Skip to content

Release Notes — v0.7.10

RTCM3 MSM: BeiDou-3 B2b signal IDs restored (decoder and encoder)

Release date: 2026-09-22 Type: Bug fix (RTCM3 MSM BeiDou signal table) — no positioning change in the regression suite Branch: release/v0.7.10

Overview

Field comparisons on RENAG stations and a mosaic-X5 receiver (#333, reported by @philippebourcier) showed that MRTKLIB decoded no BDS-3 B2b observations from RTCM3 MSM streams that carried them. The cause: the BeiDou MSM signal-ID → RINEX code table msm_sig_cmp[32] in src/rtcm/mrtk_rtcm3.c was byte-identical to the MALIB parent and left signal IDs 25–27 empty. RTCM 10403.3 Amendment 2 assigns those IDs to the B2b signals 7D / 7P / 7Z, so every B2b cell resolved to CODE_NONE and was skipped with only a level-2 trace (unknown signal id=N).

The encoder was affected too. to_sigid() in src/rtcm/mrtk_rtcm3e.c performs the reverse lookup against the same table, so RTCM3 output built from SBF / NovAtel / BINEX input — whose decoders already emit CODE_L7D — silently lost B2b as well.

Fix

PR #335 adds the three missing entries ("7D", "7P", "7Z" at IDs 25–27) and replaces the stale "Tentative IDs for B1C and B2a" comment with the source citation. The change is table-only; the decoder (save_msm_obs) and encoder (to_sigid) paths are untouched.

Provenance: PocketSDR's reading of table 3.5-108 lists exactly these three at IDs 25–27 as standard entries; demo5 carries the same three (marking 7P/7Z tentative) and MADOCALIB has 7D only. The further tentative PocketSDR / demo5 extensions (1S/1L/6D/8D, …) are deliberately not imported.

docs/dev/pitfalls-public.md P-23 now records that the encoder shares the per-constellation tables (a missing entry drops the signal in both directions with no error) and that the tables must track RTCM amendments.

Verification

  • Red → green. New utest_msm_bds_b2b encodes one BDS epoch (BDS-3 C21 with 2I + 7D/7P/7Z, BDS-2 C06 with 2I + 7I as controls) as MSM7 (type 1127) via gen_rtcm3() and feeds it byte-by-byte into a second rtcm_t via input_rtcm3(). On the unmodified table exactly the three B2b presence checks fail while the 2I/7I controls pass (67-byte frame); after the fix all checks pass (98-byte frame; P deltas < 2e-4 m, L < 6e-5 m, SNR exact). This is the suite's first MSM encode→decode round-trip coverage (#296).
  • Mutation check (wire IDs). The encoded MSM header signal mask is inspected directly: the combined frame must set exactly bits {2, 14, 25, 26, 27}, and one-signal frames must each set only their own bit. A swapped table entry (7D ↔ 7P) fails here even though the symmetric round-trip and the combined-frame ID set would both still pass.
  • Full regression gate: 122 tests, 121 pass; the single failure is the documented environment-only madocalib_pppar_ion_check (LAPACK-vs-reference ~1.6 cm vs 0.5 cm tolerance, reproduces on clean develop). All RT replays passed at full length. No positioning test in the suite decodes BDS MSM observations (the only .rtcm3 positioning input is IGS-RTS SSR), so the suite is unchanged by construction; the round-trip test is the coverage.
  • Review: Copilot approved with no findings; Greptile's P2 (pin the wire IDs independently of the shared table) is the mutation check above.

Behaviour note (slot placement)

Decoded 7D takes the B2 slot (index 2), not an extended slot: sigindex() applies the per-band code priority (7D > 7I > 7Q > 7X, src/data/mrtk_obs.c) across the signal list of each MSM message. In a message that also carries BDS-2 7I (typical mosaic-X5 base), 7I moves to an extended slot (7P / 7Z go to extended slots). This is the same placement the RINEX reader's set_index() already produces for C7D / C7I from the same receiver, so RTCM input now matches the RINEX path; it is also upstream RTKLIB's per-message semantics. Override lever: RTCM option -CL7I (getcodepri()); sigcfg does not govern RTCM3 observations yet.

  • #333 — fixed.
  • #296 — partially addressed (first MSM round-trip test; broader RTCM3 coverage remains open).
  • #189 / #234 — follow-ups: positioning use of B2b and sigcfg control of RTCM3 observation slots.

Upgrade notes

None — table-only change: no struct layout change, no configuration change.