AUD · Audio tools

LTC Timecode Generator and Reader

First 16 bits of 01:00:00:008.0 ms

01:00:00:00 to 01:00:59:24 · 1,500 frames

Shareable link

Settings are written to the URL as you change them. Nothing differs from the defaults yet.

Oxyd To put the result on a track with other audio, Oxyd is a multitrack editor with the same effects. Open Oxyd

Write linear timecode as audio, or read it back out of a recording. LTC is the form of SMPTE timecode that travels as an audio signal: 80 bits per frame, one frame of code for every frame of picture, on a spare track or a cable between machines.

Generate a timecode WAV

Pick the frame rate, type the start address as HH:MM:SS:FF and set the duration. The caption under the scope shows the first and last address and the frame count. Download writes a mono WAV at 44.1, 48 or 96 kHz in 16, 24 or 32-bit PCM.

Level is the peak relative to full scale. -6 dB suits most line inputs. Readers work from the timing of the edges and not from the level, so lower it for a camera’s microphone input, where a hot square wave bleeds into the neighbouring track.

User bits are 32 spare bits in every frame, entered as 8 hex digits. Productions put a date, a reel number or a camera letter there. They repeat unchanged in every frame of the file.

Frame rates

Setting Frames counted per second Real speed
23.976 24 24000/1001 fps
24 24 24 fps
25 25 25 fps
29.97 drop-frame 30, with skipped numbers 30000/1001 fps
29.97 non-drop 30 30000/1001 fps
30 30 30 fps

At 29.97 fps and 48 kHz a frame is 1601.6 samples long, which is not a whole number. The generator places every edge from the sample index, so the fraction never accumulates: an hour of code is exactly an hour of frames.

Drop-frame timecode

NTSC video runs at 30000/1001 frames a second. Counting 30 frames to the second makes the code run slow against a wall clock by 108 frames an hour, which is 3.6 seconds.

Drop-frame code skips frame numbers 00 and 01 at the start of each minute, except minutes 00, 10, 20, 30, 40 and 50. That removes 108 numbers an hour. 00:00:59;29 is followed by 00:01:00;02, and 00:09:59;29 by 00:10:00;00. The semicolon before the frames is the convention for marking drop-frame.

A start address that drop-frame skips, such as 00:01:00;00, is moved to the next one that exists and the hint under the field says so.

Inside one frame

Bits 0 to 63 carry the frames, seconds, minutes and hours as binary-coded decimal digits, least significant bit first, with a 4-bit user group after each digit. Bit 10 is the drop-frame flag and bit 11 the colour-frame flag, which this generator leaves clear. Bits 64 to 79 are the sync word 0011 1111 1111 1101, or 0x3FFD.

The bits are sent as biphase mark: the signal flips at every bit boundary and flips once more in the middle of a 1. That keeps the signal free of DC and makes it readable at any polarity. The polarity correction bit (bit 27, or bit 59 at 25 fps) is set so each frame holds an even number of ones, which makes every frame start on the same edge direction.

SMPTE 12M asks for edges with a controlled rise time instead of a hard square wave. Each edge here is a raised cosine with a 40 µs rise, centred on the bit boundary.

Read LTC from a recording

Open a file. Timecode sits on one channel of a stereo or multitrack recording, so each channel is read separately and the one with the most frames is reported. Set the channel yourself when two channels both carry code. Mixing a timecode channel with programme audio destroys it, so the mixed option is only for dual-mono files.

The report lists:

  • the first and last timecode read, and the frame count
  • the frame rate from the highest frame number seen, and the rate measured against the file’s sample rate
  • the drop-frame flag and the user bits, printed in the digit order ltcdump uses
  • whether the code runs forward, in reverse, or both

Jumps, repeats and dropouts

Every frame is checked against the one before it. Each break in the sequence becomes a row with its time and sample position in the file:

  • Jump: the address does not follow the previous one. An edit across two takes, or a recorder that was stopped and restarted on time-of-day code, looks like this. The row gives the size of the jump in frames.
  • Repeat: the same address twice in a row.
  • Dropout: some frames could not be read, and the count on the far side is what it would have been had the code kept running. The signal was damaged or muted; the timecode underneath is intact.
  • Direction change: the code turns from forward to reverse or back.

Copy the table or download it as CSV with the first and last frame included. A file with no rows between first and last is continuous.

Clock recovery

The reader times each zero crossing to a fraction of a sample. Every gap between crossings is either a whole bit or half of one, and the reader learns the bit length from the signal: 40 gaps in a row that fall into two groups, one twice the other. From then on the bit length follows the signal, so a transfer that runs 4 % fast or a tape with some wow still reads.

Nothing depends on level or polarity. A fader move, an inverted cable and a DC offset all leave the crossings where they were.

Limits

The longest file is one hour. A playback device sets the real speed of a generated file, and consumer sound cards drift by tens of parts per million, so the WAV does not replace a hardware generator for long-form sync.

Heavy lossy compression, clipping into a limiter, and code mixed with other audio all cost frames, and they show up as dropouts in the table.

Frequently Asked Questions

Generate LTC writes a mono WAV of SMPTE linear timecode from a start address, a frame rate and a duration. Read LTC takes a recording, or the microphone or line input, and reports the first and last timecode, the frame rate, the drop-frame flag, the user bits, the frame count and every discontinuity.

WAV, FLAC, AIFF, MP3, M4A, Ogg and the audio track of a video file. Each channel is read on its own and the one with the most frames is reported, or you can name the channel. Lossless files read best; a low-bitrate MP3 can smear the edges enough to lose frames.

Both run at 29.97 frames a second. Non-drop counts 30 frames to every second of code, so the code falls 3.6 seconds behind the clock each hour. Drop-frame skips frame numbers 00 and 01 at the start of every minute except each tenth minute, which keeps the code within about 3 frames of the clock over a day. No pictures are dropped, only numbers.

Only when the recording runs at its true speed. Those pairs carry identical frame numbers and differ by 0.1 % in speed, so the reader measures the frame length against the file's sample rate. If the measured rate is more than 0.03 % from both, it names the pair and shows the measured figure. A set drop-frame flag settles it as 29.97.

No. The file is exact to the sample, but a player's clock is not, and a sound card that runs 20 parts per million fast gains about 2 frames an hour. Use the file to stripe a track, feed a slate or test a reader. For locking equipment over a long session, use a hardware generator.

No. Listening shows the code as it arrives and logs discontinuities. It does not output timecode, regenerate it, or control other equipment.

Yes. The sync word reads differently from each end, so a reversed file or a tape played in reverse is read, counted downwards and marked as reverse. A change of direction partway through is listed as an event.

Explore Our Tools

Browse all tools