Intuition The one core idea
The physical layer paints bits (1s and 0s) as voltage-vs-time pictures on a wire, then asks a single question: how fast can those pictures change before they blur into each other? Everything else — NRZ, Manchester, Nyquist, Shannon — is just a different answer to "how fast, and how do I draw it?"
Before you can read the parent note, you need a small toolbox. Below is every symbol and idea the parent leans on, built from nothing, each one earning the next.
A bit is one yes/no choice: 1 or 0. Nothing more. On a wire we agree on a code, e.g. "1 means +5 volts, 0 means 0 volts."
Definition Signal (voltage vs time)
A signal is a graph. The horizontal axis is time (seconds, moving left→right as the world ticks forward). The vertical axis is voltage (how hard the wire is being pushed). The receiver never sees your bits — it only sees this wiggly line and must guess the bits back.
Look at s01: the same three bits 1 0 1 drawn as a voltage picture. The whole chapter is about choosing how to draw that picture.
T b
The bit period T b is how long (in seconds) the wire holds the picture for one bit. Small T b = each bit gets little time = you are going fast.
Definition Frequency and the hertz (Hz)
Frequency counts how many times something repeats per second . Its unit is the hertz , written Hz : 1 Hz = "once per second." A wiggle that completes 3000 full up-down cycles each second is 3000 Hz .
Picture: a wave; count the peaks that pass a fixed point in one second — that count is the frequency.
Intuition Why the topic needs frequency
A wire is picky: it can pass slow wiggles happily but smears fast ones. To say "how fast can I wiggle" we must measure wiggling — and frequency is exactly that measurement. Without Hz there is no Nyquist and no Shannon.
B
Bandwidth B (in Hz) is the range of frequencies a channel lets through :
B = f high − f low
where f high is the fastest wiggle it passes and f low the slowest.
In s02, the shaded band is the set of frequencies that survive; anything outside is chopped off. A wider band = more room to wiggle fast = more bits/sec possible. This single number B is the star of every capacity formula.
Common mistake Bandwidth is not "amount of data"
In everyday speech "bandwidth" means data-per-second. Here it is strictly a width in Hz — a range of frequencies. The link between the two is exactly what Nyquist and Shannon compute.
Encoding often uses more than two voltage levels. To count how many bits that buys, we need one tool.
2 n
2 n means "multiply 2 by itself n times." 2 3 = 2 × 2 × 2 = 8 . Picture: a branching tree — each new bit doubles the number of leaves. n bits ⇒ 2 n distinct patterns.
s03 shows the doubling tree: 1 bit → 2 patterns, 2 bits → 4, 3 bits → 8.
log 2 and not some other tool
We want: given M distinct voltage levels, how many bits does one symbol carry? Since M patterns need n bits with M = 2 n , the answer is by definition n = log 2 M . The log is simply "how many yes/no questions to pin down one of M choices." No other function answers that question.
S , noise power N
S = strength (power) of your wanted signal. N = strength of the random junk (noise) also on the wire. See Signal-to-Noise Ratio .
Definition Signal-to-noise ratio
S / N
A plain division : how many times stronger the signal is than the noise. S / N = 1000 means the signal is 1000× the noise power. It has no units — it's a pure ratio.
Definition The decibel (dB)
A dB is a squished way to write huge ratios using a logarithm base 10:
( N S ) dB = 10 log 10 ( N S ) ⟺ N S = 1 0 ( dB /10 )
Why: ratios like 1000 or 1000000 are clumsy; log 10 turns "×10" into "+10 dB," compressing the scale.
Common mistake dB is already logarithmic
"30 dB" is not the ratio 30. Convert first: 1 0 30/10 = 1 0 3 = 1000 . Feeding 30 straight into Shannon is the classic error the parent warns about.
One held picture on the wire — one voltage level for one time-slot. With 2 levels a symbol = 1 bit; with 4 levels a symbol = 2 bits.
Definition Baud rate vs bit rate
Baud = symbols per second (how many pictures change per second).
Bit rate C = bits per second.
They link through §3:
C = baud × log 2 M
See Bandwidth vs Bit Rate . When M = 2 , log 2 2 = 1 , so baud = bit rate — which is why people confuse them.
Intuition The seed of Nyquist
A channel of bandwidth B can produce at most 2 B cleanly-distinguishable symbol changes per second. This "2 × " comes from the Sampling Theorem : a wave whose fastest wiggle is B Hz is fully pinned down by 2 B samples per second — sample slower and you miss wiggles, faster and you learn nothing new. Hence the Nyquist rate is 2 B symbols/sec.
You don't need the proof yet — just carry the fact: bandwidth B → up to 2 B symbols/sec. That single fact powers both Nyquist and Shannon in the parent note.
log base 2 = bits per symbol
Nyquist and Shannon capacity
Everything on the left feeds the single box on the right: the capacity laws of the parent note, the parent topic .
Cover the right side and test yourself — if any line stumps you, re-read its section above.
A signal graph: what is on each axis? Horizontal = time (seconds), vertical = voltage
What does 1 Hz mean? One full cycle (repetition) per second
Bandwidth B in one sentence? The width of the frequency range a channel passes, B = f high − f low , in Hz
Compute log 2 8 and say what it means. 3 — three bits are needed to label 8 distinct levels
How many bits does one symbol carry with M levels? log 2 M bits
Convert 20 dB to a linear S / N ratio. 1 0 20/10 = 1 0 2 = 100
Baud vs bit rate — when are they equal? When M = 2 (binary), since log 2 2 = 1
Where does the "2 " in 2 B come from? The sampling theorem — a bandwidth-B signal needs 2 B samples/sec to be captured