Changelog
What's new on gnsscalc.com, newest first. Library-level changes to the underlying gnss-js engine have their own changelog.
Changelog
User-visible changes to gnsscalc.com, newest first. The site deploys continuously, so entries are grouped by date rather than version. Library-level changes live in gnss-js’s own CHANGELOG.
2026-11-16
- /campaign: tidier time-series charts and station table on phones. The position and residual charts’ axes end on round values (whole millimetres, half years; the first of the month on series under a year) instead of steps like 0.9 mm and an end label of 2025.9. On phones the station table fits without scrolling: each ± sits under its value and the role toggle reads Stn / Ref.
2026-11-15
-
/campaign: clearer setup and results. The default span is the last two years (monthly days ending at least three weeks ago, so every day can be ambiguity-fixed) instead of a fixed 2024–2025 window; network alignment is offered only with three or more stations, where it applies. Each day now records the products that solved it, and the result and the report name them (“CODE final orbit + 30 s clock + code/phase biases”) instead of a list of possibilities; the report also says on how many days the ambiguities were fixed. The 2020.0 coordinate epoch says what it is for when it lies outside the data. With fewer than five days per station the report card no longer promises velocities and time series. The local-network intro states what the baselines are validated to (2–5 mm up to 50 km) instead of a 10–15 km limit from the older engine. On phones the coordinate table no longer breaks its numbers, and every report card puts its download button under the title instead of squeezing it.
-
/positioning (RTK): a survey report for static baselines. With a known base coordinate, a static session’s result offers the same certificate as PPP’s: the rover’s coordinates in ITRF2020, the regional frame and the national grid and heights, derived from the frame the base coordinate is in (now chosen beside it — ETRS89, ITRF2020 or NAD83, by default the one used where the base is); the base and the baseline (ΔE/ΔN/ΔU, length); both antennas and receivers; the 95 % uncertainty relative to the base; a sign-off block. Annex A gives the processing (orbits, constellations, ambiguities, the PPP check, both files’ SHA-256) and an independent check against the epoch-by-epoch solution with broadcast orbits, plotted over the session; Annex B the rover file’s data quality, Annex C site photos. It can be registered with a report number like the PPP report. The static baseline’s 95 % uncertainty is now at least 6 mm horizontally and 10 mm vertically, what its validation on EPN baselines supports (the formal figure said 4–5 mm).
-
/positioning (RTK): a known base coordinate, and static sessions solved as a whole. The base’s coordinate can now be entered (any common format) instead of taken from the base file’s header, whose approximate position the rover used to inherit — centimetres to decimetres off. A static session is also solved as one baseline over the whole session from precise orbits (the method /campaign’s network mode uses), checked against both stations’ PPP; the result leads with the rover’s coordinate, its 95 % uncertainty and how it was solved, and the epoch-by-epoch solution (fix rate, track) stays below for the map and plots. Wettzell, 68 m baseline, 24 h: the rover within 1.5 mm horizontally and 3 mm in height of EUREF’s coordinate. Data from the last day (no precise orbits yet) still gets the epoch-by-epoch solution, and the page says so.
-
/positioning (PPP): clearer results, and a sample to try. The result states one coordinate and one uncertainty — the report’s 95 % horizontal and vertical figures — instead of several different accuracy numbers and a second, unlabelled coordinate further down; the scatter card now says it shows how steady the solution was, not its accuracy. The RINEX header’s approximate position is compared without a warning (it is an a-priori value, not a surveyed coordinate). The RTK-base tools have their own “Use as RTK base” tab, the setup guide speaks to surveying a point, constellations are named and the products shown as final or rapid. “Try a sample” processes a real 24 h file from Wettzell and compares the result with EUREF’s combined solution for that day (a few millimetres). The FAQ states the validated accuracy.
-
Keyboard shortcuts. ⌘K (Ctrl+K) anywhere opens a palette: type a tool’s name or what you want to do (rtk, ntrip, antenna, geoid…) to go there, or type a time to open it in the time calculator. On the time page, / jumps to the paste box. The palette lists all shortcuts when empty.
-
Time calculator: step the time. Under each card’s RINEX epoch, ‹ › move the time back or forward by a second, minute, hour, day or week (pick the unit between them); hold to repeat, or use ↑/↓ in the RINEX field.
-
Survey report: the station’s own coordinate is labelled for what it is. When the comparison is against the RINEX header’s APPROX POSITION XYZ, the report now says that is an approximate position, not a survey reference — a few centimetres there are not an error of the result.
-
/ppp and /campaign: the troposphere may change faster during a session. The wet delay’s allowed variation is now set per hour (3.5 mm/√h) instead of per observation epoch, so 1-second and 30-second files of the same site get the same model, and it is looser than before (1.1 mm/√h at 30 s), which was too tight for a summer or tropical hour. One-hour sessions in Europe end closer to the truth (East 10 → 7 mm RMS) and fix more often; 24 h coordinates are unchanged, and /campaign’s ZTD agrees better with EUREF’s (4.5 → 4.1 mm RMS).
-
/ppp and /campaign: ambiguity-fixed runs use GLONASS again. The precise products publish no GLONASS phase biases, and fixed runs used to drop GLONASS altogether; it now stays in the solution with float ambiguities (only GPS, Galileo and BeiDou are fixed). One-hour sessions fix about 2 minutes sooner and end closer to the truth; 24 h /campaign days improve slightly (1.4/0.8/2.1 → 1.3/0.7/2.0 mm on the EPN benchmark).
-
Time calculator: the paste box reads far more, and shows what it read. Dates as people write them — 17/11/1994, 11/17/1994, 17.11.94, 17 Nov 1994, Nov 17th 1994 2:30 pm, 19941117T123000, 2026-11, March 3, 2026-W46-7 — plus a time alone (today), now/today/yesterday, doy 319, GPS week alone, decimal years, µs and ns timestamps, and TAI or TT times. While you type, a list shows every reading with its GPS and UTC time; pick one with the arrows or the mouse, Enter sets the Initial Time and Shift+Enter the Final Time, and Undo puts the time back. Where day and month could swap (05/11/1994) both readings are listed. Fixed: a date without a time (1994-11-17) was not read, and long numbers could give dates thousands of years away.
-
/visibility: the World view is open by default on the Sky tab (satellites in view and PDOP everywhere at the plan time). Its map is built when you scroll to it; close it and it stays closed.
-
/visibility: a reassigned PRN no longer shows the wrong satellite. Past the broadcast orbits, satellites are filled from CelesTrak element sets matched to their PRN; a satellite whose CelesTrak name carried a PRN could take it from the one the broadcast actually identifies by position. C12 is now BeiDou-3 M26 (named C48 at CelesTrak), but plans showed the retired BeiDou-2 M4, still named C12, about 60° away in the sky, with a jump where the two sources met. Position matches now always win over names, and a name no longer stands in for a PRN the day’s broadcast carries but no element set matches: C06, C08 and C40 were filled from the BeiDou-2/-3 IGSOs still named so, which are other satellites today, and drew figure-eights with a kink where the broadcast ended. They now end with the broadcast. The receiver profiles get the right satellite generation too.
-
Pick a satellite once, see it everywhere. Click a satellite on a sky plot, the world map, the pass schedule, the passes table, the elevation chart’s satellite list (/visibility) or the elevation heatmap (/qc), and every view on the page picks it: its whole path is drawn bold on the sky and on the map, the rest fades, and a card gives where it is now (azimuth and elevation, or the point it is over) and, on /visibility, the pass it is on or its next one. On the world map the picked satellite’s whole orbit is drawn too, dashed: a sidereal day of ground track, the closed figure-eight of a BeiDou IGSO, both revolutions of a GPS satellite; and it stays on the map (a hollow ring) while it is below your horizon. The /visibility world map now shows every satellite at the plan time, those below your horizon small and faint, with their tracks at half strength (the track mode, Trail by default, keeps the amount in hand). Click it again, or the ✕, to let go.
-
One view panel on every sky plot and ground-track map (/visibility, /qc, /stream, /nmea, /orbits). Beside a wide sky plot it stays open (on narrow ones and on the map it is behind a “View” button), and its constellation list is also the plot’s legend, with the number in view; the map’s legend hides and shows constellations the same way. It sets which constellations are drawn, the tracks on the sky and on the map (off, a trail of 10 min to 2 h, up to the time cursor, or the whole window), satellite names, satellites the receiver didn’t track (grey or hidden), the elevation-mask ring, and the night side on the world map. The choices are kept in the browser and apply on every page until Reset; each page keeps its own defaults until then. It replaces the constellation checkboxes on /qc and the “Full tracks” box on /visibility, and /qc’s world map now also follows the time cursor.
-
/visibility: plans for tomorrow and later come in about 6 s instead of 15–20 s. Past the broadcast orbits the planner waited for the almanac feeds, two of which take ~15 s to fail; they now get 6 s, after which the CelesTrak element sets (already loaded) carry the forecast, and the sources line says so.
-
/visibility shows daylight. Night is shaded on the plan’s timeline, with sunrise and sunset at the site beside it (or polar night / midnight sun); the plan report lists them in its summary.
-
Time calculator: paste any time. One box above the calculator reads what you have at hand: a RINEX or SP3 epoch line, a RINEX 2/3/4 or IGS product file name (or its URL), NMEA ZDA/RMC, GPS week + time of week, year + day of year, a SINEX epoch, Unix or GPS seconds, JD/MJD, or a date in almost any form. Pasting sets the Initial Time; the line under the box says how the text was read, and offers the Final Time and any other reading (1478779200 is GPS seconds in 2026 or Unix time in 2016).
-
Time calculator: products for this day. Under the calculator, the files people download for the Initial Time’s day — broadcast navigation; IGS final, rapid and ultra-rapid orbits, clocks and ERP; CODE and ESA multi-GNSS finals with CODE’s biases; GFZ rapid; ESA and CODE ionosphere maps — with their archive names (copy or download), their folders, and whether each is on the server yet, read from the archives’ listings. Older days get the names used then (short IGS names at CDDIS, WRD broadcast files). Plus a link to the archive explorer for station data that day.
-
/ppp uses BeiDou by default (GPS + GLONASS + Galileo + BeiDou; /campaign unchanged). BeiDou is read on B1C/B2a, the signals today’s precise clocks refer to, with or without ambiguity fixing (B1I/B3I stays under Advanced for receivers that track nothing newer), and is left out where the receiver antenna has no BeiDou calibration. One-hour sessions: with ambiguity fixing the first fix comes 1–4 minutes sooner and more sessions fix; float solutions are a little closer to the truth (Asia-Pacific: 67 → 57 mm after 15 minutes). On 24 h sessions it makes no difference, which is why /campaign keeps its constellations.
-
/ppp: the default engine also waits for the float solution to settle before fixing (5 cm in 5 minutes): wrong fixes on one-hour sessions about halve, the first fix comes 2–4 minutes later; final and 24 h results unchanged.
-
/visibility warns about space weather. Under the plan’s headline, a line gives the NOAA geomagnetic forecast (Kp) over the window and a rough scintillation risk at the site, from its magnetic latitude (IGRF-14), local time, season and solar flux: after sunset near the magnetic equator, in the auroral zone at night and in storms, and at mid latitudes in severe storms. Risk periods are shaded on the timeline and flagged beside the verdict; the plan report (PDF) has it in the summary and a section of its own with the risk periods and the Kp per 3-hour slot.
-
/ppp and /campaign, uncombined engine: fixes wait for the float solution to settle. No fix is tried while the float position still moves more than 3 cm in 3 minutes. Near the equator a short session’s float can drift by decimetres while its error estimate says a few centimetres, and fixes taken then were wrong. One-hour sessions: wrong fixes about 3–4× rarer (Europe and Asia-Pacific), the first fix about 2 minutes later; final results unchanged, 24 h results identical.
-
/ppp and /campaign fix BeiDou ambiguities when you add BeiDou (Advanced → Constellations). Before, ambiguity-fixed runs silently dropped BeiDou: CODE’s standard final products have none. With BeiDou chosen they now come from CODE’s MGEX final (BeiDou on B1C/B2a), BeiDou-3 satellites built by CAST are fixed and those built by SECM stay float (fixing them caused wrong fixes), and each BeiDou family gets its own receiver clock. On one-hour sessions this fixes 1–4 min sooner and more often than GPS + GLONASS + Galileo, in Europe and Asia-Pacific; the default constellations are unchanged. BeiDou is also left out where the receiver antenna has no BeiDou calibration in IGS20 (the result says so): there the GPS offsets would stand in, which moved those stations by 1–2 mm in height on 24 h sessions.
-
/ppp and /campaign, uncombined engine: fixes wait for a 12 cm float. Near the equator one-hour float solutions stay decimetres off and the fixing could accept wrong integers; it now waits until the float is better than 12 cm (Asia-Pacific: about a fifth fewer sessions with wrong fixes; Europe unchanged).
-
/visibility: a better Where. A large map with the place search on it (it also takes a pasted position) and a my-location button; beside it the site’s name (clicking the map names the place), the position in any format — decimal, DMS, UTM, MGRS, locator or ECEF — with DMS and UTM readouts, and the height, filled in from the terrain model and the EGM2008 geoid whenever the site moves.
-
/visibility, reorganised around the answer. After Compute, a headline says whether the window is good and, if not, when it is weak, with the satellites in view, PDOP, the share of time with good geometry and the best hour, over one timeline of satellites and PDOP (weak periods shaded; click it to move the time cursor). The rest sits in tabs — Sky, Timeline, Passes, RTK & base, Atmosphere, Satellites — instead of one long page. The setup is one bar — Where, When, Mask, Receiver, Compute — that reads as the question being asked; each part opens its editor (place search, map and “use my location”; quick windows such as next 4 h, next 24 h or tomorrow 07–19 at the site; the mask and horizon; the receiver as cards). “When should I go?” is a mode of its own. The planner now opens on the current time, and the passes table says “up at start” / “up at end” for passes the window cuts, with each pass’s time in view.
-
/visibility computes in about 2 s instead of 15–20 s for plans that start now. Where the broadcast orbits run out, satellites are now filled from CelesTrak element sets (already loaded, and as good as almanacs for planning); the almanac feeds, which could hang for 15 s, are asked only when the element sets can’t be loaded, and get 6 s.
-
/visibility: the plan report is a card at the top of the results, like the reports on /qc and /ppp, with what the PDF holds, its build steps, and an optional site name for the title.
-
/visibility: can I get RTK here? The streamed reference stations nearest your site — public casters, IGS and EUREF stations, and on request the RTK2go and Centipède community bases — with distance and direction, the constellations each sends (flagged when it lacks one your plan uses), and what the distance means for RTK: a verdict from ideal to too far, and the expected precision at 8 mm + 1 ppm. Says whether you are inside a permanent network (network RTK) or at its edge, lists the stations with daily RINEX for post-processing with a rule-of-thumb occupation time, and “Common view” runs the base + rover comparison with that base.
-
/orbits: a group over your site for a day. Under the sky plot, how many of the group’s satellites are above your mask every 15 minutes for the next 24 h — for Starlink at 52° N, 147–234. Crowded skies drop the per-satellite labels (hover for names).
-
/orbits: constellation slots. For GPS, GLONASS, Galileo and BeiDou, every MEO satellite by its orbital plane and its place in that plane, now: the planes as columns, empty slots as gaps, Galileo’s two satellites in eccentric orbits between them. Click one to select it.
-
/visibility: the scanned days at a glance. Under the best windows, a day × time-of-day grid of mean PDOP (or the fewest satellites in view) every half hour; click any cell to plan a session from there.
-
/visibility: shared links keep the local horizon. Share a plan and the recipient gets your horizon (computed or drawn) too, not just the mask.
-
/orbits: pass detail. Click a pass for where to point (azimuth at rise, highest point and set), the closest range, and the Doppler shift over the pass at a frequency you enter (L1 for GNSS satellites, 437 MHz otherwise), with its largest rate of change.
-
/visibility: world view. Satellites in view, or the PDOP they give, across the whole world at the plan’s time cursor, with this plan’s constellations, receiver and satellite selection. Hover anywhere for the value there.
-
/visibility: plan report (PDF). The plan as a printable report in the house style: site, window, mask and horizon, receiver and orbit sources; the numbers that decide a session (satellites in view, PDOP, share of time with good geometry, the best hour, every period of weak geometry); satellites and DOP over time; the sky with the satellite tracks and your horizon; when each satellite is up, per constellation; GPS outage notices; and every pass in an annex. Built in your browser.
-
Fixed: “From buildings (OpenStreetMap)” on /visibility gave up too early. The main OpenStreetMap query server takes 30–45 s for a busy town centre; it now gets a minute before the next server is tried.
-
/visibility: GPS outage notices. Every NANU that touches the plan’s window — announced maintenance or manoeuvres, satellites unusable until further notice, and for past dates the outage as it happened — with a link to the notice. By default the satellite drops out of counts, DOP, sky plot and passes for exactly those minutes (a pass that crosses an outage is split). Notices come from CelesTrak’s NANU archive; Galileo’s NAGUs are linked, not read.
-
Fixed: some SBAS satellites were placed 42 million km away on /visibility, /status and /orbits about half the time. Part of the merged IGS broadcast files write SBAS positions in metres instead of km; they are now read correctly.
-
/orbits: SP3 from TLEs. Every PRN-matched GNSS satellite for the next 24 h at 5 min as an SP3-d file, for software that only reads SP3 (labelled in the file: about 1 km, no clocks).
-
/visibility: base + rover common view. Give a second site and see, over the plan’s window, the satellites each end sees, the ones both see (what a static or RTK baseline can use) and the PDOP of that common set.
-
/visibility: find the best time. Scans up to two weeks ahead for the sessions with the best geometry at your site (your mask, horizon, receiver profile and satellite selection), with your thresholds for satellite count and PDOP. Each window shows local solar time and an ionosphere note (afternoon peak; low-latitude evenings), and “Plan it” opens that window as a full plan.
-
Orbit Explorer has a 3D globe next to the map: a Sun-lit Earth with the day/night line, every satellite of the group (all ~11 000 Starlinks too), the selected one’s orbit over one period, and your position. Click or hover a satellite to pick it. Switch between Earth-fixed (as a receiver sees it: a GEO satellite sits still, an IGSO draws a figure-8) and inertial (fixed to the stars: orbits are closed ellipses and the Earth turns inside them).
-
/visibility plans for your receiver. Pick single-frequency L1, dual-frequency L1 + L2 (survey), L1 + civil L2C, L1 + L5 (mass-market dual-frequency) or triple-frequency, and counts, DOP, sky plot and passes keep only the satellites that transmit those signals. Each satellite’s block (GPS IIR / IIR-M / IIF / III, GLONASS-M / K, BeiDou-2 / 3, …) comes from the CelesTrak catalogue, matched to its PRN by position.
-
/visibility fills every constellation, any date. Where neither broadcast orbits, predicted orbits nor almanacs cover a satellite (often BeiDou, GLONASS, QZSS and NavIC in far-ahead forecasts), it is propagated from its CelesTrak element set.
-
Satellites in this plan: name, block, launch year, signals and element-set age of every satellite, each linked to its orbit on /orbits.
-
New on /visibility: automatic local horizon. Under “Local horizon / obstructions”, From terrain, From buildings (OpenStreetMap) or Both computes what blocks the sky at your site — mountains and valley walls out to 30 km from a global elevation model (with Earth curvature and refraction), and buildings within 300 m from their OpenStreetMap heights or floor counts — and turns it into the per-direction elevation mask. The computed skyline is drawn over the editor, a line says what blocks most and where (“up to 19° to the north”), and the mask can still be painted over by hand.
-
/orbits: how good is a TLE? For every GNSS satellite, its CelesTrak element set against the broadcast orbit of the PRN it was matched to, over the day so far: radial, along-track and cross-track error per constellation (about 1 km, mostly along-track), the worst offenders, error against element-set age, and the time series for the selected satellite.
-
/orbits: a year of Sun beta angle for the selected satellite, with its eclipse seasons shaded and dated.
-
/orbits exports the selected satellite’s next 24 h as a CCSDS OEM (orbit ephemeris message) and as CZML for CesiumJS, besides CSV and KML.
-
Orbit calculator on /orbits: from altitude, period or revolutions per day, the speed, J2 node and perigee drift, sun-synchronous inclination, coverage radius, longest pass and ground-track repeat, with presets (ISS, sun-synchronous, Starlink, GPS, Galileo, GEO).
2026-11-14
- New: Orbit Explorer. Where every GNSS satellite is right now, and any other satellite with a public element set (space stations, weather, Starlink, OneWeb, … from CelesTrak, or a TLE/OMM you paste): ground tracks, your sky, the orbit’s elements and type (LEO, MEO, IGSO, GEO, HEO), passes over you and Earth-shadow passages for the next 24 h, with CSV and Google Earth (KML) exports. Orbits are propagated in the browser with SGP4, checked against the 666 reference states of Vallado et al. (2006) to 0.1 mm.
- GNSS objects are matched to the PRN broadcast from the same position in the day’s IGS broadcast ephemeris (about 1 km apart). That gives the Galileo, GLONASS and NavIC PRNs that catalogue names don’t carry, and flags objects whose name shows a different PRN than the one they broadcast.
2026-11-12
-
Fixed: receiver antenna offsets were applied with North and East swapped in PPP (/positioning, /campaign) and in RTK between two different antenna types. The horizontal coordinates move by the size of the antenna’s horizontal offset: a few millimetres for most antennas, about 2 cm for a few (e.g. JAVRINGANT_G5T). Heights are unchanged. Checked against the EUREF combined solution on 40 EPN stations and against the ANTEX 2.0 definition of the antenna frame.
-
/antex reads ANTEX 2.0 (.atx2, the format IGS now publishes alongside igs20.atx) as well as 1.4, and saves a file as the other version (“Save as ANTEX 1.4 / 2.0”). Receiver offsets are always shown as North / East / Up — 2.0 writes them as X / Y / Z = East / North / Up. For 2.0 files: satellites are listed by SVN, and the offsets’ origin, code and gain patterns and the reference frame are shown; converting to 1.4 takes the satellites’ PRNs from the IGS satellite metadata.
-
Trajectory report for kinematic PPP. A kinematic run now has a Report tab: page 1 shows the track in plan, its start and end in ITRF2020 and the national grid, its length and speed, and the formal precision (labelled as a lower bound); Annex A is the processing (position and its σ over time, residuals, troposphere, receiver clock) and Annex B the file’s /qc report.
-
PPP processing options (Advanced settings): elevation mask, processing interval (a long 1 Hz file runs much faster), and start from a known position — on a surveyed mark the filter starts converged; kinematic runs hold the position for the minutes spent on the mark. A session split into several files (hourly RINEX, a split log) is processed as one. The PPP Export menu adds the track (or converged positions) as KML.
-
PPP exports: the static coordinate as SINEX (with σ from the validated 95 % uncertainties) and the receiver clock as RINEX CLOCK.
-
Fixed: the Export menu on /positioning didn’t download anything (PPP, RTK, Compare): picking an item closed the menu before the download started.
-
Survey reports get a number you can verify. A report from /positioning’s PPP can now be registered when it is made. The number (like GC-2026-K7Q4-9XHM) is printed in the letterhead and in every page’s footer, with the address to check it.
- At gnsscalc.com/verify/
, anyone holding the report sees the summary and coordinates it was issued with, and when. - Dropping the PDF there tells whether it is byte for byte the file issued. Dropping the observation file tells whether the report was computed from it. Files are checked in the browser, not uploaded.
- Registration is on by default and can be switched off. It stores the certificate’s summary and coordinates only, not notes, photos or data.
- At gnsscalc.com/verify/
-
The PPP survey report is a one-page certificate with annexes.
- Page 1: the coordinates in every frame, the antenna, the 95 % uncertainty and a sign-off block.
- Annex A, the processing: settings, satellites and PDOP over the session, observations used and rejected per satellite, residuals per satellite and against elevation, position over time with its formal σ, convergence and the accuracy basis, ambiguity fixing (with PPP-AR), zenith delay and the receiver clock (drift, Allan deviation).
- Annex B, the data quality: the file’s /qc report. Annex C: your site photos.
-
The survey report covers every national system on /coordinates. Germany, Belgium, Poland, Switzerland, Ireland (including Northern Ireland), Portugal, Spain and the US State Plane are now found from the point, not only chosen by hand. The certificate now gives LV95 with the LN02 and LHN95 heights in Switzerland, and the national heights from the official geoid grids: DHHN2016, RH 2000, N2000, NN2000, DVR90, Cascais, Alicante, and Malin Head or Belfast depending on the side of the border. The national frame is named as each country does (
ETRS89/DREF91/2016,CHTRF95), and Kehl, across the Rhine from Strasbourg, is now Germany on /coordinates too. -
The /qc PDF report has a new layout, the same as the survey report: the quality verdict first (graded as on the page, with your receiver class), then the file, tracking and signal strength, multipath (now also by direction, from the antenna), cycle slips and completeness. Charts are black and white.
-
Clearer QC drivers. When a file is missing most of the day’s epochs, Tracking now says so (“only 28 % of the epochs are in the file”) and no longer suggests interference or the antenna. When a single band loses lock while the others hold, Interference names that (“band 7 only: lost lock on most satellites”), since a jammer usually hits neighbouring bands too. Grades are unchanged.
-
Header position no longer blamed for jamming or spoofing. The check now ignores position fixes with no spare satellite or off the ground: a jammed hour on 3–6 satellites had pulled the result 6,000 km away (KRRS, 6 Nov; now 2 m). When the fixes themselves wander by kilometres, as under spoofing (POLV, 24 Sep: 1 km every 30 s), it says so rather than calling the header wrong.
-
National systems are now part of the Point and the transform.
- In a supported country, the Point card shows the national grids and heights alongside ECEF, geodetic and UTM. They’re editable like every other field: type RD, Lambert-93, LV95, a grid reference or a national height to move the Point.
- The frame transform targets the country’s national frame by default, with a one-line spec of the realisation (
Germany: ETRS89/DREF91/2016 = ETRF2000 @ 2008.46 · UTM, DHHN2016), and shows the same expressions with the result. - The separate national-grids card, with its country picker and input boxes, is gone.
-
/campaign reaches back to 1994. Days before 2014 are solved with the IGS’s third reprocessing (GFZ’s solution), which recomputed 1994–2022 orbits and clocks consistently with ITRF2020, the frame the results are in. Wettzell from 1999 to 2025 (313 monthly days, through three antenna changes taken from its site log) gives 20.41 / 15.64 / −0.10 mm/yr, within 0.2 mm/yr of its official IGS20 velocity. Where a series spans the switch to ESA’s operational orbits in January 2014, a known step is added there: it measured −4.8 ± 1.0 mm North at Wettzell.
-
Steadier long series. When a station’s site log is available, its antenna history replaces the RINEX headers’ (old headers spell one antenna several ways, each read as an antenna change). Archive stations keep the name you asked for whatever their old files call them. A step needs three days on each side. A series that can’t be fitted now says why.
2026-11-11
-
Germany, and national heights for the Nordic countries.
- An ETRF2000 result in Germany shows ETRS89 UTM 32/33 and DHHN2016 heights (BKG’s GCG2016).
- Sweden, Finland, Norway and Denmark now also show RH 2000, N2000, NN2000 and DVR90 heights.
- Height grids are at full resolution and load in small tiles around the point.
- Country detection now has rough outlines for Germany, France, Sweden and Finland. Before, Stuttgart showed Lambert-93 and Oslo showed Sweden.
-
Ireland, Portugal and Spain on /coordinates.
- Ireland: ITM, with heights above Malin Head and Belfast (OSGM15).
- Portugal: PT-TM06, with Cascais altitudes (GeodPT08).
- Spain: ETRS89 UTM 29–31, with Alicante altitudes (EGM08-REDNAP), for the mainland and the Balearics.
Each uses the national ETRS89 realisation that EPSG records. Dublin, Barcelona and the Spain–Portugal border now land in the right country.
-
Survey report (PDF) on /ppp. A static PPP result can now be downloaded as a report to hand to a client or file with a job. The PDF is built in your browser, and project, client and surveyor are remembered on your device. It contains:
- The point’s coordinates in ITRF2020 at the observation epoch, in the regional frame (ETRS89, NAD83) and in UTM.
- The country’s national grid and height system where gnsscalc has one, picked from where the point is: RD/NAP, British National Grid/ODN, Lambert-93/NGF-IGN69, SWEREF 99, TM35FIN, Norway, Denmark, MGA, NAD83(CSRS). You can choose another grid or none.
- The EGM2008 orthometric height.
- The session, the processing and its quality, with convergence and scatter charts.
- Your notes, up to four site photos and a sign-off block.
-
Swiss LV95 and heights on /coordinates. An ETRF89 (= CHTRF95) result in Switzerland now shows:
- CH1903+ / LV95.
- The LN02 and LHN95 heights, from swisstopo’s CHGeo2004.
You can also start from LV95 coordinates and an LN02 height. ETRF89 joins the frame list as “ETRF89 (Switzerland)”.
-
State Plane, Belgium and Poland on /coordinates.
- A NAD83(2011) result in the US shows its State Plane zone (all 117 SPCS83 transverse Mercator and Lambert zones) in metres, US survey feet or international feet.
- An ETRF2000 result in Belgium shows Lambert 2008.
- An ETRF2000 result in Poland shows PL-1992 and its PL-2000 zone, x (northing) first, as written there.
- All three accept typed input; the US input also takes feet.
-
Lambert-93 and NGF altitudes for France on /coordinates. In France, an ETRS89 (ETRF2000, which is what RGF93 v2b is) result now also shows:
- Lambert-93.
- The CC zone for its latitude.
- The NGF-IGN69 altitude from IGN’s RAF20 quasi-geoid, on the mainland only.
You can also start from Lambert-93 or CC coordinates. The national-grids card’s country picker now tells apart countries that share a frame.
-
The British National Grid on /coordinates. In Great Britain, an ETRS89 (ETRF97, as OS Net) result now also shows:
- National Grid easting and northing.
- The OS grid reference.
- The height above ODN, or above the local island datum.
This is Ordnance Survey’s OSTN15/OSGM15, using OS’s 1 km grid loaded in 100 km tiles. It reproduces all 80 of OS’s test results. You can start from an easting and northing or a grid reference, with an optional height.
-
RD and NAP for the Netherlands on /coordinates. An ETRS89 (ETRF2000) result in the Netherlands now also shows RD x, y and the NAP height, by RDNAPTRANS™2018 with NSGI’s grids. You can also type RD coordinates, plus a NAP height if you have one, to set the Point. Checked against NSGI’s 100 certified points: within 0.2 mm in RD and 0.05 mm in NAP.
-
National frames and grids for five more countries on /coordinates.
- Finland: EUREF-FIN, with TM35FIN and the ETRS-GK zones.
- Norway: EUREF89, with UTM 32/33/35 and NTM.
- Denmark: ETRS89-DNK, with UTM 32/33 and DKTM.
- Australia: GDA2020 and GDA94, with MGA.
- Canada: NAD83(CSRS), with UTM and MTM.
The Nordic frames use the same NKG2020 land-uplift chain as SWEREF 99, plus Kartverket’s shift grid for Norway. The SWEREF 99 card is now a national-grids card with a country picker. It opens by itself when the Point is in one of these countries, and you can type grid coordinates in each country’s usual order. All transformations and grids match PROJ to well under a millimetre.
-
Quality history on /station. Each IGS and EUREF station archived at BKG (~560 a day) now shows the /qc check of each of its daily files: one verdict per day, the checks of any day you click, and charts of tracking, multipath, slips, antenna C/N0 and continuity over time. Computed once a day and shared by everyone.
-
Fixed: BeiDou in RINEX 3.02 files. Files written in RINEX 3.02 (many Trimble NetR9 and Leica GR50 stations) label BeiDou B1I differently from later versions. It was read with the wrong frequency, so /qc reported cycle slips on half of all BeiDou epochs and graded those stations Poor (GMSD: 107 slips per 1000 → 0.3). Positioning with BeiDou from these files is fixed too.
-
Tracking no longer counts a receiver’s channel limit against the site. Some Leica receivers are set to record only 4 or 6 Galileo satellites at a time, so they tracked half of the Galileo satellites in view. /qc now detects this, grades tracking on the other constellations and names the limit.
-
SBAS no longer counts in the multipath grade. As already for cycle slips: one SBAS satellite with an anomalous signal on 9 November (19 m) graded 13 stations from Kyrgyzstan to Australia Poor. SBAS is also left out of the “worst direction” hint. A station tracking from ~15° all round is now described as probably the receiver’s own elevation mask, not as interference.
-
Fewer false cycle slips. A satellite already checked on a complete signal pair is no longer also tested on a pair that lacks code on one band (some Javad receivers record BeiDou B2I phase without code on the geostationary satellites: 3,836 false slips → 4). Satellites that aren’t in the broadcast navigation, while the rest of their constellation is, are left out of the grade (a NavIC satellite out of service caused all of KUNZ’s slips).
-
Interference is graded by what it costs. Poor now means an hour in total, or 10 minutes of a drop of 15 dB or more. A short, shallow dip is marginal (it can also be rain on the antenna).
-
Caster logins, in one place. Every tool that connects to a caster (/stream, /ntrip, the RTK base, /network) now uses the same logins: kept on this device by default, or for this tab only, and listed in Settings with whether the caster accepted each one, and Forget. A refused login now stops at once and says which caster refused it. Fixed: an RTK base on the rover’s own caster (two BKG stations, say) connected without the login; the ?rtk= link and the other-caster form now use the kept login too. The connect form’s note is corrected: logins pass through the relay (which stores and logs nothing), they don’t go “straight to the caster”.
-
/network redesigned as a watchlist, built on the station catalogue. Search the catalogue by id, name, city or network, pick on its map, or take the stations nearest to you — 484 permanent stations with a live stream; BKG’s and Kadaster’s casters ask for your (free) login once. The watchlist is a list beside a map, kept in your browser (never with passwords) and resumable; pick a station for its catalogue identity, signals, live position check and events, or open it in /stream. Every /station page with a stream has “Watch on /network”, and a streaming filter in the catalogue offers “Watch these on /network”.
-
/network watches 16 stations. Connections, decoding and live positioning now run in a background thread (a Web Worker): sixteen TU Delft stations keep the page itself under 3 % busy, and status alerts keep their timing when the tab is in the background.
-
Live Network Grid (/network). Watch up to eight stations at once — NTRIP mountpoints from any mix of casters, or raw TCP receiver streams — with a live OK / WARN / DOWN verdict per station and the reason, desktop alerts when one changes, an event log, a map and a compact table sorted worst first. Beyond “no data”, it notices a stream that is connected but has stopped sending something it used to (a message type, a whole constellation), a station announcing a moved position or new equipment, and C/N0 drops shared by several stations at once (regional interference). Pick stations on a map of the permanent-station registry or the ones nearest to you, save lists in your browser, and export the session as RINEX (every station, one zip), CSV or JSON.
-
/network and /stream share their streams. Stations keep running while you use other pages and show in the Live pill; “Open in /stream” hands one to the stream inspector on the same connection, nothing reconnects.
-
Faster first fix on some casters. A stream that broadcasts only a sliver of its own ephemerides (only QZSS or NavIC, say) no longer holds the live position back for 90 s waiting for the rest; it plots once its own coverage stops growing.
-
Local network mode on /campaign. Millimetre relative positions for a dam, a landslide or a monitoring grid. In every session each point is tied to a reference point by a static baseline with fixed integer ambiguities, small sessions add the baselines between points, and the session is adjusted by least squares with the reference held fixed. Bad baselines are found by a w-test and loops are checked. Each point’s series then gives its motion relative to the reference. At the Wettzell observatory it agrees with EUREF’s solution within 2 mm, with 1–2.4 mm day-to-day scatter.
-
RTK post-processing on /positioning applies antenna heights. Static baselines used to run from antenna to antenna even though the base position is the marker. Mixed receiver makes also gain Galileo (E1C against E1X), and BeiDou no longer biases them.
-
Known jumps from site logs and earthquakes on /campaign. Each station’s equipment history is read from its IGS site log (or EUREF’s, for European stations), and earthquakes close enough to move it come from the USGS catalogue (within 10^(M/2 − 0.79) km: 160 km for M6, 510 km for M7). Antenna changes and earthquakes become steps in the fit, each listed with its cause; receiver changes are listed but not forced. An earthquake sequence between two sampled days becomes one step, which used to make the fit fail outright.
-
Older years on /campaign. Days before about 2020 are archived as RINEX 2 files compressed with Unix
compress(.Z), which the site couldn’t read. It reads them now, everywhere files are accepted, so /campaign reaches back to 2014, where ESA’s multi-GNSS precise orbits begin. -
Episodic campaigns on /campaign. Dropped files now open a review table first: what each header says about the occupation (station, day, antenna and radome, antenna height, receiver), editable, with antennas that lack an IGS20 calibration flagged. Field RINEX headers are often wrong, and a wrong height moves the solution by exactly that much. A change of antenna height alone no longer counts as a jump in the series: heights are applied, and in a tripod campaign they differ at every visit. New option: fix ambiguities (PPP-AR) with CNES’s products, worth it for sessions under about 6 hours.
-
Fewer good days rejected as outliers. The outlier test was far too eager on short series: clean 12-day series lost a day 14% of the time, now 1%. Real blunders are still caught.
-
Ambiguity fixing reports fixed sessions as fixed. On /ppp and /campaign, a static session whose very last fixing attempt failed was reported as float even when it was fixed most of the time. It now uses its last accepted fix (Brest, 2 hours: height error 33 → 22 mm).
2026-11-10
- /campaign redesigned, and much faster. Station-days now download and solve on several background workers at once, and the series are fitted in the background too, so the page never freezes (it used to stall for about a second per day). A 72-station-day network now takes 24 s instead of 80. Stations are entered as chips, and the ones you type stay the stations of interest when you add reference stations. A live grid shows every station-day as it is downloaded, solved or failed, with the time left and “Retry failed”. Results stream in as days are solved, and the page works on phones. The archive is chosen automatically: Wettzell, the default example, failed in the EUREF archive, which lists it but only serves it from the IGS one.
- Automatic ocean tide loading. /ppp and /campaign now correct ocean tide loading everywhere without pasting coefficients: our own global grid (GOT4.10c tides loaded onto an elastic Earth, 1.8 MB of tiles, fetched only for the station’s area) agrees with Onsala-style coefficients to 0.3 mm in the median and cuts Brest’s 2-hour height scatter from 48 to 16 mm. Pasted BLQ still takes precedence; the PDF report says which was used.
- SINEX export on /campaign. One click exports the stations’ positions and velocities with their uncertainties as SINEX 2.02, the format IGS, EUREF and survey software exchange. Wettzell’s exported velocity agrees with its official IGS20 one within its error bars.
- Campaign reports and ocean tide loading. /campaign can download a PDF campaign report: the processing, the reference stations, velocities with realistic error bars, a time-series chart per station with jumps and outliers, and the daily coordinates. /ppp and /campaign also accept a station’s ocean tide loading coefficients (BLQ, from Onsala’s free service); at a tidal coast like Brest that cut the height bias of 2-hour PPP from 25 to 5 mm.
- Campaign processing: realistic error bars. Velocities on /campaign now carry a realistic ± that allows for the correlated (flicker) noise in GNSS series, instead of the white-noise error that typically claims 2–5× too much precision. Networks can also be aligned with a 7-parameter Helmert, which absorbs the orbit product’s frame scale (Wettzell’s height came within 2.3 mm of IGS20).
- Campaign processing: networks (phase 2). /campaign takes several stations at once; ”+ nearest references” adds the closest IGS20 stations, each day is aligned to them (removing the error the region shares), bad references are screened out with the reason shown, and a map shows the velocity field in ITRF2020 or relative to the plate. On Wettzell with five neighbours, alignment cut the horizontal scatter by a third and brought the two-year velocity to within 0.1 mm/yr of IGS20.
- Campaign processing (new page, phase 1). /campaign turns many days of one station into a coordinate time series. Drop daily RINEX files, or name a station in the IGS/EUREF archives with a date range and a step. Each day is solved with static PPP, then gnsscalc fits the velocity (ITRF2020 and relative to the regional plate frame), seasonal motion, jumps (antenna/receiver changes are found from the headers) and outlier days, with charts and a CSV. On Wettzell, 3.4 years of monthly days matched the official IGS20 velocity to 0.1 mm/yr horizontally.
- Fixed: /ppp with receivers whose clock drifts. Satellite positions were timed without the receiver’s clock offset, which Septentrio (by default), some Javad receivers and phones let drift up to a millisecond. That put up to ~10 cm of East error into their PPP results, varying from day to day. Steered-clock receivers were unaffected. Found while validating campaign processing against IGS stations (GRAZ was 11 cm off; now 2–9 mm).
- Android phone logs. Raw GNSS logs from Google’s GnssLogger app
(
gnss_log_….txt) now open everywhere a receiver log does: /convert turns them into RINEX, /qc analyses them, and /ppp runs on dual-frequency phones that log carrier phase (GPS L1+L5, Galileo E1+E5a). The phone’s own fix seeds the approximate position. Pseudoranges match Google’s reference processing exactly. - SBAS for dropped receiver logs. When a u-blox, Septentrio or NovAtel log carries SBAS messages (EGNOS, WAAS…), the single-point positioning on /qc now also solves every epoch with those corrections. You can switch between plain and SBAS, compare their RMS, and see the protection levels (HPL/VPL) next to the actual error. On 90 minutes from TU Delft’s DLF500, the vertical RMS dropped from 2.3 to 1.0 m, and the protection levels bounded the error in every epoch.
- /ppp can fix ambiguities (PPP-AR, beta). A new option resolves the carrier-phase ambiguities to integers using CNES’s integer-clock products (GFZ orbits and clocks with CNES code/phase biases; a ~45 MB download, published about a week after the observation day; GPS + Galileo). It helps most in short static sessions: on seven IGS stations, 1-hour files went from 1.7 to 0.8 cm median horizontal error, and 30-minute files from 2.7 to 1.0 cm. The result says whether the ambiguities fixed and when. When the products aren’t out yet, /ppp runs float as before and says why.
2026-11-09
-
/ppp is now centimetre-accurate. Static PPP now applies satellite and receiver antenna phase-centre calibrations (IGS20), the antenna height from the RINEX header, solid-earth tides and phase wind-up, and several model errors in the engine were fixed. On five IGS stations a 24-hour file lands within 0.2–0.8 cm of the station’s ITRF2020 coordinate (it was 10–33 cm off); 6 hours give about 1–2 cm, 1–2 hours about 4 cm. Kinematic mode, which was metres off per epoch, now tracks to a few centimetres once converged. The result shows which antenna calibration was used and the antenna height.
-
Clear reference frames everywhere. Coordinates now carry a small tag saying which frame (and, for ITRF, which epoch) they are in, and one site-wide “Show coordinates in…” choice — ITRF2020/WGS84, ETRS89 or NAD83, defaulting to your region’s — applies to /ppp results and the base coordinate card. On /stream, the station’s announced coordinate is read in its own frame: declared by RTCM, detected once Galileo HAS PPP converges (ETRS89 vs ITRF differ by ~0.85 m in Europe), or set by you — so a regional frame no longer shows up as a metre of error. RTK exports state that they inherit the base’s frame, and /methods explains it all.
-
“Is my base where it says it is?” on /ppp. A static PPP of a base station’s own data is now compared with the coordinate the station announces — RTCM 1005/1006, the Septentrio receiver setup or the RINEX header — in each reference frame (ITRF2020, ETRS89, NAD83), so a base set up in ETRS89 isn’t flagged for the ~1 m frame difference. It tells you whether the base is right, slightly off or misplaced, by how much in which direction, and opens the base coordinate in the frame it was set up in. Septentrio logs and RTCM captures now also get their antenna calibration and height applied (from the receiver setup / RTCM 1006 and 1033).
-
Compact RINEX (Hatanaka) files decode correctly in every tool. Files whose satellite count changes between epochs could read some satellites’ observations under the wrong satellite. All tested IGS files now decode exactly like the reference
crx2rnx. Satellite ids written asG 2are recognised too. -
Galileo authentication on /stream. For Septentrio streams, the Quality tab now verifies Galileo’s OSNMA signatures live — public key, signed root key, chain keys and each satellite’s navigation data — and a banner warns if any data fail authentication (a spoofing symptom). Septentrio files dropped on /qc get the same verification.
-
Live HAS PPP on /stream uses corrected GPS satellite antenna offsets.
-
BeiDou-3’s new navigation messages in RINEX 4 exports. Septentrio logs dropped on /convert now also yield the BeiDou-3 B1C, B2a and B2b ephemerides (RINEX 4 CNV1/CNV2/CNV3 records), counted under “CNAV records” and written to the RINEX 4 navigation export. RINEX 4 nav files added to /convert keep their CNAV and CNV1–3 records through the export too.
-
Live quality on /stream works again — with the same charts as /qc. The Quality tab’s cycle-slip, multipath and ionosphere panels reported “no carrier phase / no dual-frequency observations” on healthy streams: an elevation-mask unit mix-up hid every satellite. Fixed, and the tab now shows /qc’s own charts over the live stream. Quality runs in the background (as does live RTK), the charts update every few seconds without stalling the page, and a pulse shows it’s live.
-
Satellite trails on the live sky plot, and one 10° elevation mask across the sky plot, the quality grading and the live solvers.
-
Live RTK reconnects a silent base. A base stream that stays connected but stops sending is now detected after 20 s and reconnected.
-
MGRS and batch conversion on /coordinates. Positions now show their MGRS reference, and a list of points (latitude/longitude or ECEF, pasted or dropped as a file) converts in one go to geodetic, ECEF, UTM and MGRS as a CSV.
-
Check & repair in /convert. Resampling lands on the interval’s grid (:00/:30 for 30 s), as teqc and gfzrnx do; files from receivers hit by a GPS week-rollover bug can be shifted back to their real dates; and the header editor can rewrite antenna and receiver names to their exact IGS spelling. Also fixed: removing sparse signals could silently drop them from later exports of the same file.
-
Geoscience Australia in the /qc archive browser — about 1,000 ARGN/APREF stations a day across Australia, the Pacific and Antarctica, with map pins for most.
-
Compact RTCM (MSM1–3) streams and files now work. Some bases send the compact MSM messages, which carry each satellite’s range only to within a ~300 km ambiguity; they used to show no observations at all. /stream, live RTK, /network and /convert now resolve them from the station’s position and broadcast orbits — positions, quality checks and RINEX recordings work on these streams like any other.
-
Co-located mountpoints share one marker on the /stream caster map, listing every stream at that spot with its own Connect button.
-
/network keeps caster passwords out of the station list when you add mountpoints from the caster browser.
-
Multipath scatter tooltips on /qc no longer read “RMS = NaN” or show units twice.
2026-11-07
- Survey your base station on /positioning. PPP now reads receiver logs directly — u-blox UBX, Septentrio SBF, NovAtel, Trimble, BINEX or an RTCM3 capture — as well as RINEX, starting from a code solution on the precise orbit when the file has no header position. A static result now ends in a “use as your base-station coordinate” card: the coordinate in ITRF2020, ETRF2000/ETRF2020 or NAD83(2011), ready to paste into RTKBase/Centipede, RTKLIB str2str, a u-blox ZED-F9P fixed mode (TMODE3) or a Septentrio receiver, with the EGM2008 orthometric height and an honest note on what decimetre-level PPP is good for. Converted logs can be downloaded as RINEX for NRCan CSRS-PPP or AUSPOS when you need centimetres — including when the precise orbit is not out yet.
- Interference watch on /stream. Live streams are now checked for a C/N0 drop on most satellites at once — the signature of a jammer — against a slow per-signal baseline, per frequency band. A banner raises it from any tab; the Quality tab logs each event with its duration, peak drop and bands. Over 30 recorded TU Delft streams it raised no false alarms, and it caught an injected 8 dB L1 drop on every one within 2 seconds.
- New /methods page — how each tool computes what it shows, and what each decoder and engine has been validated against (RTKLIB, teqc, HASlib, IGS products), including where the checks stop.
2026-07-24
- Sharper live positioning on /ntrip. The live single-point position now models the ionosphere from a global ionosphere map (GIM) and the troposphere with a Saastamoinen model, replacing the broadcast Klobuchar and fixed-zenith approximations — on a mid-latitude station the height error drops from several metres to sub-metre. File-based positioning benefits from the new troposphere model too. (gnss-js 1.25.)
- /convert inspector header fixed. The frozen column header in the message field-by-field view was transparent, so rows scrolled through it; it is now solid.
- /convert now opens any receiver file, even with no observations. A navigation- or almanac-only log (e.g. an SBF almanac dump) used to fail with “no observation epochs”; it now shows the full record census, opens the field-by-field inspector, and offers a RINEX navigation download when it carries ephemerides — so you can see what’s inside a file regardless of what it holds.
- Live RTK on /ntrip — connect a base and a rover mountpoint for a real-time carrier-phase baseline: FIX / FLOAT / DGNSS status, ratio, per-epoch ΔENU and a horizontal scatter of fixed solutions. Integer ambiguity resolution runs in the browser (gnss-js 1.24, MLAMBDA). Verified live against two TU Delft stations — fixed solutions within ~2 cm of the surveyed inter-station baseline.
- Visibility planner forecasts BeiDou (and GLONASS when available) — future-date mode already covered GPS and Galileo from public agency almanacs; it now also pulls a TU Delft reference receiver’s SBF almanac feed, adding BeiDou and — whenever the receiver has broadcast them — GLONASS. Constellations light up in forecast mode exactly when almanac data exists for them; orbits stay km-class, pass times good to seconds.
2026-07-23
-
The visibility planner now works for future dates — pick next week and plan: positions come from broadcast almanacs (GPS from the US Coast Guard’s daily YUMA file, Galileo from the European GNSS Service Centre), propagated with the new gnss-js almanac engines (validated against the ICD’s own worked example and real receiver almanacs). Km-class orbits — pass times good to seconds, exactly what planning needs. GLONASS/BeiDou show as unavailable in forecast mode (their public almanac feeds don’t exist; a receiver-file source is planned).
-
Fixed: BeiDou GEO satellites (C01-C05, C59-C63) could be positioned hundreds of km off (gnss-js 1.23) — GEO records whose broadcast inclination exceeded an internal threshold were propagated with the wrong frame convention. Affected BDS GEO orbits on /rinex, sky plots, visibility and SPP whenever such records were in use; positions now verify against an RTKLIB reference harness at the metre level. gnss-js 1.23 also ships RTK stage 1 (DGNSS + carrier float engine, validated to centimetres against RTKLIB on a public base/rover pair) — the /ntrip live-RTK feature builds on it next.
-
u-blox files are now four-constellation navigation sources (gnss-js 1.22): Galileo I/NAV, BeiDou D1/D2 and GLONASS strings decode from RXM-SFRBX raw frames alongside the existing GPS path — a single F9P drop yields the full nav picture (validated: 62/62 E, 17/17 C, 37/37 R records vs RTKLIB on a 75-min capture). gnss-js 1.20/1.21 also land the same raw-frame decoding for Septentrio logs, almanac propagation for all constellations, and Galileo HAS decoding (E6-B → Reed-Solomon → precise corrections; 49,630 fields identical to the FGI reference decoder) — app-facing HAS features to follow.
-
RINEX 4 navigation export with GPS CNAV (gnss-js 1.19): receiver drops now also assemble CNAV ephemerides from raw L2C/L5 bits (Septentrio raw blocks, u-blox SFRBX; validated 21/21 against IGS RINEX 4 files — RTKLIB cannot decode these at all). /convert shows the CNAV count, and selecting RINEX 4 writes a v4 nav file carrying the CNAV and ionosphere records alongside the classic ephemerides.
-
Fixed: exported GLONASS navigation records carried a flipped clock sign (the v3 writer double-negated τₙ). If you exported a nav file from /convert or /rinex with GLONASS records before today, re-export it — clock-dependent processing of those satellites was biased by 2·τₙ (microsecond-scale, hundreds of metres in range).
-
NovAtel logs now yield GPS ephemerides from GPSEPHEM (gnss-js 1.18) — the decoded-fields log many OEM7 captures use instead of RAWEPHEM; validated 42/42 against same-day IGS BRDC files. The public OEM719 sample now decodes a complete GPS+GAL+BDS navigation set from one file. gnss-js 1.18 also decodes GPS CNAV (L2C/L5) from Septentrio raw blocks and u-blox SFRBX (IS-GPS-200 implementation — RTKLIB has none; 21/21 records vs IGS RINEX 4 files) — RINEX 4 export lands next.
-
The inspector’s NovAtel coverage now includes GALEPHEMERIS, BDSEPHEMERIS, GPSEPHEM/QZSSEPHEMERIS and IONUTC field tables — drift-guarded like the rest.
-
Message inspector on /convert: the record census is now clickable — browse every instance of a message type (own timestamp, size, CRC verdict) and dissect it field by field with vendor-doc names, scale factors and units, next to a hex dump that highlights each field’s bytes on hover. Full layout tables for observation, ephemeris and almanac messages on all three formats — including NovAtel RANGECMP’s bit-packed records and RAWEPHEM opened down to the individual GPS LNAV subframe fields; every other known type shows its dissected frame header + hex. Layout tables are drift-guarded by tests against the oracle-validated gnss-js decoders.
-
Single-file drops are now complete navigation sources (gnss-js 1.16/1.17): u-blox ephemerides assemble from raw RXM-SFRBX subframes (22/22 vs RTKLIB on a full F9P capture), NovAtel adds Galileo/BeiDou/QZSS ephemerides (107/107 on a public OEM719 capture), and all three formats now deliver the broadcast Klobuchar coefficients and leap seconds — so SPP on a dropped receiver log applies the iono correction (historically ~2× better 3D RMS) and exported nav files carry proper GPSA/GPSB/leap-second header lines.
-
/convert: one Download RINEX button for the whole dataset (obs in the selected version + nav when present) instead of two.
-
Fixed: a truncated final frame in a receiver log (e.g. the classic RTKLIB OEMV sample, cut at exactly 256 KiB) crashed the inspector.
-
Format detection is stricter and more tolerant at once: it now requires two checksum-valid frames (text can’t misdetect) and scans the first 128 KB, so captures that open with wrapper preamble — a public OEM719 file starts 4.6 KB in — are recognized.
-
Navigation data straight from receiver logs (gnss-js 1.15): Septentrio SBF (GPSNav/GLONav/GALNav/BDSNav/QZSNav) and NovAtel (RAWEPHEM, GLOEPHEMERIS) navigation blocks are now decoded alongside the observations. A single SBF or NovAtel drop on /rinex brings its own ephemerides — satellite orbits, SPP and Doppler prediction with no separate BRDC file — and /convert gains a “Download NAV” button that writes them as RINEX 3.04 NAV. Decoding is validated field-for-field against RTKLIB conversions (83 records, zero disagreements) and cross-checked against Septentrio’s own sbf2rin. SBF almanac blocks (GPS/GAL/GLO/BDS) decode in the library too — groundwork for planning beyond ephemeris validity.
-
Receiver file converter (/convert): u-blox UBX (RXM-RAWX), Septentrio SBF (Meas3 + MeasEpoch) and NovAtel (RANGE/RANGECMP) logs → RINEX 2/3/4, entirely in the browser. Formats are detected from frame syncs, gzip accepted. The same logs can be dropped straight onto /rinex for the full QC/SPP/export pipeline — one decoder path, two doors. Decoders (gnss-js 1.14) are validated observable-for-observable against RTKLIB conversions of reference datasets.
-
Live SPP card: ΔE/ΔN/ΔU now share one timeline chart.
-
/convert now lists every record in the file with its count — message and block names resolved from the vendor documentation (111 SBF blocks, 500+ NovAtel logs, common UBX messages): see at a glance whether a log carries almanacs, ephemerides, raw nav frames or just observations.
-
Fixed: live SPP on /ntrip was kilometres off (and stream-recorded RINEX epoch tags were 18 s early) — the MSM epoch conversion returned UTC where the whole pipeline expects GPS-scale time, computing every satellite 18 s early (gnss-js 1.14.1). Validated against a public caster: 1.1/1.0/4.8 m E/N/U RMS vs the station’s surveyed position. The live-SPP BRDC warm-up also walks back days now instead of failing silently at night.
-
Live positioning on /ntrip: while streaming, the page now solves SPP from the stream’s own decoded pseudoranges in real time — current fix, horizontal scatter and Up timeline against the station’s broadcast position (RTCM 1005/1006), with today’s BRDC seeding ephemerides so the first fix lands in seconds instead of minutes, and Klobuchar applied.
-
Broadcast orbit accuracy monitor on /status: pick a day and see broadcast-minus-precise orbit errors (vs ESA rapid/final products) per satellite, decomposed radial/along/cross with per-constellation RMS and orbit-only SISRE — the CoM/APC radial bias separated out honestly — plus broadcast clock errors against the precise clocks (constellation-ensemble and per-satellite bias removed) and the full SISRE per constellation. Click a satellite for its radial/along/cross/clock time series (GLONASS’s 30-minute sawtooth is clearly visible).
-
RINEX observation writers (2.11/3.04/4.01) moved into gnss-js 1.12.0, which also gains an SP3 precise-orbit parser with Lagrange interpolation — groundwork for a broadcast-vs-precise orbit comparison view.
2026-07-22
- Doppler chart on /rinex — observed D-observables per satellite, plus an observed-minus-predicted residual view from broadcast-orbit range rates (needs a nav file; validated at ±0.07 Hz on BRUX).
- ROTI mode on the ionosphere card — 5-minute sliding std of the rate of TEC, the standard irregularity/scintillation index.
- “vs GIM” mode on the ionosphere card — measured slant TEC against the ESA rapid GIM at each pierce point; exposes DCB-calibration bias directly.
- SPP now applies the broadcast Klobuchar model (gnss-js 1.11.0) — on a 24 h single-frequency test the 3D RMS dropped from 6.1 m to 2.4 m.
- Visibility planner buildout (/visibility): map location picker with search and shareable deep links, start time with midnight-crossing BRDC merge, paintable per-azimuth horizon mask, unified satellite selector with constellation groups and health flags, per-constellation visible-count and DOP charts, pass Gantt, sky plot + world view with comet-trail animation, elevation chart with cutoff band and satellite hover/pin, ionosphere card (TEC world map, TEC at your location vs the broadcast model, NOAA Kp strip).
- SPP folded into /rinex with a ΔENU accuracy plot; the separate /spp page now redirects there.
- Animated TEC sky plot on /rinex.
- Archive browser on /rinex — browse IGS/EUREF/NOAA archives by day with a station map, one click loads obs + matching nav.
- NTRIP proxy relay: unit-tested decision logic (SSRF guard, cache TTLs, chunked-body handling).
- Archive browser: all TU Delft DPGA stations now appear on the station map (coordinates taken from their own RINEX headers).
- Every chart can now be downloaded as a PNG (hover a chart card for the button).
- Deep links for /positioning (
?x&y&z&rx&ry&rz) and /spectrum (?band&sys), with Share buttons like the planner’s. - UTC/local time toggle on the NMEA track statistics.
- The processing backdrop now also covers /nmea and /antex loading.
2026-07-21
- Ionosphere view modes: per-arc TEC variation, rate of TEC (ΔTEC/Δt), and second difference.
2026-07-20
- DCB-calibrated slant TEC on the RINEX ionosphere card (ESA bias product; receiver bias from the night floor).
- Load RINEX files directly from archive URLs (
?url=deep links). - Ionospheric slant TEC chart on the RINEX QC page.
- Satellite visibility planner (first version) and single-point positioning pages.
- Reference-frame transform tool (ITRF/ETRF, 14-parameter Helmert) on /positioning.
- NTRIP client deep links; navbar grouped into dropdowns.
Earlier
- Per-code-band multipath, cycle-slip and completeness QC; RINEX header editor, export/convert, PDF report (/rinex).
- Live broadcast ephemeris status page with all decoded fields and satellite generations (/status).
- NMEA reader with track map and GPX export; ANTEX PCV viewer; NTRIP sourcetable browser and stream monitor; GNSS signal spectrum viewer; GPS time tools.