Sharing a Medium: Multiplexing

Why share at all?

Pulling a wire or laying a fiber is expensive; the air's radio spectrum is finite and licensed. So the single most valuable resource in the physical layer is the medium itself, and we cannot afford one dedicated cable per conversation. Multiplexing is the family of tricks that lets many independent signals travel over one shared medium at the same time, while a matching step called demultiplexing pulls them cleanly apart at the far end.

Think of a four-lane motorway. The road is one shared medium, yet many cars travel without colliding because they keep to separate lanes (a spatial split) or take turns at a merge (a time split). Multiplexing does the same to a wire, only the lanes are made of frequency, of time slots, or of colours of light. Crucially, this is a physical-layer idea about sharing a link; do not confuse it with the transport-layer connection machinery higher up, where port numbers multiplex many apps onto one host. Same word, different floor of the building.

FDM: everyone gets their own frequency lane

The oldest scheme is frequency-division multiplexing. The medium's total bandwidth (its band of usable frequencies, measured in hertz) is sliced into narrow sub-bands, and each signal is modulated up onto its own sub-band, sitting there for the whole session. This is exactly the broadband idea from the last guide put to work: lift each conversation onto a different carrier frequency and they coexist without ever touching, because they occupy different parts of the spectrum.

You meet FDM everywhere. FM radio stations each own a frequency you dial in; your home's cable line carries TV channels and your internet on separate frequency bands of the same coax; Wi-Fi splits its band into channels so neighbouring routers can avoid each other. To stop neighbours from bleeding into one another, FDM leaves an unused guard band between sub-bands, a little empty buffer of frequency, which is part of why you never quite get the full theoretical bandwidth.

TDM: everyone gets their own turn

Time-division multiplexing shares the medium the other way: instead of giving each signal a slice of frequency forever, it gives every signal the whole frequency band for a tiny slice of time, in a strict repeating rotation. Source 1 gets slot 1, source 2 gets slot 2, and so on; after the last source, the cycle (a frame) repeats. Because the turns come around thousands of times a second, each conversation feels continuous even though it is really transmitting in rapid bursts.

FDM (split by frequency)        TDM (split by time)
  freq                            time runs left to right
  high [==== user C ====]         | A | B | C | A | B | C |...
   |   [==== user B ====]         [___ one frame ___] repeats
  low  [==== user A ====]         each user owns the WHOLE band
       all flow at the same time  but only inside its own slot

There is an honest catch. In classic synchronous TDM, a source's slot is reserved even when it has nothing to send, so an idle conversation wastes its turn and the link runs below capacity. Statistical TDM fixes this by handing each empty slot to whoever actually has data right now, squeezing out more useful throughput; but it then needs a label on each burst to say who it belongs to, which is the seed of packet switching. So the neat reserved-slot picture is a simplification that real high-efficiency links deliberately break.

WDM: many colours down one fiber

On optical fiber, the same FDM idea reappears with a beautiful twist and a new name: wavelength-division multiplexing. Light of different wavelengths is just light of different frequencies, which our eyes read as different colours. WDM sends many independent data streams down one glass fiber at the same time, each on its own colour of laser light. The colours pass through each other without interfering, and a prism-like device at the far end separates them back out, one per receiver.

This is how the global Internet backbone reaches staggering capacity. A single fiber can carry dozens or even hundreds of colours (dense WDM), and each colour itself runs a high-speed modulated signal using tricks like QAM from the last guide. Multiply hundreds of colours by tens of gigabits per colour and one hair-thin strand of glass carries many terabits per second. WDM is really FDM done with light: same principle of frequency lanes, just up in the optical band where the available bandwidth is enormous.

Walking one signal through a multiplexed link

Let us follow one phone call across a TDM link, end to end, so the mechanism becomes concrete. The device that combines the inputs is a multiplexer (a mux); the one that pulls them apart is a demultiplexer (a demux). The two must agree perfectly on timing, because in TDM the only thing that tells the demux who a burst belongs to is exactly when it arrives.

1. Several callers feed the mux at once; the mux assigns each one a fixed slot number inside the repeating frame (caller 3 always lands in slot 3).
2. Each frame, the mux grabs a small chunk of caller 3's audio bits and places them into slot 3, then moves on to slot 4, sending the assembled frame onto the shared medium.
3. The bits cross the medium as one combined high-rate stream; the medium neither knows nor cares that several conversations are interleaved inside it.
4. At the far end the demux locks onto the frame's start using a sync marker, counts off the slots, and copies whatever lands in slot 3 to caller 3's output line.
5. Reassembled slot by slot, caller 3 hears a smooth conversation, never noticing that their voice traveled as thousands of tiny bursts taking turns with strangers.

Now you can see why these schemes feel like the postcard-and-envelope analogies from earlier rungs but a layer below. Multiplexing does not add an address; it gives each conversation a reserved frequency, colour, or time slot, and the receiver demultiplexes purely by where or when a signal shows up. That is the whole foundation the higher layers stand on: by the time a frame, a packet, or a stream reaches them, the messy job of sharing one piece of copper, glass, or air among many talkers is already solved down here.