GNSS-SDR v0.0.22 released

GNSS-SDR v0.0.22 is a major step forward for the project. The receiver now processes four new signals: BeiDou B1C and B2a, QZSS L1 C/B, and SBAS L1 (EGNOS and WAAS, not yet used as ranging sources). The PVT block gains real-time kinematic (RTK) positioning fed by RTCM 3 corrections from NTRIP casters, enabling centimeter-level relative positioning under suitable observing conditions, together with a new integer ambiguity resolution management layer. Observation and navigation files can now be written in RINEX 4.02 format, and a new CUDA acquisition engine can offload the PCPS search grid to NVIDIA GPUs, including Jetson platforms. Support for RF hardware grows with new signal sources for bladeRF (without going through gr-osmosdr), Pocket SDR, and SAPHYRION EVK1029 front-ends, and the Galileo OSNMA implementation now handles chain, public key, and Merkle tree renewals and revocations, as well as alert messages. This release also fixes several biases in the group-delay and ionospheric corrections applied by the single-point positioning solver.

Many of these features are community contributions, credited in the list below. The most relevant changes with respect to the former release are:

Improvements in Accuracy:

• Added real-time kinematic (RTK) positioning using RTCM 3 corrections received from NTRIP casters, enabling centimeter-level relative positioning under suitable observing conditions.
• Fixed the sign of the group-delay correction applied to GPS and QZSS L1+L2C dual-band observations in the SPP solver when a single-frequency ionospheric model is used: the L1 C/A pseudorange is now corrected as P1 - c*TGD, as specified in IS-GPS-200 20.3.3.3.3.2, instead of P1 + c*TGD, removing a per-satellite bias of twice the broadcast group delay (up to several meters).
• Fixed the GPS L1+L5 dual-frequency ionosphere-free correction in the SPP solver: the gamma-weighted L1 term now applies ISC_L1C/A as specified in IS-GPS-705 20.3.3.3.1.2.2, instead of erroneously reusing ISC_L5I5 on both terms of the combination.
• The variance of the broadcast ionospheric delay estimate is now scaled consistently with the delay itself when converting from the GPS L1 frequency to the L1/B1 frequency of other constellations (delay scales with f^-2, its variance with f^-4).
• Fixed a double-counting of the ionospheric delay in the Single Point Positioning (SPP) solver for dual-band satellites processed with a single-frequency ionospheric model (e.g., PVT.iono_model=Broadcast): the ionosphere-free pseudorange combinations formed for GPS/QZSS L1+L5 and Galileo/GLONASS/BeiDou dual-band observations no longer get the modeled ionospheric delay applied on top, which biased those residuals by the (elevation-dependent) modeled delay. The measurement variance of these combinations now also receives the same noise-amplification factor already used in the ionosphere-free positioning mode, instead of a Klobuchar-based variance term that did not correspond to the measurement.
• The modeled ionospheric delay in the SPP solver is now scaled to the observed frequency band for single-band measurements outside the L1/E1/B1 band (GPS L2C-only or L5-only, Galileo E5a-only or E5b-only, GLONASS L2-only, BeiDou B3I-only): the L1 delay is multiplied by (f_L1/f_band)^2 (about 1.65 for L2 and 1.79 for L5/E5a), and its variance by the square of that factor. Previously the unscaled L1 delay was applied to those measurements, undercorrecting the ionosphere by the same factor.
• Improved PVT processing of GPS L2C, GPS L5, and QZSS signals using CNAV navigation data: satellite positions now include the CNAV semi-major axis and mean-motion rate terms, and group-delay / inter-signal corrections follow IS-GPS-200 / IS-GPS-705 in both single-band and L1+L5 dual-band configurations. Contributed by @vladisslav2011.
• Carrier-phase discontinuities are now detected and flagged: cycle slips after a signal reacquisition, and half-cycle jumps caused by a change in the telemetry-resolved phase polarity. RINEX observation files report them with standard loss-of-lock indicator values, and the optional carrier smoothing filter restarts instead of smoothing across the jump.

Improvements in Availability:

• Added Acquisition_XX.full_grid_search (default: false) for acquisition implementations using the CPU PCPS block. When enabled, each search stage accumulates all max_dwells non-coherent integrations before accepting or rejecting the strongest peak. This also applies to both stages of make_two_steps and to narrowed Doppler searches. The default preserves early acceptance; max_dwells=1 is unchanged. bit_transition_flag=true takes precedence and still uses a single double-length dwell. Waiting for all dwells increases acquisition latency. Contributed by @joebre.
• Improved TOW rollover handling in Telemetry Decoder blocks.
• Galileo F/NAV and I/NAV ephemerides are now retained independently instead of overwriting each other when they have the same PRN. PVT automatically uses the ICD-consistent service for the enabled bands (E1/E5a uses F/NAV; E1/E5b uses I/NAV, with I/NAV taking priority when E5a and E5b are both enabled), while RINEX, RTCM MT1045, monitoring, and assistance-data persistence preserve the navigation-message source. XML persistence keeps the legacy gal_ephemeris.xml view and automatically adds gal_inav_ephemeris.xml and gal_fnav_ephemeris.xml; no configuration change is required.
• Galileo E1 observations can now use the I/NAV ephemeris while F/NAV is still being decoded in E1/E5a configurations, with the E1/E5b BGD applied to match that clock model, and return to F/NAV once it is available. This removes the cold-start delay in which Galileo could not contribute to PVT until F/NAV was fully decoded on E5a. E1 can also use Reduced CED in the same situation. E5a, E5b and E6 observations always keep the clock reference of their configured service. The reverse fallback (E1 using F/NAV when I/NAV is stale) requires use_unhealthy_sats=true, since F/NAV carries no E1B health information.
• The PVT iono model now decides the pseudorange model for every satellite of a system: only PVT.iono_model=Iono-Free-LC combines two bands, and any other model uses the first band alone with its TGD/BGD (second band alone only when the first is missing). Previously a satellite with two bands in the record was silently switched to the iono-free combination while single-band satellites of the same system kept the single-frequency model, mixing two clock references within one solve and folding the receiver’s uncalibrated inter-band delay (e.g. differing input-filter group delays) into the solution, which could make the chi-square test reject every epoch. The ISC-aware GPS L1/L5 combination is now applied in Iono-Free-LC mode.
• Fixed Galileo single-frequency broadcast group-delay corrections in SPP and PPP, selecting the E1/E5a or E1/E5b BGD from the active navigation service and applying the ICD frequency scaling to E5a and E5b observations.
• Hardened Galileo I/NAV and F/NAV handling by rejecting alert pages from the nominal decoder, validating unavailable GGTO data, and preventing stale Word 5 data from completing Reduced CED or Reed-Solomon-recovered ephemerides.
• Improved the availability of navigation data by making histogram-based bit synchronization more resistant to weak or ambiguous prompt transitions, which could select the wrong bit-boundary phase, prevent telemetry frame synchronization, and delay TTFF. Candidate edges are now scored with normalized coherent prompt averages, require a configurable margin over the second-best phase bin, and are validated with fresh matching transition events. Histogram stability and tentative-lock validation advance concurrently to avoid unnecessary synchronization delay. New configuration parameters are Tracking_1C.bs_runner_up_margin (default: 0.3), Tracking_1C.bs_transition_window_epochs (default: 4), Tracking_1C.bs_transition_confidence (default: 0.6), and Tracking_1C.bs_tentative_events_required (default: 2).
• Added an optional frequency-refinement scan to help tracking lock onto signals whose initial Doppler estimate is displaced by a navigation-bit or secondary-code transition during acquisition, particularly Galileo E1. Enable it per signal with Tracking_<Sig>.f_error_step_num (default: 0, disabled). This selects the number of Doppler bins around the acquisition estimate; even nonzero values are rounded up to an odd count. f_error_doppler_step sets their spacing (default: 250 Hz), and f_error_accumulation sets the code periods accumulated per bin (default: 20; zero is replaced with one, with a warning). The scan adds a startup delay of one code period per accumulation per bin and supports high_dyn=true. The pull_in_time_s and bit_synchronization_time_limit_s budgets start after the scan, allowing the tracking loops their full settling time. Contributed by @joebre.
• Added an optional CSV dump of the frequency-refinement scan, enabled with Tracking_<Sig>.f_error_dump=true (default: false). The tested Doppler frequencies, their correlation power, and the selected frequency are written to Tracking_<Sig>.f_error_dump_filename (default: ./f_error_dump.csv). Channels sharing a filename write to the same file, with scan, satellite and channel identifiers; the first scan overwrites any previous file, and later scans in the same receiver run append their results. The Octave scripts load_f_error_dump.m, find_f_error_scans.m and plot_f_error_scan.m (in utils/matlab/libs) and plot_all_f_error_scans.m (in utils/matlab) load and plot the dumped scans, and utils/matlab/libs/f_error_sim.m provides a Monte Carlo simulation of the scan for sizing f_error_step_num, f_error_accumulation and f_error_doppler_step without a live capture.

Improvements in Efficiency:

• Optimized CPU DLL/PLL VEML tracking for pilot/data signal pairs by computing the pilot correlators and the data prompt in one multicorrelator pass. The data prompt now reuses the same carrier wipe-off as the pilot correlators instead of invoking a separate one-tap correlator, while non-pilot tracking keeps the previous correlation path.
• When dual-frequency assistance provides the Doppler of a satellite already tracked in the primary band (GNSS-SDR.assist_dual_frequency_acq=true), the PCPS acquisition in the secondary band now searches a single Doppler bin instead of the full grid, and recalibrates the pfa-based threshold to the number of bins searched. New parameter Acquisition_XX.reference_bin_min_sidelobes (default: 4) sets the Doppler separation, in correlation sidelobes, that decides whether a full-grid CFAR search needs dedicated noise-reference bins. Acquisition .mat dumps include doppler_center, doppler_narrowed, and doppler_num_candidates: the first doppler_num_candidates columns of acq_grid are Doppler bins at doppler_center - doppler_max + doppler_step * col, and any remaining columns are noise-reference bins. Contributed by @joebre.
• Added an optional visibility-aware acquisition search, enabled with GNSS-SDR.enable_visibility_aware_search=true (default false, which leaves the existing search order untouched). Once a receiver position is available, either from a fix or from GNSS-SDR.AGNSS_ref_location, satellites are continuously classified as visible, excluded (elevation at or below GNSS-SDR.search_elevation_mask, default 0 degrees, or flagged unhealthy), or not yet known, using the freshest ephemeris or almanac decoded for GPS, Galileo, BeiDou, GLONASS, and QZSS. Idle channels then favor visible satellites over unknown ones, at the ratio given by GNSS-SDR.visible_vs_mayvisible_search_ratio (default 3), and skip known excluded ones, so less CPU is spent acquiring satellites that are below the horizon. The classification is refreshed whenever new navigation data arrives, when the receiver moves more than GNSS-SDR.visibility_recompute_position_threshold_m (default 1000 m), every GNSS-SDR.visibility_recompute_interval_s (default 120 s), and when almanac data becomes older than GNSS-SDR.visibility_almanac_max_age_s (default 3 days). A satellite that is already being tracked is never released because of this classification, and PVT.elevation_mask still decides which observations enter the navigation solution. Contributed by @joebre.
• Added opt-in almanac/ephemeris Doppler prediction for secondary signals with Acquisition_<signal>.alm_ephe_assisted_doppler_narrowing=true (default false, also supported per channel). To acquire secondary signals without waiting for a tracked primary band, set GNSS-SDR.assist_dual_frequency_acq=false. Pre-fix prediction additionally requires GNSS-SDR.doppler_prediction_before_fix=true, an AGNSS_ref_location (and AGNSS_ref_utc_time for replay), and explicit, finite, nonnegative values for both GNSS-SDR.clock_frequency_max_error_ppm and GNSS-SDR.receiver_max_velocity_m_s. Missing or invalid bounds preserve the full Doppler search; explicit zero bounds assert no uncertainty in that component. The predicted center uses GNSS-SDR.clock_frequency_offset_ppm (default 0) and zero receiver velocity before a fix. Search widening honors --doppler_max and --doppler_step overrides, and falls back to the regular full search centered at 0 Hz when the uncertainty window is not narrower than the configured Doppler grid. Live-fix prediction refreshes its timestamp before both idle and channel-event acquisition attempts.
• New CUDA acquisition engine: with -DENABLE_CUDA=ON, any PCPS acquisition block can evaluate its Doppler x code-phase search grid on the GPU with batched cuFFTs by setting Acquisition_XX.use_cuda=true (or GNSS-SDR.use_cuda_acquisition=true). Peak search and detection statistics are unchanged, so results match the CPU implementation; the block falls back to the CPU if the device cannot be initialized. Added benchmark_pcps_grid (CPU baseline vs. GPU) and unit tests checking the GPU grid against the CPU reference and running the full GPS L1 C/A adapter on a real capture. Contributed by @phillipvu.

Improvements in Interoperability:

• Added the BeiDou B1C receiver chain, with signal identifier 1D: acquisition (BEIDOU_B1C_PCPS_Ambiguous_Acquisition, with optional QMBOC local replica), tracking (BEIDOU_B1C_DLL_PLL_VEML_Tracking, tracking the pilot component by default), and B-CNAV1 telemetry decoding (BEIDOU_B1C_Telemetry_Decoder, including LDPC decoding of subframes 2 and 3 and BCH decoding of subframe 1). The PVT engine uses the B-CNAV1 ephemeris, clock, and group-delay corrections (TGD_B1Cp / ISC_B1Cd), implements the BDGIM ionospheric model broadcast in B-CNAV1, and supports both B1C-only and mixed B1I+B1C configurations, keeping DNAV and B-CNAV1 ephemerides isolated and preferring B1C over B1I when both signals are available from the same satellite. B-CNAV1 ephemerides are also written to RINEX navigation files (native CNV1 records in RINEX 4.02, D1-style stand-in records in RINEX 3.02) and to the XML assistance-data storage. A sample configuration file is provided at conf/File_input/Beidou/gnss-sdr_BDS_B1C_geb_if20k_fs18m_ibyte.conf. Contributed by @OuWenhao16.
• Added the BeiDou B2a RNSS receiver chain (B2a_I data / B-CNAV2), with signal identifier 5D: PCPS acquisition (BEIDOU_B2A_PCPS_Acquisition), DLL+PLL tracking (BEIDOU_B2A_DLL_PLL_Tracking; BPSK(10), 1 ms primary code, data component only), and B-CNAV2 telemetry decoding (BEIDOU_B2A_Telemetry_Decoder), including soft-decision 64-ary LDPC(96,48) decoding of the 576 coded bits into 288 information bits before CRC-24Q and PRN validation. The decoder reuses the B1C GF(64) arithmetic and fixed-path decoder, with a full-alphabet sum-product fallback for B2a. Carrier polarity and tracking gain are normalized before decoding. The PVT engine uses B-CNAV2 ephemeris, clock, and group-delay corrections (TGD_B2ap / ISC_B2ad), and RINEX 4.02 navigation files contain native CNV2 records. GEO and BDS-2 satellites (PRN 1-18 and 59-63) are not assigned B2a channels and are not used in PVT. Sample configuration files are provided at conf/File_input/Beidou/gnss-sdr_BDS_B2a_file.conf and conf/File_input/Beidou/gnss-sdr_BDS_B2a_cu_l5_if20k_fs18m.conf. Contributed by @huangchuhan.
• Added support for the QZSS L1 C/B signal (PRNs 203-206), broadcast by satellites configured to transmit it in place of L1 C/A. Observables and ephemerides from L1 C/B PRNs are attributed to the PRN of the satellite’s nominal PNT signals in PVT and output products, following the RINEX 4.00 convention. Contributed by @vladisslav2011.
• Added reception of SBAS L1 signals (EGNOS and WAAS, PRN 120-138), with signal identifier S1: PCPS acquisition (SBAS_L1_PCPS_Acquisition), DLL+PLL tracking (SBAS_L1_DLL_PLL_Tracking), and telemetry decoding (SBAS_L1_Telemetry_Decoder) with Viterbi FEC decoding, CRC-24Q verification, and message-type reporting. Decoded frames carry traceback-corrected reception timestamps and can be dumped to per-PRN text files in an EMS-like layout with TelemetryDecoder_S1.dump=true. SBAS satellites are not used as ranging sources yet. A sample configuration file is provided at conf/File_input/SBAS/gnss-sdr_SBAS_EGNOS_rx.conf. Contributed by @kalmancito.
• Added support for RINEX 4.02 output, activated by setting PVT.rinex_version=4 in the configuration file (or with the -RINEX_version=4.02 command-line flag). Observation files are generated in the 4.02 version format, and navigation files make use of the data record structure introduced in RINEX 4.00. The default behavior when PVT.rinex_version is not set remains unchanged (RINEX 3.02).
• Added an opt-in RTK path fed by NTRIP corrections. The PVT block can now connect to an NTRIP caster, decode the RTCM 3 base position and base observations, and feed time-aligned reference data to its RTKLIB relative-positioning solver. Supported receiver configurations, per constellation and freely combined: GPS L1 C/A alone or together with L2C or L5, Galileo E1 alone or together with E5a, and BeiDou B1C (single-frequency). Single-band sets run single-frequency RTK, viable on the short effective baselines of VRS services. GPS L5 and Galileo E5a share the same center frequency, and BeiDou B1C shares the GPS L1 / Galileo E1 center, so the combined GPS L1+L5 / Galileo E1+E5a / BeiDou B1C dual-frequency receiver needs only two RF channels, and a single-frequency GPS+Galileo+BeiDou receiver needs one. The RTCM 3 MSM decoder gained the BeiDou B1C signal identifiers and prefers B1C over B1I when a base station broadcasts both in the shared first frequency slot. The client prefers NTRIP v2 and, after a fully-sent v2 exchange closes or times out before receiving response bytes, or returns HTTP 400, 501, or 505, retries on a fresh NTRIP v1 connection (PVT.ntrip_version=1 forces the legacy protocol). It supports TLS 1.2 or newer with system-CA certificate and hostname verification (PVT.ntrip_tls_enabled=true). It reconnects without blocking the GNU Radio work function, filters station changes and stale corrections, redacts credentials from RTKLIB traces, and retains an explicitly labeled single-point fallback when configured. VRS and nearest-station mountpoints are supported: the client periodically reports the rover position upstream as an NMEA GGA sentence (PVT.ntrip_send_gga, enabled by default, with the cadence set by PVT.ntrip_gga_period_ms, 10 s by default), starting as soon as the receiver produces its first position solution.
• The ionospheric Klobuchar coefficients and the UTC(NICT) offset parameters broadcast by QZSS satellites, both in the L1 C/A LNAV message and in the L5 CNAV message, are now stored separately from the GPS ones instead of overwriting them. This enables the QZUT / QZSS ION RINEX 4 data records (with the compulsory WIDE subtype for the CNVX Klobuchar set, broadcast in CNAV Message Type 30), prevents QZSS-sourced parameters from being mislabeled as GPS corrections in mixed GPS + QZSS configurations, feeds the QZSS slots of the RTKLIB navigation structure, and adds qzss_utc_model.xml, qzss_iono.xml, qzss_cnav_utc_model.xml, and qzss_cnav_iono.xml to the XML storage output.
• QZSS ambiguities are now resolved in their own group instead of jointly with GPS, avoiding integer fixes across the GPS-QZSS inter-system bias, and the RTCM 3 decoder accepts the final RTCM 3.3 BeiDou ephemeris message type 1042 (in addition to the draft type 63), with the a2 clock drift rate term now scaled per the BeiDou ICD (2^-66 instead of 2^-55).
• Cycle-slip detection by phase-doppler difference is now available: the detector removes the common receiver clock error as the median range-rate residual over all satellites before thresholding. It is enabled by setting PVT.slip_threshold_doppler (in m/s; 0, the default, disables it). The innovation rejection threshold in relative positioning and PPP is now split between carrier-phase and code observables: PVT.threshold_reject_innovation_phase complements the existing PVT.threshold_reject_innovation (which now applies to code) and defaults to the same value, so existing configurations behave identically; a value of 5.0 m for the phase threshold is recommended in RTK modes.
• The double-difference ambiguity transformation now uses the index-based formulation, and integer ambiguity resolution is driven by a new management layer that can skip AR while the float position variance is still high (PVT.ar_max_position_variance, default 0.25 m^2; 0 disables the gate), reject newly-risen satellites and retry when the AR ratio degrades (PVT.ar_filter, default true), cycle a single satellite out of AR when no fix is achieved with many satellites in view (PVT.min_drop_sats, default 10; 0 disables), scale the AR ratio threshold with the number of ambiguity pairs (PVT.ar_ratio_min/PVT.ar_ratio_max; equal values keep the fixed PVT.min_ratio_to_fix_ambiguity), and gate fixing and holding on minimum satellite counts (PVT.min_fix_sats, default 4; PVT.min_hold_sats, default 5) with a configurable fix-and-hold pseudo-measurement variance (PVT.var_holdamb, default 0.1 cycle^2). The reference satellite for double differencing is now selected by lowest measurement variance instead of highest elevation, excluding slipped satellites, which behaves better in urban conditions where SNR is a better quality proxy than elevation. Also fixed an out-of-bounds risk in the double-difference bias bookkeeping when five constellation groups are active.
• Added SNR-dependent and receiver-reported-stdev terms to the observation weighting model of the single-point and RTK solvers: PVT.error_factor_snr (m; a recomended value is 0.005) adds a term driven by the C/N0 of rover and base observations relative to PVT.error_snr_max (default 52 dB-Hz), and PVT.error_factor_rcv_std weights observations by receiver-reported pseudorange/carrier-phase standard deviations (new Pstd/Lstd fields in the observation structure, ready to be populated from the tracking-loop variance estimates). Both terms default to 0.0 (disabled), preserving the elevation-only error model.
• The single-point solver can now estimate a separate QZS-GPS inter-system bias instead of assuming QZSS shares the GPS receiver clock. The estimated offset is reported in sol.dtr[4]. It is opt-in via PVT.estimate_qzss_isb=true (default false) because the extra unknown requires one more satellite in mixed GPS+QZSS epochs, which degrades availability under limited sky visibility; enable it only in open-sky scenarios with six or more satellites in view.
• Added a Bladerf_Signal_Source for interoperability with Nuand’s bladeRF front-ends (bladeRF x40, x115, and bladeRF 2.0 Micro xA4/xA9), streaming RX samples directly through libbladeRF (requires the -DENABLE_BLADERF=ON building flag) instead of going through gr-osmosdr. Supports single-channel (SISO) reception, an optional RX bias tee for powering an active antenna on the 2.0 Micro, and exposes a single overall RX gain (unlike the if_gain / rf_gain split used by Osmosdr_Signal_Source). A sample configuration file is provided at conf/RealTime_input/gnss-sdr_GPS_L1_bladeRF_native.conf. Contributed by @MrCry0.
• Added a new Signal Source implementation Pocket_SDR_Signal_Source, which supports Pocket SDR FE 2CH/4CH/8CH GNSS RF front-ends through the gr-pocketsdr GNU Radio out-of-tree module. It requires the -DENABLE_POCKETSDR=ON building flag. Check the Signal Source documentation. Contributed by @minhaj6.
• Improved support for Keysight (formerly Spirent) GSS6450/GSS6425 format sample files. The Signal Source implementation is now named GSS6450_File_Signal_Source, while retaining Spir_GSS6450_File_Signal_Source as a backward-compatible alias. It can auto-detect .gns file layout information, unpack 2-, 4-, 8-, and 16-bit samples, and expose multi-channel recordings as independent RF output streams.
• Reworked ION_GSMS_Signal_Source support for ION GNSS SDR metadata files. The source now validates and deduplicates requested streams, honors file offsets, block headers/footers, and omitted cycle counts, and stops finite captures cleanly with a guarded valve tail. Chunk unpacking now handles word endianness, padding, shifts, repeated lump patterns, repeated stream IDs, standard integer encodings, and FP32 streams as float or gr_complex outputs.
• Improved Labsat_Signal_Source support for LabSat 2, LabSat 3, and LabSat 3 Wideband recordings, including more robust header parsing, corrected 2-bit sample decoding, multi-channel output handling, and unit-test coverage for the supported layouts.
• Added an opt-in EVK1029_Signal_Source for the SAPHYRION EVK1029, a dual-band (E1/E5a) or triple-band (E1/E5a/E6) GNSS evaluation kit built around the SY1009 RF front-end and SY1019 ADC/DSP space-grade ASICs. Reads the EVK1029 host application’s raw capture files directly (a continuous, header-less stream of OBA-encoded 4-bit samples, two per byte, 16 samples per little-endian 64-bit word), without going through the generic XML-metadata-driven ION_GSMS_Signal_Source path. Disabled by default; build with -DENABLE_EVK1029=ON to enable it.
• Improved Galileo HAS robustness and ICD compliance, including stricter MT1 validation, correct cache/Do-Not-Use handling, TOW fallback for E6 HAS pages, preserved mask/IOD correction context, and corrected HAS application in RTKLIB/PVT.
• Fixed bugs in the generation of RTCM MSM messages.
• Fixed identification of GLONASS satellites.
• Fixed bug in the generation of the spreading code for QZSS L5 PRN 196.
• Improved validation of GPS/QZSS CNAV Clock, Ephemeris, Integrity (CEI) dataset.
• Implemented QZSS LNAV almanac/auxiliary pages decoding.
• Hardened BeiDou DNAV and Glonass GNAV decoding.
• Completed BeiDou D1/D2 DNAV decoding, including almanac, time, integrity, differential-correction, and ionospheric-grid data, with BeiDou almanacs wired into RTKLIB-assisted satellite visibility.
• Corrected RINEX 3/4 navigation and observation output for GPS, QZSS, Galileo, and BeiDou, including DNAV metadata and refreshable RINEX 4 ION/STO/EOP records. CNAV-only GPS/QZSS configurations now automatically use RINEX 4.02 for both files, avoiding lossy RINEX 3 navigation records.
• Improved performance of Galileo’s Viterbi decoder.
• Fixed edge cases in the retrieving of GPS L1 C/A navigation data.
• Fixed Glonass carrier phase and time annotations in RINEX files.
• Implemented handling of the GLONASS notification of a forthcoming leap second event (KP word in the GNAV message), improving timekeeping across leap second transitions.
• The NMEA printer now generates QZGSA and QZGSV sentences, reporting the QZSS satellites used in the PVT solution and in view (with elevation, azimuth, and C/N0), using the QZSS system and signal identifiers defined in NMEA 0183. Contributed by @vladisslav2011
• Fixed NMEA GSV C/N0 reporting for non-L1 configurations, including GPS L2 and L5: RTKLIB now preserves per-frequency signal strength and observation-code metadata in satellite status, and the NMEA printer emits the strongest available C/N0 with the corresponding NMEA signal identifier. Contributed by @vladisslav2011.
• The custom output stream defined by monitor_pvt.proto now includes a tracked_satellites list. Each entry reports one tracked signal (system, prn, signal), its azimuth_deg and elevation_deg, whether it was combined with another signal of the same satellite (e.g., the Galileo E1+E5a ionosphere-free combination), and a used flag telling whether it contributed to the reported fix. Satellites that were tracked but left out of the solution (below PVT.elevation_mask, or excluded by RAIM) are listed with used = false. Unhealthy satellites are listed with healthy = false. Contributed by @joebre.

Improvements in Maintainability:

• Refactored main Acquisition, Tracking, and Telemetry Decoder adapters, simplifying interfaces and improving consistency across processing chains. This reduces code duplication, enhances maintainability, and eases the integration of new GNSS signals. Contributed by @MathieuFavreau.
• Merged the GLONASS L1 and L2 C/A telemetry decoder blocks, as well as the BeiDou B1I and B3I ones, which were almost identical since each pair of signals broadcasts the same navigation message (GNAV and DNAV, respectively), into single blocks parameterized by the frequency band, following the approach already used for the Galileo telemetry decoder. No changes are required in configuration files.

Improvements in Portability:

• Refactored Python interpreter detection and improved CMake portability and robustness across dependency discovery, distro detection, and cross-compilation handling.
• The CUDA build (-DENABLE_CUDA=ON) works again with current toolkits and on NVIDIA Jetson: removed the hardcoded sm_30 (Kepler) architecture, which CUDA >= 11 rejects; CMAKE_CUDA_ARCHITECTURES is now honored and detected automatically on Jetson (Orin -> 87, Xavier -> 72, TX2 -> 62, Nano -> 53) or set to native with CMake >= 3.24; the CUDA language standard follows the host C++ standard (C++17); imported CUDA::cudart/CUDA::cufft targets are linked explicitly; -Wno-psabi is no longer passed to nvcc. Contributed by @phillipvu.
• Added docs/JETSON.md, a build/verify/benchmark guide for NVIDIA Jetson. Contributed by @phillipvu.

Improvements in Reliability:

• Hardened the Galileo OSNMA protocol implementation, adding support for Chain Renewal, Chain Revocation, Public Key Renewal, Public Key Revocation, Merkle Tree Renewal, and OSNMA Alert Message events. Improved the management of OSNMA cryptographic material and added unit tests to ensure compliance with the OSNMA Receiver Guidelines v1.3, including edge-case handling. Added the new configuration value GNSS-SDR.osnma_mode=replay, which disables the receiver wall-clock GST alignment check for OSNMA tag processing, enabling replay of previously captured Galileo signals while keeping all other OSNMA verification steps active.
• Fixed the decimation logic of the Monitor, AcquisitionMonitor and TrackingMonitor blocks: decimation_factor now selects every N-th epoch and always consumes all the input items, instead of grouping Gnss_Synchro objects into bursts and skipping others, and empty datagrams are no longer sent. Fixed a use-after-free memory corruption caused by google::protobuf::ShutdownProtobufLibrary() being called from the destructor of Serdes_Gnss_Synchro, before the protobuf library was actually used; the library is now shut down only once, at program exit. Added a unit test for the monitor decimation. Contributed by @vladisslav2011.
• GPS_L1_CA_DLL_PLL_Tracking_GPU: fixed a cross-block data race in the CUDA multi-correlator kernel (the carrier wipe-off and the correlation were in the same launch, synchronized only with __syncthreads()), fixed the cudaHostAlloc flags (cudaHostAllocMapped || cudaHostAllocWriteCombined evaluated to cudaHostAllocPortable), stopped calling cudaDeviceReset() from a per-channel destructor (it tore down the context under the other channels), and stopped cudaFree()-ing device aliases of host-mapped buffers.

Improvements in Usability:

• The PVT Monitor now reports per-signal details for satellites used in the position solution, including PRN, constellation, signal, azimuth, elevation, and whether multiple signals were combined. Contributed by @joebre.
• The Monitor (Monitor.enable_monitor=true) now also reports channels that are tracking a signal but do not have a valid time reference yet, filling their entries with the latest raw tracking data (C/N0, Doppler, carrier phase) while keeping their observable validity flags unset. This makes the Monitor usable in Galileo E6-only configurations, where the time of week cannot be obtained from HAS pages, as well as during the initial seconds of operation, before the telemetry decoders attain synchronization.
• Added Galileo System Time (GST) annotations to HAS outputs when GST is decoded from an I/NAV channel, enabling the HAS Time of Hour (TOH) to be associated with an absolute UTC timestamp.
• Galileo E6 observables are now generated by default, making them available in RINEX files and other receiver outputs when E6 channels are configured. Since Galileo E6 HAS pages do not broadcast the time of week, the receiver configuration must also include other Galileo channels providing the time reference for the E6 observables, either E1 or E5b (I/NAV), or E5a (F/NAV). Their generation can be disabled by setting Observables.enable_E6=false. This setting is now independent of PVT.use_e6_for_pvt, which keeps controlling whether E6 observables are used in the PVT solution.
• The console now reports the RTKLIB solution status. The First position fix and periodic Position at lines are tagged with [RTK FIXED], [RTK FLOAT], [DGNSS], [SBAS] or [PPP] (color-coded), and status transitions are announced once when they happen, including the LAMBDA ambiguity-resolution ratio and its threshold when an RTK fix is acquired or lost. The [PPP] label is only shown when precise ephemeris and clock products are actually loaded; PPP-mode processing on broadcast products is not labeled as PPP. Plain single-point operation keeps the classic, unmodified console output. The underlying outputs (PVT dump, Monitor, NMEA) keep reporting the raw RTKLIB solution status.
• A new global parameter GNSS-SDR.observation_date allows specifying the approximate date of the signal capture, in YYYY-MM-DD or YYYY format (e.g., GNSS-SDR.observation_date=2014-12-20), when post-processing recorded signal files. It is used to resolve the GPS mod-1024 week-number rollover: each broadcast week number is expanded to the 1024-week era closest to the given date. This works for recordings from any era, including files captured after the April 2019 rollover replayed far in the future, and also fixes the applied leap-second offset, which is derived from the resolved date. If the parameter is not set, the era is derived from the system clock, as before, which is the right choice for live operation. The GNSS-SDR.pre_2009_file flag, which could only select the August 1999 - April 2019 era, is now deprecated: it keeps working exactly as before, but the receiver prints a notice suggesting GNSS-SDR.observation_date instead, and it is ignored if the new parameter is also set.
• Reworked the Python plotting utilities under utils/python (acquisition, tracking, telemetry, observables, and PVT diagnostics). Each script now exposes a command-line interface (run with --help) and can be executed from any directory with configurable input and output locations, instead of requiring edits to the source to change file paths. The --file-prefix option takes the value of the corresponding block’s dump_filename configuration parameter directly, reconstructing the dump file names the same way the receiver does. A new utils/python/README.md documents all the utilities and their options. Includes fixes to the acquisition grid and tracking dump readers and plotters contributed by @minhaj6.
• Fixed the time tags of position solutions reported in the terminal and in NMEA, KML, GPX, and GeoJSON outputs for configurations without GPS channels (e.g., Galileo-only receivers): the reported epoch was shifted by the residual receiver clock offset, which in the absence of GPS satellites is absorbed by an inter-system bias state instead of the receiver clock state. Reported epochs now fall on the same integer-millisecond grid as the observables, as they already did in configurations including GPS. RINEX files were not affected.
• Abseil logging now creates a unique timestamp/PID logfile for each run, preserving previous logs across the receiver, calibration tool, and test runners. On POSIX systems, an atomically updated relative symlink points to the latest logfile.

As usual, compressed tarballs are available from GitHub and Sourceforge.