Also known as: G.722
G.722 is the ITU-T wideband speech codec that carries roughly 50 Hz–7 kHz audio at 64 kbps — double the audio bandwidth of narrowband telephony — using sub-band adaptive differential PCM (SB-ADPCM).1 Standardised in 1988, it was the first widely deployed “HD Voice” codec, and because it fits into the same 64 kbps as G.711 it slots directly into existing telephony infrastructure while sounding noticeably clearer.
How it works
G.722 samples audio at 16 kHz and uses a quadrature mirror filter (QMF) to split the signal into two equal sub-bands: a lower band (0–4 kHz) carrying most of the speech energy and an upper band (4–8 kHz) carrying the high-frequency detail that makes wideband sound “HD”. Each sub-band is then coded independently with ADPCM — adaptive differential PCM, which transmits the quantised difference from a predicted sample and adapts its quantiser step to the signal level, exploiting the redundancy between successive samples.
The bit budget is deliberately lopsided: the lower band is coded at up to 6 bits per sample and the upper band at 2 bits per sample, reflecting where the ear needs precision. This gives three operating modes — 64, 56, and 48 kbps — where the lower rates free 8 or 16 kbps of the channel for an auxiliary data path while dropping to 5- or 4-bit lower-band coding. Compared with the lower-complexity, narrowband G.711, G.722 trades a little more processing for a much wider, more natural-sounding voice at the same channel rate.
Relevance to SDR
G.722 is a wireline and cordless-telephony codec — used in VoIP, video-conferencing, and DECT CAT-iq handsets — not an over-the-air radio waveform, so it is never decoded from RF by a scanner. It appears in this guide as the classic example of sub-band ADPCM and as the first mainstream wideband voice codec, a useful contrast with the low-bitrate vocoders used on digital land-mobile radio. GopherTrunk does not use G.722 in its decode chain.