Signal Encoding

The Problem

You have a stream of bits: 1 0 1 1 0 0 1

You need to send this as electrical voltage over a wire.

The simplest idea: high voltage = $1$, low voltage = $0$.

This is called NRZ-L (Non-Return-to-Zero, Level).

NRZ: Simple But Flawed

The problem: when you send many $1$s in a row, the voltage just stays high.

The receiver sees a flat line. How does it know how many $1$s there were?

Without transitions, the receiver has no way to count bit boundaries.

The Synchronization Problem

The transmitter and receiver need synchronized clocks to agree on bit timing.

But clocks drift. Even a tiny mismatch compounds over time.

With a long run of identical bits:

• Transmitter: “I sent eight $1$s”
• Receiver: “I counted… seven? Nine?”

We need transitions in the signal to keep clocks synchronized.

Manchester Encoding

The clever solution: instead of voltage levels, use transitions.

• $1$ = transition from low to high (in the middle of the bit)
• $0$ = transition from high to low (in the middle of the bit)

Every single bit has a transition in the middle.

This means:

• Self-clocking: the receiver recovers timing from the signal itself
• No flat lines: guaranteed transition every bit period
• No DC component: average voltage is zero

The Tradeoff

Manchester encoding uses twice the bandwidth of NRZ.

Why? Because there’s a potential transition at every bit boundary AND in the middle of each bit.

Manchester is used in 10BASE-T Ethernet where reliability matters more than bandwidth efficiency.

The DC Component Problem

There’s another issue with NRZ that Manchester also solves: DC offset.

If you transmit more $1$s than $0$s, the average voltage isn’t zero. This DC component causes problems:

• Transformers block DC: many transmission systems use transformers that can’t pass constant voltage
• Capacitors charge up: DC builds up and distorts the signal over time
• Wasted power: maintaining a constant offset wastes energy

Manchester naturally fixes this: every bit spends equal time at high and low voltage, so the average is always zero.

Can We Get DC-Free Without Doubling Bandwidth?

Manchester solves both problems (sync and DC), but at the cost of 2x bandwidth.

What if we want DC-free transmission without that penalty?

This leads us to a different approach: using three voltage levels instead of two.

AMI: A Different Approach

Alternate Mark Inversion uses three voltage levels: +V, 0, -V

The rules:

• $0$ = zero voltage
• $1$ = alternates between +V and -V

So 1 0 1 1 0 1 becomes: +V, 0, -V, +V, 0, -V

Why Alternate?

By alternating polarity on each $1$, the positive and negative voltages cancel out.

No DC component: the average voltage is zero.

Bonus: if you ever see two +V in a row, you know an error occurred. Built-in error detection!

AMI’s Weakness

AMI still has a problem with long runs of zeros.

If you send 0 0 0 0 0 0 0 0, the signal stays at zero voltage. No transitions, same synchronization problem as NRZ.

This led to variants like B8ZS and HDB3 that insert artificial transitions during long zero runs.

Comparison

When To Use What

NRZ: Simple, efficient, but only for short distances or when you have a separate clock signal.

Manchester: When you need guaranteed synchronization. Used in Ethernet.

AMI: When you need DC-free transmission (like over telephone lines) and can handle the occasional long-zero problem.

The right encoding depends on your channel characteristics and requirements.