Field Guide · technology

Also known as: EVS, Enhanced Voice Services

EVS (Enhanced Voice Services) is the 3GPP speech and audio codec introduced with LTE to carry telephone calls at far higher quality than earlier cellular vocoders.1 It is the codec behind “HD Voice+” and is used for VoLTE and 5G VoNR calls. EVS spans narrowband all the way to full-band (up to 20 kHz audio), degrades gracefully under packet loss, and includes an AMR-WB interoperable mode so it can hand calls off to legacy networks without transcoding.

NB 4kWB 8kSWB 16kFB 20k AMR-NB / AMR-WB EVS (NB → FB)
EVS covers audio bandwidths from narrowband to full-band in one codec, extending well beyond the range that AMR reaches.

How it works

EVS is a switched codec: it chooses, per frame, between an ACELP-based coder for speech and an MDCT transform coder for music and generic audio, so a single codec handles both cleanly. It supports four audio bandwidths — narrowband, wideband, super-wideband, and full-band — at bitrates from 5.9 kbps up to 128 kbps, with a variable-rate mode that can drop to comfort noise during silence. Backward compatibility is built in through an AMR-WB IO (interoperable) mode that is bit-exact with AMR-WB, letting a 5G/LTE call cross into a 3G network without a lossy transcode.

Two robustness features set EVS apart. Its packet-loss concealment reconstructs missing frames well enough that occasional loss is inaudible, and a channel-aware mode proactively embeds partial redundancy of earlier frames so the decoder can recover a lost packet from a later one — essentially source-integrated forward error correction tuned for VoIP jitter and loss.

Relevance to SDR

Like AMR, EVS lives inside the cellular packet core (VoLTE/VoNR) rather than on open land-mobile channels, so it is not something a trunking scanner decodes. It appears in this guide as the current state of the art in cellular voice coding and as the natural comparison point to open-source VoIP codecs such as Opus, which occupy a similar speech-plus-audio, loss-resilient design space. GopherTrunk does not implement EVS; its decode chain is land-mobile digital voice, whose vocoders are the MBE family and ACELP rather than EVS.

Sources

  1. Enhanced Voice Services — Wikipedia, on EVS bandwidths, bitrates, the ACELP/MDCT switch, AMR-WB IO, and channel-aware mode. ↩

See also