PositionLatitude / longitude
The phone’s geographic position. Useful for plotting the flight and matching it to nearby places; a noisy fix is still an estimate, not a precise point.
THE ZEPHYR FIELD GUIDE · RESEARCHED 25 OCT 2026
Zephyr is a phone-only flight recorder. Here is what iPhone, iPad and Android devices can measure, what phone sensors and weather can help us estimate, and which measurements sit outside that scope.
Archived research snapshot — 25 October 2026. The implementation labels and findings below describe that earlier development version, not the current mobile app. Read current release and support information for today’s availability. The mobile apps have not been publicly released.
Timestamped location, groundspeed, altitude and position accuracy; derived distance, climb and time; an armed liftoff/landing heuristic with manual override; route review and sharing. Existing manual TAS and field pressure-altitude calculators are separate from the planned automatic estimates.
Course/heading capture, barometer and motion recording, phone-estimated maximum bank, weather-based estimated TAS, satellite diagnostics and terrain. IAS and aircraft connections are excluded. The browser is a demo; signed iOS and Android releases are not available yet.
Both platforms expose a location estimate assembled by the operating system. It may use GNSS, network positioning and other sensors. Calling every fix “raw GPS” overstates what we know. Apple Core Location · Android Location
The phone’s geographic position. Useful for plotting the flight and matching it to nearby places; a noisy fix is still an estimate, not a precise point.
When a fix was produced, which may differ from when the app received it. Retain original timestamps, detect gaps and reject stale fixes; Android also offers an elapsed-realtime clock.
The phone’s speed relative to the ground, useful for a flight profile. This is not indicated or true airspeed. Invalid readings must not become zero-speed events. Apple speed
Where the track moves relative to true north, not where the nose points. Course change over time yields ground-track turn rate. Wind changes the ground path, so groundspeed and course alone cannot generally establish actual bank. Apple course · FAA ground reference maneuvers
Height relative to the API’s reference surface. Apple exposes separate MSL and ellipsoid values; Android’s traditional altitude is WGS84 ellipsoid height. Normalize the datum before comparing to terrain. Altitude reference note
A quality estimate in meters, not a guarantee of accuracy. Use it to qualify route detail and climb estimates. Apple marks altitude invalid when vertical accuracy is zero or negative. Apple vertical accuracy
Available platform fields can express how uncertain speed and bearing are. Store missing values as unavailable, not as perfect precision; smooth trends only after checking quality. Apple speed accuracy · Android availability
Android can report satellite use, constellations and carrier-to-noise density on supported devices. Standard iOS Core Location does not offer the same public satellite-status feed. Source/mock indicators are useful provenance, not proof against cheating. GNSS status · Apple source information
These are Zephyr algorithms and product opportunities, not extra phone sensors. The estimates are only as good as the samples beneath them. Long gaps, taxiing and a poor fix must stay visible in the review.
Connect accepted fixes to show the route and accumulate horizontal distance. Filtering should remove impossible jumps without cutting legitimate turns. A gap should not quietly become a perfectly recorded straight line.
Calculate time between detected or manually corrected events. Keep airborne time distinct from recording duration; neither automatically equals Hobbs, tach or legally required logbook time.
Combine sustained movement and altitude change to propose a takeoff event. A classifier needs aircraft-specific testing, confidence and correction; driving quickly or a GPS jump can look similar.
Propose the end of an airborne segment from the track. Touch-and-goes, low passes and brief stops need separate handling. A detected landing time is not a wheel-contact measurement.
Estimate vertical speed from altitude changes over a time window. Smoothing reduces noise but adds delay. It cannot make poor altitude data equivalent to a calibrated aircraft vertical-speed instrument.
Infer taxi, climb, cruise and descent; suggest circuits and airport visits using a vetted airport/runway database. Let the pilot correct labels before these affect totals or achievements.
Build consistent distance, duration, climb and airport-history summaries. Compare like aircraft and keep questionable segments out of records. Useful progress metrics do not require rewarding increasingly steep banks.
Review the route and turn flight statistics into a shareable image. Synchronized playback, source badges and privacy trimming around home airports would add depth to the current route/share experience.
Engineering inference: confidence-aware event detection, gap handling and filtering need recorded test flights with known outcomes. Phone-platform documentation establishes the inputs; it does not validate Zephyr’s takeoff or landing algorithm.
FOUR DISTINCTIONS THAT MATTER
QNH is an altimeter setting, not a measurement of the wind or airspeed. At the same QNH and 100 kt groundspeed, a 20 kt headwind means 120 kt airspeed; a 20 kt tailwind means 80 kt in level flight along the wind. The pressure setting cannot tell those cases apart. IAS is excluded from Zephyr. NWS pressure definitions · U.S. Navy wind triangle
Our phone-only research path is estimated TAS from ground velocity minus a weather wind vector, matched to position, altitude and time. That yields horizontal airspeed and can approximate TAS near level cruise; significant climb or descent needs additional treatment. Forecast uncertainty must stay visible, and nearby surface METAR wind must not stand in for wind aloft. The existing manual airspeed/pressure-altitude/temperature calculator is a different method. Aviation Weather Center data guide · FAA airspeed definitions
A rigidly mounted, calibrated phone can support a gyro-based bank estimate. Start from a known aircraft attitude, use all three gyro axes to follow orientation in three dimensions, and correct bias and drift with a flight-tested fusion method. Mounting alone or integrating one axis is insufficient. Sustained turn acceleration, vibration and magnetic interference must be handled rather than mistaken for gravity. Android motion sensors · Apple device motion · Android attitude limitations
Label the result estimated bank / estimated maximum bank, with quality checks and no unqualified peak from a moved phone. An independent attitude reference can validate the algorithm during development; it is not a required aircraft connection when Zephyr runs.
A barometer senses its local environment. In a pressurized cabin, its reading follows cabin pressure, not the aircraft’s outside static pressure. Even an unpressurized cockpit needs installation and airflow validation. Do not convert a cabin reading into claimed aircraft pressure altitude. FAA airframe handbook: pressurization
Apple exposes separate altitude values relative to mean sea level and the WGS84 ellipsoid. Android’s traditional altitude is ellipsoid-relative, with separate optional MSL fields on newer APIs. Those reference surfaces can differ; terrain subtraction needs a compatible datum. Apple altitude fields · Android altitude fields · NOAA geoid models
Current implementation finding: a local source audit on 25 October 2026 found that Zephyr’s installed Expo location library (58.0.19) passes MSL altitude through its iOS foreground-watch API, but ellipsoid altitude through its iOS background-task recording API and on Android. These are recording paths, not simply whether the screen is visible. Zephyr does not yet store a reference marker. This finding applies to the inspected library paths, not every version or configuration.
Hardware varies. These signals are technically available on supported phones, but Zephyr does not currently record barometer or IMU data. Most describe the phone itself; translating them into aircraft quantities needs more work.
Pressure around the device. Apple reports kilopascals; Android pressure sensors report hectopascals. Convert units explicitly and preserve the original value. Not every Android handset includes a barometer. Apple pressure · Android environmental sensors
Track pressure-related height changes from a starting point. Apple exposes relative altitude; Android can derive a pressure-based estimate. Weather, cabin pressure and sensor drift can all change the interpretation. Apple relative altitude · Android calculation
Apple can supply sea-level-relative absolute altitude with uncertainty and recommended resolution on supported hardware. Check availability at runtime; the API is not an aircraft altimeter setting or a certified pressure-altitude source. Apple absolute altitude
Sensor fusion can estimate device orientation in a defined reference frame. Record axes, mount alignment and calibration state. Device roll is not yet aircraft bank. Apple attitude · Android orientation
How fast the phone rotates about its axes. Useful for a synchronized motion trace or detecting handling. Integrating a gyro alone accumulates drift; a turn rate about a phone axis is not automatically heading rate.
Record acceleration along each phone axis, and supported gravity-compensated estimates. Handling, vibration, mounting and filtering affect peaks. Do not label a raw peak “aircraft G” or landing quality. Apple motion components
The direction a configured phone axis points. Nearby metal and magnetic fields can interfere; true-heading conversion also needs location. This is separate from course over ground and from the aircraft’s nose. Apple true heading · Android magnetic sensors
Estimate field pressure altitude from entered field elevation and altimeter setting. That result describes the field conditions, not current flight pressure altitude. A phone barometer in a pressurized cabin cannot supply outside static pressure. NWS pressure definitions
These are data integrations, not quantities a phone can sense by itself. Show the station or dataset, observation time, distance, units and age. A nearby airport observation is not a measurement at the aircraft.
Match position to airport coordinates and published field elevation. Useful for suggested departure/arrival labels. Nearest does not necessarily mean visited; runway geometry and pilot confirmation improve the match.
Show the latest reported setting from nearby stations with station ID, distance and observation age. This is reduced pressure for altimeter use, not raw pressure sensed by the phone. NWS METAR fields
Display airport wind and, with runway orientation, estimate headwind/crosswind components. METAR direction is referenced to true north; runway labeling and other weather sources can use different references. Surface wind is not wind aloft.
Combine a station’s reported temperature and pressure/elevation inputs for a field density-altitude estimate. A phone temperature reading is not aircraft outside-air temperature; an airport reading is not necessarily representative at altitude. NWS altitude definitions
Attach observed visibility, cloud layers and derived flight-category context to a flight review. Clouds reported in METAR use height above the observing station. Keep observations distinct from conditions all along the route.
Bring in valid-time forecasts for a preflight or retrospective view. Position-, altitude- and time-matched winds aloft could also support estimated TAS. Keep forecast age, resolution and uncertainty visible; this would still be an estimate.
Query ground elevation under a coordinate, then subtract it from an altitude on a compatible vertical datum. Terrain resolution, obstacles and altitude uncertainty all matter. This is an estimate, not a radio-altimeter reading. NOAA datum conversion
Overlay relevant runway geometry, airspace boundaries and effective dates in a review. Vertical references and data currency need explicit handling. Phone position alone does not know a runway’s dimensions or current airspace status.
A useful Zephyr feature: a “conditions at this moment” strip could pair each flight segment with the nearest suitable observation and its age. The matching would be a disclosed estimate, especially between distant stations or across weather changes.
Zephyr uses the phone and supporting weather/terrain data. Estimated maximum bank remains planned; weather-based estimated TAS is a research path. IAS and aircraft connections are excluded. The entries below make those decisions explicit without confusing a phone estimate with an aircraft instrument reading.
IAS is an aircraft airspeed-system reading. Phone GPS and QNH do not provide it, and weather-based estimated TAS is not IAS. Zephyr will not present an IAS instrument or add an aircraft connection to obtain it. FAA airspeed definitions
CAS corrects indicated airspeed for instrument and position errors using aircraft-specific calibration. Without the underlying measurement, a phone-only app cannot supply a direct CAS reading. This is a scope boundary, not a planned aircraft integration.
Subtract the weather wind vector from the GPS ground-velocity vector. The magnitude estimates horizontal airspeed, approximating TAS near level cruise. QNH alone is insufficient. Show source age and uncertainty; suppress an unsupported estimate. Wind-triangle basis
Match forecast wind to position, altitude and time, converting its direction into the same reference frame as the ground track. Wind resolution and forecast error limit estimated TAS; surface METAR wind is not a substitute aloft. NOAA weather data
Initialize a calibrated, rigidly mounted phone to a known attitude; follow orientation using all three gyro axes with bias/drift correction and flight-tested fusion. Qualify the peak by signal quality and reject moved-phone samples. Aircraft hardware is not a runtime requirement.
The same validated 3D phone orientation could support a pitch/roll trace; heading needs careful reference and magnetic-error handling. Track angle is not aircraft pitch, and phone acceleration alone is not a calibrated slip/skid reading.
Zephyr can research phone acceleration, but will not turn its raw peak into a claimed aircraft load factor or angle of attack. These are different physical quantities requiring validation and inputs beyond this phone-only scope.
The phone does not directly measure fuel flow or engine temperatures. Aircraft telemetry connections are excluded, and a microphone is not a replacement for those instruments. This information will not be presented as automatically recorded by Zephyr.
The experience can feel consistent while the collection layer handles different hardware, permissions, units and lifecycle behavior. Runtime capability checks are part of the product.
Before altitude-based features ship: normalize the different references in iOS foreground/background recording and Android, then preserve the reference and uncertainty with every sample. At the time of that audit, the engine could retain a numerical altitude after its vertical accuracy becomes unavailable, and some derived uses accept that missing quality. Harden those validity checks before flight validation; add terrain and estimated AGL only after the reference and quality handling is consistent. These are findings from the local source audit of 25 October 2026.
“Works on my phone” is the beginning of validation. Screen lock, a warm cockpit, lost service and a moved mount all affect the quality of a flight.
Validation should cover: representative iPhones and Android devices; the actual aircraft and mount; screen lock and process interruption; no network; degraded GNSS; taxi, low passes and touch-and-goes; long recording sessions; and comparison against time-aligned aircraft instruments. Numerical error bounds should come from those tests, not from the number of decimal places in the UI.
The opportunity is a clear, trustworthy flight story: what happened, what the data supports and what the pilot can learn from it. This ordering is a product recommendation based on the capability limits above.
Verify the altitude reference, improve sample quality/gap handling, persist recovery state, and validate liftoff/landing across aircraft. Give pilots an easy way to correct events. These improvements make every later metric more useful.
Collect course and available quality fields; combine route, altitude, groundspeed and climb profiles on one timeline. Tap a moment to see its source and confidence. Make privacy trimming part of sharing.
Show nearby altimeter settings, reported wind and field conditions with age and distance. Add vetted field elevation and terrain only after vertical references are normalized. Keep station observations distinct from conditions at the phone.
Guide rigid mounting, initial-attitude setup and calibration; record all three gyro axes and supporting sensors; correct bias and drift with tested fusion. Validate against independent reference data during development, then provide an estimated maximum bank with confidence and moved-phone rejection. No aircraft connection is required in use.
Pair GPS ground velocity with suitable winds aloft, preserve forecast uncertainty and restrict the estimate to validated flight conditions. QNH is useful pressure context, not a TAS solution. Keep the phone-only scope: no IAS or aircraft telemetry. Build distance, time and airport progress around quality-checked data, with maximum bank available as contextual flight review.
Reviewed 25 October 2026. Platform documentation describes available APIs, not a guarantee of sensor hardware, aviation accuracy or Zephyr implementation. Derived metrics, feature priorities and validation recommendations are engineering judgments. Zephyr is being built for flight recording and post-flight review; the data preview is not a flight instrument.