The IGC flight-recorder file, decoded.

One letter per line

The first character of every line is the record type — a single letter A–L. The rest of the line is fixed-width and/or delimited data whose layout depends on that letter. A recorder writes the file once, in order:

A  recorder identification          (one, first line)
H  header metadata                  (pilot, glider, date, hardware)
I  B-record extension schema        (extra fix fields)
J  K-record extension schema        (optional)
C  task declaration                 (optional, pre-flight)
F  satellite constellation          (at start, then on change)
B  position fix  ── repeated ──      the track / barograph
E  events    K  periodic data       (interleaved with B)
G  security signature               (last; optional on soft-loggers)

The fix, byte by byte

A B record is a 35-byte fixed base, optionally extended by fields the I record declares. Columns are 1-indexed from the B. Hover a field:

Paste IGC, read it back in English

Edit the file on the left. Every line is parsed live — the record letter is colour-coded, B-record coordinates are converted to decimal degrees, and the barograph redraws from the fixes. Nothing leaves the page.

Degrees, minutes, thousandths

Latitude and longitude are not decimal degrees. They are degrees + minutes + thousandths of a minute, WGS84, with a hemisphere letter. Longitude degrees take three digits, latitude two.

Altitude — two of them, both in metres

Pressure altitude (cols 26–30) is on the ISA 1013.25 hPa datum — not the pilot's QNH. GNSS altitude (cols 31–35) is the WGS84-ellipsoid height. When fix validity is V (2D), trust pressure altitude and treat the GNSS value as unreliable.

Every record type

Filter by name or letter, or narrow by category. Colour encodes what a record does.

The H record

One line per field: H, a source letter (F recorder · O observer · P pilot), a 3-letter subtype, then the data.

The I record & three-letter codes

The I record declares the extra fields appended to every B record: a count, then for each a start byte, end byte, and a three-letter code (TLC). J/K do the same for periodic data.

I 02 3638 FXA 3940 SIU
   │  │    │    │    └ bytes 39–40 → satellites in use
   │  │    │    └ code
   │  │    └ bytes 36–38 → fix accuracy (m)
   │  └ start/end byte
   └ number of extensions

The C record

A pre-flight declared task: one header line (declaration date/time, flight date, task id, turnpoint count, description), then one line per point reusing the B-record coordinate format.

C230726210000230726000102Demo Triangle
C5052000N00412000ETakeoff
C5100000N00500000ETP1
C5052000N00412000ELanding

Point count = takeoff + start + turnpoints + finish + landing. The coordinates are identical in format to a B-record fix, so the same decoder handles both.

Where parsers go wrong

• Read dates day-first. HFDTE is DDMMYY, never ISO — and accept both HFDTEDATE:DDMMYY,NN and legacy HFDTEDDMMYY.
• Minutes are thousandths. MMmmm means MM.mmm minutes, not decimal degrees. Apply LAD/LOD/TDS when the I record declares them.
• Longitude is three degree-digits (DDD), latitude two. Off-by-one shifts every fix.
• Trust A over V. When validity is V, the GNSS altitude is unreliable — prefer pressure altitude or drop the fix.
• Pressure altitude is on 1013.25 hPa, in metres — not the pilot's QNH, not feet.
• Slice by column, not whitespace. B records have none; extensions are byte-positional per the I record.
• G may be missing. Absence means "unvalidatable," not "invalid" — soft-loggers routinely omit it.
• Times wrap at midnight. Detect a decreasing HHMMSS and roll the date forward.

What this reconciles