Skip to content

Release Notes — v0.7.5

Single-band CLAS PPP-RTK — regression coverage and capability record

Release date: 2026-07-06 Type: Test coverage + capability record (no positioning change) Branch: release/v0.7.5

Overview

v0.7.5 is not a feature release — there are no positioning-code changes. It locks in two single-band CLAS PPP-RTK configurations that the v0.7.4 engine already handles, backing each with a regression test anchored to independent truth, and records the mechanics in a new design note.

The v0.7.4 engine already positions these configurations correctly without any dedicated single-frequency machinery: the ionosphere Gauss-Markov decay bounds the unobservable per-satellite iono residual, so a single-band solution stays well-behaved. This release captures that behavior as tests so it cannot silently regress, and documents why it works.

The two configurations are:

  1. Single-frequency (nf=1) CLAS PPP-RTK.
  2. Galileo E1/E5b band mismatch — u-blox ZED-F9P-class receivers tracking E1+E5b under CLAS E5a-only biases.

1. Regression coverage for single-band CLAS PPP-RTK

Two new claslib test triples lock in the v0.7.4 behavior. Each triple pairs a consistency reference with an absolute check against independent truth.

Single-frequency (nf=1)

  • New config conf/claslib/rnx2rtkp_sf.toml.
  • 2019/239 dataset, validated against GSI F5 truth.
  • Absolute accuracy: 2D 1σ 2.5 cm / 95 % 3.9 cm.

Galileo E1/E5b band mismatch

  • ZED-F9P-class receivers track E1+E5b, but CLAS carries E5a-only biases, so the mismatched satellites run E1-only.
  • Self-collected ECJ02 dataset, validated against an independent IGS-products coordinate.
  • Absolute accuracy: 3D 1σ 18.5 cm / 95 % 20.6 cm.
  • The E1-only satellites bound the solution at dm-level — this is not DF-equivalent accuracy, and the test enforces that bound rather than a dual-frequency target.

2. Tooling

compare_nmea_abs.py gains a --llh fixed-coordinate option, so an absolute accuracy check can be anchored to a supplied lat/lon/height (the ECJ02 IGS-products coordinate) rather than an averaged solution.

3. Design record

docs/design/sf-ppp-rtk.md records:

  • The mechanics of single-band CLAS PPP-RTK, and the dependency on the ionosphere Gauss-Markov decay that bounds the unobservable iono residual.
  • A post-mortem of the explicit SF iono constraint that was prototyped, then subsumed by the v0.7.4 engine changes, and deliberately not shipped.

Compatibility & migration

No configuration changes are required for existing workflows; the new conf/claslib/rnx2rtkp_sf.toml is additive. There are no positioning-algorithm, matrix, physical-constant, or test-tolerance changes — v0.7.5 only adds tests, one tooling option, and documentation.

Issues / PRs

  • PR #253 — regression coverage, compare_nmea_abs.py --llh, and docs/design/sf-ppp-rtk.md.
  • PR #254 — release.

References

  • docs/design/sf-ppp-rtk.md — single-band CLAS PPP-RTK behavior record and the not-shipped SF iono-constraint post-mortem.