Field Guide · term

Also known as: voice superframe, DMR voice superframe

A DMR voice superframe is the repeating unit of a DMR voice call: six bursts labelled A–F, spanning 360 ms, carrying 18 on-air AMBE+2 voice frames of 72 bits each.1 Burst A is framed by a voice sync word; bursts B–F carry no sync and instead put embedded signalling in their central field, so they are located not by a sync search but by TDMA cadence relative to burst A.2

A · sync B · LC1 C · LC2 D · LC3 E · LC4 F each burst = 3 × 72-bit AMBE+2 frame → 18 frames / superframe bursts B–E fragments 1–4 → one 72-bit embedded Link Control 360 ms · TDMA cadence locates B–F from burst A
Six bursts carry 18 AMBE+2 voice frames; burst A is sync-framed while bursts B–E each contribute one of the four embedded link-control fragments, and burst F carries voice only.

Structure

GopherTrunk decodes the superframe in internal/radio/dmr/voice/superframe.go into a VoiceSuperframe holding 18 AMBE frames in transmission order (bursts A–F, three per burst). The central-field roles across the six bursts are:

Burst Central field Voice frames
A voice sync word (anchors the frame) 3
B embedded LC fragment 1 (LCSS First) 3
C embedded LC fragment 2 (Continuation) 3
D embedded LC fragment 3 (Continuation) 3
E embedded LC fragment 4 (LCSS Last) 3
F embedded signalling (no LC fragment) 3

The four fragments from bursts B–E reassemble into a 72-bit embedded Link Control; when it passes its BPTC + checksum, its destination and source addresses label the call. A single-fragment (LCSS == Single) field in any of those bursts instead carries a Reverse Channel word or the null idle.

Cadence and phase lock

Only burst A produces a sync match, so the decoder locks onto it and slices B–F at a fixed same-slot stride. On a contiguous single-slot stream that stride is 132 dibits; on a real 2-slot TDMA carrier the other timeslot’s burst sits between each of a call’s bursts, making same-slot bursts 264 dibits apart (no inter-burst CACH) or 288 when a 12-dibit CACH precedes each burst on live outbound air. NewInterleavedDecoder auto-detects which stride is in use per call — a CRC-valid embedded LC is authoritative and locks immediately, and absent one it scores each candidate by AMBE FEC quality (the Golay-corrected-bit count across all 18 frames) and locks the clearly-best cadence. That AMBE-quality fallback is what let a 288-cadence CACH carrier decode even when its embedded LC never did (issue #644). Each superframe also carries a Phase — the burst-A anchor’s parity over the physical-burst period — which is a stable relative discriminator between the two interleaved calls, not an absolute TS1/TS2 label.

Relevance to SDR

The superframe is the frame GopherTrunk’s DMR voice pipeline actually operates on: locking burst A, slicing the six bursts at the right cadence, pulling 18 AMBE+2 frames for the vocoder, and reassembling the embedded LC to name the call. Handling the 264-vs-288 cadence and the phase lock correctly is what keeps a decoder from pulling dibits out of the wrong timeslot on a busy carrier — the difference between clean audio for one talkgroup and a garble of two interleaved calls. Because burst A’s BS-sourced sync is identical on both slots, the embedded LC (or Phase bound to a talkgroup) is the only thing that tells the two apart.

Sources

  1. Digital mobile radio — Wikipedia, on the DMR standard and its voice superframe. ↩

  2. Time-division multiple access — Wikipedia, on the two-slot framing the superframe’s cadence rides. ↩

See also