The idea
A physical link is expensive to lay and often carries far more bandwidth than any one signal needs. Running one cable per conversation would waste almost all of it. Multiplexing is the fix: a technique that lets multiple signals share a single data link at the same time, instead of one link per signal.
The lecture gives three ways to do it, and they differ in exactly one question: what do you divide up to make room for everyone — frequency, or time?
FDM — dividing the frequency
How it works
Frequency-Division Multiplexing
FDM is an analog technique. It applies when the bandwidth of the link is greater than the combined bandwidths of the signals being carried — there is spare spectrum, and FDM slices it up.
Each channel gets its own strip of frequency. Between neighbouring channels
sits a guard band — a strip of unused bandwidth that stops one channel’s
signal bleeding into the next. For n channels there are n − 1 guard bands,
one fewer than the number of channels, because a guard band only exists
between two neighbours.
Worked example
Five 100 kHz channels, 10 kHz guard bands
Channel bandwidth. Five channels at 100 kHz each:
5 x 100 = 500 kHz.Guard bands. Five channels need
5 - 1 = 4guard bands, at 10 kHz each:4 x 10 = 40 kHz.Total.
AnswerTotal bandwidth = 540 kHz
Where marks get lost
n channels need n − 1 guard bands, not n
The most common slip in this calculation is using n guard bands instead of
n − 1. Picture it: three channels sitting side by side only have two gaps
between them, not three. Draw the channels as boxes if the count is not
obvious — the guard bands are the spaces, and there is always one fewer space
than there are boxes.
Aside
Not every FDM question includes guard bands. If three 4 kHz voice channels are packed directly into a 12 kHz link with no guard bands, they simply sit edge-to-edge: 20–24 kHz, 24–28 kHz, 28–32 kHz. The formula above is for when guard bands are explicitly part of the question — read carefully for whether they are.
TDM — dividing the time
How it works
Synchronous Time-Division Multiplexing
TDM divides the usage of a high-capacity link into time slots, one per device, repeating every frame. It only works with digital signals — this is the direct counterpart to FDM being analog-only. It is also called synchronous TDM, because every device gets its slot on a fixed schedule whether or not it has anything to send.
Worked example
1 Mbps input, 4 channels, synchronous TDM
Input slot duration. Each channel arrives at 1 Mbps, so one input bit takes:
Output slot duration. Four channels share the output link, so each output slot is a quarter of the input slot:
0.25 μs.Output bit rate. The output carries all four channels’ bits: four times the per-channel rate.
Frame duration and frame rate. One frame carries exactly one bit from each of the four channels, so a full frame completes at the same rate as a single channel’s own bit rate:
AnswerOutput bit rate 4 Mbps, frame rate 1 Mbps
STDM — dividing the time, only when there is data
How it works
Statistical Time-Division Multiplexing
Synchronous TDM’s fixed slots waste bandwidth on any device that has nothing to send that frame. Statistical TDM (also called asynchronous TDM) fixes this: the frame is built from a statistical analysis of which input lines are likely to be transmitting at any given moment. The multiplexer scans the input lines, and only when a frame is filled with active data does it send that frame across the link.
The cost is addressing. Because a slot’s position no longer tells the receiver which device it belongs to — an idle device is simply skipped rather than sent as an empty slot — every slot needs an address, so the receiver can tell whose data just arrived.
| FDM | TDM (synchronous) | STDM (statistical) | |
|---|---|---|---|
| Domain divided | Frequency | Time | Time |
| Signal type | Analog | Digital | Digital |
| Slot/band allocation | Fixed strip per channel | Fixed slot per device, every frame | Variable — only active devices get a slot |
| Idle device | Its band still exists, unused | Its slot is still sent, empty | Its slot is skipped entirely |
| Needs addressing? | No | No — position identifies the device | Yes — a skipped slot means position alone is not enough |
Check yourself
Ten 4 kHz voice channels are combined with FDM using 500 Hz guard bands between them. What is the total bandwidth?
Ten channels: 10 x 4 = 40 kHz. Nine guard bands (10 - 1) at 0.5 kHz each:
9 x 0.5 = 4.5 kHz. Total: 40 + 4.5 = 44.5 kHz.
Putting it together with PCM
Worked example
100 voice channels in a 20 kHz passband
This is where multiplexing, PCM and spectral efficiency from the earlier topics meet in one question.
Bandwidth per channel. 100 digitized voice channels share a 20 kHz passband:
Bit rate per channel. Each channel is standard telephone PCM — 8000 samples/s x 8 bits/sample:
Spectral efficiency. Bits per second, per hertz of bandwidth used:
Notice this is the same spectral-efficiency idea from the very first topic in this module — more data rate for the same bandwidth — now applied across a whole multiplexed bundle rather than one signal.
Answer200 Hz per channel; 64 kbps per channel; 320 bits/s/Hz
In the exam
- Name the three techniques. FDM, synchronous TDM, statistical (asynchronous) TDM. Tutorial 2 asks for this list directly.
- FDM is analog, TDM is digital. State this pairing plainly — it is a one-mark distinction that gets missed under time pressure.
- Guard bands: n − 1, not n. Draw the channels if you are unsure.
- TDM slot and frame arithmetic. Input slot =
1/input rate. Output slot shrinks by a factor of the channel count. Output bit rate multiplies by the channel count; frame rate matches the per-channel input rate. - TDM vs STDM. Fixed reserved slots vs only-active-devices slots, and the addressing cost STDM pays for the efficiency it buys.
- WDM. Recognise it as the fibre-optic name for the same idea as FDM — the lecture gives no more detail than the name and the medium it applies to.