Time-division multiplexing(TDM) is a term invented by telephone engineers to describe a system design where a common resource cable, backplane, radio-frequency spectrum-is divided into channels that are separated by time. This is in contrast to space-division multiplexing (SDM), which separates channels by modulating them in to different frequency bands. An example of SDM would be POST (plain old telephone service) telephone lines, and an example of FDM would be the analog microwave radios that carried telephone calls in the 1950s and 1960s. Radio broadcasting is, of course, another example of FDM.

TDM was a natural companion to first-generation digitation. Once audio is made into bits, it is quite easy to offset these into timeslots. Simple logic functions comprised of counters and muses are up to the task. In the early 1960s, engineers had to use the building blocks at their disposal.

The first application of TDM was the T1/E1 line, which was used as a “pair gain” scheme to obtain more channels from existing copper pairs. It was invented in 1962 and remains widely used to this day. It multiplexes 24 channels of 8 bit audio onto two copper pairs, taking advantage of the connection’s ability to pass frequencies much higher than the usual 3.4-KHz speech audio. Switching caught-up 20 years later with the introduction of the AT&T 5ESS central office switch in 1982. The line interface part of the 5ESS was comprised of many racks full of card cages holding “circuit packs” that adapted analog POTS lines to a digital backplane where the voice channel were divvied-up into timeslots. Like all TDM equipment, the switching subsection read a full cycle of timeslots into a memory and then wrote them out in different order. This pattern was followed by other vendors of central office (CO) switches, such as Northern Telecom, Ericsson, Alcatel, and Siemens. Smaller versions were made for PBX applications, as exemplified by the popular Nortel Meridian family.

TDM systems have only audio data within their timeslot. Because there are no signaling or routing instructions in the TDM slots, there needs to be an external mechanism to keep track of where everything is located and to make the needed associations for switching. For the PSTN (public switched telephone network), this is performed by a combination of the logic and storage inside the computers that drive the individual CO switches and the Signaling system 7 (SS7) protocol that runs between exchanges. The SS7 messaging is car       ried on data channels independent from those used for speech. In contracts, IP packets “know where they are going” because the dentitions address is contained within the header of the packet itself. IP routers make all the needed decisions about what to do with the packet based only on the information contained within its header.

In the pro-audio world, AES3 is a TDM transport system. The left and right channels are timeslot multiplexed onto a single cable. MADI (Multichannel Audio Digital Interface) extends the principle to more channels over wider-bandwidth coax cable.

Just as TDM transport led eventually to TDM switching in telephony, first genera ology. The designers of these products borrowed both the “cards in a cage backplane” and the “timeslots i cables” architectures from the telephone industry, and scaled them up to sever the requirement of high fidelity audio.

1. Statistical Multiplexing

Statistical multiplexing is the unsung of the Internet age. Without it, the Internet would not as we know it. Long haul bandwidth is much more expensive than local area bandwidth. That was the in sight of the Internet’s creators that guided many of their design choice. The first Internet was built upon 56kpps Telco data service links. With this speed, there was always more demand for bandwidth than was available, and it had it be rationed both fairly and efficiently.

The Internet’s designers looked switched phone network, and didn’t like it much. The engineers who built the PSTN had to build in a lot of expensive band width that was wasted most of the time. Long distance carriers in the United States love Mother’s Day because it motivates lots of revenue generating calling. (In fact, when a telephone engineer refers to the “Mother’s Day Effect,” he or she is talking about any event that fills some part of the network to capacity and denies service to many who want it.) Accordingly, the PSTN is designed so that all those doting sons and daughter don’t get frustrating busy signals and turn to letter writing. But that means that a lot of precious bandwidth lays unused of the rest of the year.

Imagine if each Websurfer needed to open a 64 kbps channel each he or she went online. Any time spent reading a page after downloading it would waste all of the channel’s bandwidth. Conversely, the surfer’s maximum bitrate would be limited to 64 kbps. Would the Web have been practical and successful in this case? You wouldn’t have YouTube, that’s for sure.

2. IP “Backplane”

AoIP receives no benefit from statically multiplexing because it needs a fixed and continuous bitrate for to each audio stream. That’s okay because we are running it over a LAN where bandwidth is plentiful and free. But it invites the question: Why bother with all this IP stuff when we don’t receive the main networking benefit, and TDM works just fine, thank you? Well we’ve already covered the big theme reasons for using IP (low cost, common infrastructure, native interface to PCs, in the IT and telephone mainstream, telephone/data/audio integration, etc.), but now we can look at this topic from another angle, with a pure network design perspective.

Think about those circuit packs and backplanes in TDM. They are all proprietary you can’t plug a Siemens circuit pack in to a Nortel switch. The same is true in pro audio you can’t use a vendor’s TDM (i.e., AES3) audio router in another’s card cage. On the other hand, in an AoIP system, the Ethernet RJ 45 becomes the equivalent of the TDM’s backplane, giving the advantage that a wide variety of equipment may interconnect via a standard interface. Also, Ethernet allows the circuit pack equivalents to be physically distant from the central switch. We can now enclose them in a box and locate them near audio inputs and outputs they serve.