Field Guide · technology

Also known as: AMR, AMR-NB, AMR-WB

AMR (Adaptive Multi-Rate) is the family of speech codecs standardised by 3GPP for cellular telephony.1 At its core AMR is an ACELP coder — an algebraic code-excited speech model — but its defining feature is adaptivity: it can switch among eight bitrates frame by frame so that, when radio conditions worsen, bits shift from the voice codec to stronger channel coding while keeping the total rate constant. AMR became the mandatory codec for GSM and UMTS voice and, as AMR-WB, is the original “HD Voice” heard on VoLTE.

good channelpoor channel voice 12.2kFEC voice 4.75kFEC (more) same total gross bitrate — the split adapts
AMR keeps the gross bitrate fixed and reallocates bits between the ACELP speech coder and error protection as the radio channel changes.

How it works

AMR narrowband (AMR-NB) operates on 8 kHz speech in 20 ms frames and defines eight codec modes from 4.75 to 12.2 kbps, all built on ACELP. Each mode uses the same short-term LPC analysis and adaptive-plus-algebraic excitation; the modes differ mainly in how many algebraic pulses and how much gain/LSF precision they spend, so higher modes sound better but leave fewer bits for forward error correction. A source-controlled rate mechanism plus discontinuous transmission (DTX) with comfort-noise generation lets the encoder go silent between talkspurts to save capacity.

AMR-WB extends the same idea to a 16 kHz sampling rate, coding audio out to ~7 kHz with nine modes from 6.6 to 23.85 kbps. The wider band is what makes voices sound markedly clearer, and AMR-WB was standardised in parallel as ITU-T G.722.2. Crucially, the link — not the codec — decides the mode: the network signals a codec-mode request based on measured channel quality, so the two endpoints track conditions together.

Relevance to SDR

AMR is a cellular baseband codec, carried inside encrypted, tightly scheduled GSM/UMTS/LTE air interfaces rather than in the clear on land-mobile channels. It matters here as the reference point for how digital cellular voice is compressed, and as the direct ancestor of 3GPP EVS, which keeps an AMR-WB-compatible mode for interworking. GopherTrunk targets trunked land-mobile radio and does not decode cellular traffic, so AMR is context rather than a codec in GopherTrunk’s chain — the trunking systems it does follow use MBE-family or, for TETRA, plain ACELP vocoders instead.

Sources

  1. Adaptive Multi-Rate audio codec — Wikipedia, on AMR’s ACELP core, the eight narrowband modes, DTX, and its role in GSM/UMTS. ↩

See also