Interactive · Spectrum
This spectrum is not a picture. It is computed — random dibits, real pulse shaping, real frequency modulation, a real FFT — from nothing but the deviation and the filter roll-off. Change either and watch the skirts move. Then slide a neighbour in and see what "adjacent channel" means.
How this is made. Random dibits are mapped to the four symbol levels (±1 and ±3 × the base deviation), upsampled, shaped with a root‑raised‑cosine filter, integrated into a phase trajectory, and frequency‑modulated onto a complex baseband carrier. That waveform is windowed, FFT'd, and averaged over many blocks. Nothing on this chart was traced from a datasheet — it falls out of the modulation parameters. The parameters themselves are verified: DMR runs 4,800 symbols/s with deviations of ±648 and ±1944 Hz and RRC α = 0.2 (ETSI TS 102 361-1, corroborated by the Tait DMR guide, VK4PK's MMDVM notes, an open-source modem, and a Viavi service-monitor manual). M17 runs 4,800 symbols/s with ±800 and ±2400 Hz, deviation index h = 1/3, RRC α = 0.5, in a declared 9 kHz bandwidth (the official M17 protocol specification). The check that matters: load the M17 preset and read the 99% bandwidth. It computes to about 8.7 kHz — against the 9 kHz the M17 spec declares. The spec's own number falls out of the physics, which is the only reason I trust the rest of the chart. Not modelled: transmitter non-linearity, phase noise, power-amplifier splatter, and the real regulatory emission masks — all of which make a real signal dirtier than this one. 73 de N6JET