different meaning. Baud refers to the number of signal changes per second. If one bit is represented by one change then baud is the same as bps, but today encoding schemes are used to transmit multiple bits per signal change causing bps to be some multiple of baud.

A historical note on the bandwidth of voice grade lines is in order. When the original fiber voice networks were deployed, the bandwidth of the low quality copper wire was around 3000 Hz. Today, much higher quality cable is deployed but a filter is introduced at 3000 Hz to maintain the 3000 Hz bandwidth of voice grade cables.

In 1924, H. Nyquist wrote an equation expressing the maximum data rate of a noiseless channel with finite bandwidth. This was 2Hlog₂V bits per second where H is the bandwidth of the channel and V represents the discrete levels in the signal. This is important because it proved that higher sampling rates were useless since those harmonics caught by this increase in sampling rate would never exist on the cable due to capacitance.

In 1948, Claude Shannon extended Nyquist's theorem to account for thermodynamic noise on a channel. Shannon's theorem states that the maximum number of bits per second is defined as Hlog₂(1+S/N) where H is again the bandwidth and S/N is the signal to noise ratio on the cable. S/N is commonly expressed as a number instead of a ratio. The units of the number are decibels (dB).

Tanenbaum, Andrew S. Computer Networks: Third Edition"

Copyright 1996 Prentice Hall

2 Tananbaum pp125

Sprint Datasystems SONET Literature

http://www.sprintbiz.com/sonet/index.html

SONET Transmission and Architecture

http://www.sprintbiz.com:80/sonet/speed.html

BayNetworks, now Nortel

http://www.baynetworks.com

Network Associates

http://www.networkassociates.com

Cabletron Systems

http://www.cabletron.com