Evolution of Switching

As noted, the first switch was a switching matrix (board) operated by a human. The 1890s saw the introduction of the first automatic step-by-step systems, which responded to rotary dial pulses from 1 to 10 (that is, digits 1 through 9 to 0). Cross-bar switches, which could set up a connection within a second, appeared in late 1930s. Step-by-step and cross-bar switches are examples of space-division switches; later, this technology evolved into that of time-division. A large step in switching development was made in the late 1960s as a consequence of the computer revolution. At that time computers were used for address translation and line selection. By 1980, stored program control as a real-time application running on a general-purpose computer coupled with a switch had become a norm.

At about the same time, a revolution in switching took place. Owing to the availability of digital transmission, it became possible to transmit voice in digital format. As the consequence, the switches went digital. For the detailed treatment of the subject, we recommend Bellamy (2000), but we are going to discuss it here because it is at the heart of the matter as far as the IP telephony is concerned. In a nutshell, the switching processes end-to-end voice in these four steps:

1. A device scans in a round-robin fashion all active incoming trunks and samples the analog signal at a rate of 8000 times a second. The sampled signal is passed to the coder part of the coder/decoder device called a pulse-code modulation (PCM) codec, which outputs an 8-bit string encoding the value of the electric amplitude at the moment of the sample
   

2. Output strings are fed into a frame whose length equals 8 times the number of active input lines. This frame is then passed to the time slot interchanger, which builds the output frame by reordering the original frame according to the connection table. For example, if input trunk number 3 is connected to output trunk number 5, then the contents of the 3rd byte of the input frame are inserted into the 5th byte of the output frame. (There is a limitation on the number of lines a time slot interchanger can support, which is determined solely by the speed at which it can perform, so the state of the art of computer architecture and microelectronics is constantly applied to building time slot interchangers. The line limitation is otherwise dealt with by cascading the devices into multistage units.)

3. On outgoing digital trunk groups, the 8-bit slots are multiplexed into a transmission carrier according to its respective standard. (We will address transmission carriers in a moment.) Conversely, a digital switch accepts the incoming transmission frames from a transmission carrier and switches them as described in the previous step.

4. At the destination switch, the decoder part of the codec translates the 8-bit strings coming on the input trunk back into electrical signals.

Note that we assumed that digital switches were toll offices (we called both incoming and outgoing circuits trunks). Indeed, initially only the toll switches on the top of the hierarchy went digital, but then digital telephony moved quickly down the hierarchy, and in the 1980s it migrated to the central offices and even PBXs. Furthermore, it has been moving to the local loop by means of the ISDN and digital subscriber line (DSL) technologies addressed further in this part.

The availability of digital transmission and switching has immediately resulted in much higher quality of voice, especially in cases where the parties to a call are separated by a long distance (information loss requires the presence of multiple regenerators, whose cumulative effect is significant distortion of analog signal, but the digital signals are fairly easy to restore—0s and 1s are typically represented by a continuum of analog values, so a relatively small change has no immediate, and therefore no cumulative, effect).

We conclude this section by listing the transmission carriers and formats. The T1 carrier multiplexes 24-voice channels represented by 8-bit samples into a 193-bit frame. (The extra bit is used as a framing code by alternating between 0 and 1.) With data rates of 8000 bits per second, the T1 frames are issued every 125 ms. The T1 data rate in the United States is thus 1.544 Mbps. (Incidentally, another carrier, called E1, which is used predominantly outside of the United States, carries thirty-two 8-bit samples in its frame.)

T1 carriers can be further multiplexed bit by bit into higher-order carriers, with extra bits added each time for synchronization:

• Four T1 frames are multiplexed into a T2 frame (rate: 6.312 Mbps)

• Six T2 frames are multiplexed into a T3 frame (rate: 44.736 Mbps)

• Six T3 frames are multiplexed into a T4 frame (rate: 274.176 Mbps)

The ever increasing power of resulting pipes is depicted in Figure 1

Figure 1: The T-carrier multiplexing nomenclature.