Constellations & symbol mapping

Key takeaways A constellation is the set of ideal points a modulation uses on the I/Q plane. PSK places points around a circle (varying phase); QAM fills a grid (varying phase and amplitude). Each point encodes a group of bits; Gray coding ensures neighbours differ by one bit so a slip costs one error. Plotting received symbols against the ideal points — the constellation diagram — is the fastest read on link health.

Demodulation turned I/Q into symbols; this lesson is the map from those symbols to bits, and the picture that shows how well it’s working. It connects to the RF path’s digital modulation lesson.

Symbols as points on a plane

Each symbol period, the transmitter sends one point from a fixed set. Because a baseband sample is a complex I/Q value, that point has a position on the plane — a phase (angle) and amplitude (radius). The receiver, once carrier-locked, reads the incoming point and decides which ideal point it is nearest. That decision is the symbol; a lookup turns the symbol into its bits.

More points per constellation carry more bits per symbol — but pack the points closer, so noise defeats them sooner.

PSK and QAM: two families

Scheme	Varies	Points	Bits/symbol
BPSK	phase	2	1
QPSK	phase	4	2
16-QAM	phase and amplitude	16	4

PSK keeps every point at the same radius and spreads them by angle — robust, because amplitude carries no data. QAM uses a grid, squeezing more bits per symbol by using amplitude too, at the cost of points sitting closer together and thus being easier for noise to confuse. The four-level C4FM of P25 Phase 1 is a frequency-domain cousin of the same bits-per-symbol idea.

Gray coding: one slip, one bad bit

How you label the points matters as much as where they sit. Noise most often bumps a symbol into an adjacent point, so the bit labels are assigned by Gray coding — neighbours differ in exactly one bit. Then the common single-step slip flips only one bit, keeping the bit error rate far below what a careless labelling would give.

QPSK, Gray-coded:   00 | 01        a slip to any *neighbour*
                    ----+----      changes exactly ONE bit,
                    10 | 11        never two.

The diagram as a diagnostic

Plot the received symbols and the pattern is the diagnosis. Tight dots on the ideal points: healthy. A fuzzy cloud: noise — low SNR. A pattern rotating or held at an angle: an uncorrected carrier offset. Points smeared radially: multipath. The spread of each cluster from its ideal point is measured as EVM, the next unit’s first metric.

Quick check: why are constellation points Gray-coded?

Recap

A constellation is the set of ideal I/Q points a modulation uses; symbols map to bits.
PSK varies phase (points on a circle); QAM varies phase and amplitude (a grid).
Gray coding makes neighbours differ by one bit, minimizing the bit error rate.
The constellation diagram is a fast diagnostic — its shape names the impairment.

Next up: why reflections smear symbols together, and the equalizer that undoes a channel.

Frequently asked questions

What is a constellation diagram?

A constellation diagram is a scatter plot of received symbols on the I/Q plane. Each modulation scheme defines a set of ideal points — the constellation — and every received symbol lands near one of them. Tight clusters right on the ideal points mean a clean signal; a fuzzy, spread-out cloud means noise, and a rotated or warped pattern points to a specific impairment.

Why is Gray coding used to map bits to symbols?

Gray coding arranges the bit patterns so that adjacent constellation points differ by only a single bit. When noise nudges a symbol into a neighbouring point — the most common kind of error — only one bit flips instead of several. That keeps the bit error rate low for a given symbol error rate, which is why almost every real system uses it.