Also known as: pulse shaping
Pulse shaping is filtering each transmitted symbol’s pulse so the signal occupies less bandwidth and successive symbols don’t smear into one another (inter-symbol interference).1 Sharp rectangular pulses spray energy into adjacent channels; a shaped pulse keeps it contained. It is a mandatory step in almost every band-limited digital radio, because the two goals — a narrow spectrum and no ISI — are in tension and pulse shaping is how they are reconciled.
How it works
A digital symbol stream is a sequence of impulses, one per symbol. Sent as-is — or as flat rectangular pulses — the signal’s spectrum is a sinc with slowly decaying sidelobes that leak far into neighbouring channels, causing “spectral splatter.” The naive fix is to lowpass-filter hard, but any filter that narrows the spectrum also spreads each symbol’s pulse in time, so it now overlaps its neighbours at their sampling instants and corrupts them: that is ISI. Pulse shaping resolves the dilemma by choosing a pulse whose spectrum is compact and whose time-domain shape has zeros at every other symbol’s sampling instant.
That second property is the Nyquist ISI criterion. A pulse satisfying it is non-zero at its own symbol time but exactly zero at all the neighbouring symbol times, so however much its tails ring, they contribute nothing when each symbol is sampled at the right instant — the eye stays open. The universal family that meets the criterion with a tunable bandwidth is the raised-cosine pulse, whose excess bandwidth beyond the theoretical minimum is set by the roll-off factor β (0 = sharpest spectrum, hardest to build; higher β = gentler, wider, more robust to timing error).
Variants
In practice the raised-cosine response is split between transmitter and receiver: each applies a root-raised-cosine (RRC) filter, and the cascade of the two — the transmit shaping filter followed by the receive matched filter — multiplies to the full raised-cosine response. That arrangement simultaneously satisfies the Nyquist criterion (zero ISI) and maximises signal-to-noise ratio at the sampling instant, which a single filter at either end cannot do. Other shapes exist — Gaussian pulses (used by GFSK and GMSK, which deliberately accept a little ISI for an even tighter spectrum), and simple sinc or rectangular pulses in undemanding links — but RRC dominates linear PSK/QAM systems.
In practice
Real systems specify the shaping filter exactly so that transmitter and receiver agree. P25 C4FM uses a shaping filter with a raised-cosine-like response; TETRA’s π/4-DQPSK uses RRC with roll-off 0.35; many cellular and satellite links quote a specific β (0.2–0.35) as part of the air interface. The chosen roll-off is a direct spectrum-versus-robustness decision: tighter roll-off packs channels closer but leaves less timing margin in the eye diagram.
Relevance to SDR
A software receiver must apply the matching receive filter — usually the same RRC the standard specifies — to see a clean constellation and open eye; skipping it or using the wrong roll-off leaves residual ISI that smears the symbols and raises the error rate. GopherTrunk’s demod chain includes the receive-side matched pulse-shaping filter for the linear modulations it handles, sized to the protocol’s symbol rate so that the combined transmit-plus-receive response meets the zero-ISI Nyquist condition at the slicer.
Sources
-
Pulse shaping — Wikipedia, for the bandwidth-limiting and inter-symbol-interference rationale. ↩