Overview¶
Observers are components that analyze video fields and extract metadata, quality metrics, and encoded information from the video signal. They populate the observation context with their findings, which can then be used by other pipeline stages or displayed to the user.
Each observer specializes in extracting specific types of information from the video signal, ranging from VBI (Vertical Blanking Interval) data to quality measurements.
Observers are not stages and do not appear in the graph. They run on the output of whichever stage is being previewed, in the background, and their results are kept so that revisiting a frame is immediate. In the GUI the results appear in the Preview window's Observers menu — the VBI Decoder, NTSC Observer, Video Parameters and Closed Captions windows — and are also used by stage tools such as the Disc Mapper and by the analysis sinks. While the background work is running, the main window's status bar shows Computing observations… N% (frames are being analysed) or Checking observations… N% (results already stored are being confirmed), and Observations paused during playback… N% while the preview is playing. See Preview for the windows themselves.
Each observer only runs where it applies: the NTSC-specific observers (FM code, white flag) and the closed caption observer are skipped on systems they do not apply to.
Biphase Observer¶
Purpose: Extracts biphase-encoded (Manchester-encoded) VBI data from LaserDisc video signals.
What it extracts: - CAV (Constant Angular Velocity) picture numbers - Chapter numbers - CLV (Constant Linear Velocity) timecodes (hours, minutes, seconds, picture) - Lead-in and lead-out codes - Picture stop codes - User codes - Programme status information (CX audio, disc size, side number, teletext, digital video, sound mode, FM multiplex, programme dump, parity) - Amendment 2 status (copy permissions, video standard, sound mode)
How it works: The observer reads VBI data from lines 16, 17, and 18 of the video field. It decodes the Manchester-encoded biphase data and interprets it according to the IEC 60856/60857 LaserDisc standards. The decoded data is validated using parity checks and BCD (Binary Coded Decimal) decoding.
What it looks like: The extracted VBI data appears as metadata fields in the observation context, providing frame-accurate information about disc position, chapter markers, and disc configuration. The VBI Decoder window (Preview window, Observers menu, Ctrl+Shift+V) shows it for the current frame: the raw words from lines 16–18, picture number, CLV timecode, chapter number, user code, the picture-stop, lead-in and lead-out codes, and the programme status (original specification and Amendment 2 on separate tabs).
Black PSNR Observer¶
Purpose: Measures the Peak Signal-to-Noise Ratio from VITS (Vertical Interval Test Signals) black level reference signals.
What it measures: - Black level PSNR in decibels (dB)
How it works: The observer analyzes specific regions of the VBI that contain black level test signals. It extracts sample data from these regions and calculates the PSNR by measuring the noise variance relative to the expected black level. The measurement helps assess the quality of the black level region of the video signal.
What it looks like: A single numerical value in dB representing the signal quality in the black regions, shown per field in the Video Parameters window and charted over a whole recording by the SNR Analysis Sink. Higher values indicate cleaner signals with less noise.
White SNR Observer¶
Purpose: Measures Signal-to-Noise Ratio from VITS white flag test signals.
What it measures: - White flag SNR in decibels (dB)
How it works: The observer analyzes white flag test signals in the VBI to measure noise in the white level region of the video signal. It extracts samples from specific regions, uses the mean of the data as the signal reference, and calculates the SNR by comparing signal strength to noise variance. This measurement complements the Black PSNR measurement to provide a complete picture of signal quality across the luminance range.
What it looks like: A numerical value in dB representing the signal-to-noise ratio in white regions, shown alongside Black PSNR. Higher values indicate cleaner signals with less noise at white levels.
Burst Level Observer¶
Purpose: Analyzes the color burst signal amplitude.
What it measures: - Median color burst peak amplitude in 10-bit sample units (CVBS_U10_4FSC domain)
How it works: The observer examines the color burst signal (the reference signal for color decoding located after the horizontal sync pulse) on three sample lines in each field of the frame. For each line it computes the RMS of the burst window and converts it to a peak amplitude (RMS × √2 for a sinusoidal burst), rejecting outliers, then records the median across the sampled lines. The result is an AC amplitude — the peak excursion of the burst from its mean, not an absolute signal level. The burst level is important for color reproduction quality and can indicate signal degradation or processing artifacts.
What it looks like: A numerical value stored in 10-bit sample units. The GUI (for example the Video Parameters window and the Burst Level Analysis chart) converts it for display into the project's amplitude unit — 10-bit samples, IRE, or mV. A standard-amplitude burst corresponds to a peak excursion of roughly 20 IRE: the burst swings ±20 IRE around blanking (40 IRE peak-to-peak for NTSC; 300 mV peak-to-peak for PAL).
Closed Caption Observer¶
Purpose: Extracts EIA-608 closed caption data from the VBI.
What it extracts: - Closed caption presence indicator - First EIA-608 data byte (7-bit + parity) - Second EIA-608 data byte (7-bit + parity) - Parity validation for both bytes
How it works: The observer runs on 525-line systems only (NTSC and PAL-M). It reads line 21 of the first field, which carries the CC1/CC2 caption services. It decodes the NRZ (Non-Return-to-Zero) encoded caption data by detecting the start bits (00 followed by 1) and then reading 7 data bits plus 1 parity bit for each of the two caption bytes. The parity is validated to ensure data integrity.
What it looks like: Two bytes of caption data per frame, which can be decoded into text characters and control codes by caption decoders.
The Closed Captions dialog (preview window, Observers menu) is where those bytes become readable. It is offered on NTSC projects only, and is greyed out on any other standard. EIA-608 carries two bytes per frame and builds a caption out of them over a second or more, so what is on screen at any one frame is the result of a run of frames before it, not of that frame's own data; the dialog therefore decodes a trailing window of frames ending at the current frame.
What it shows is a transcript of the captions decoded so far, each with the frame it appeared at, so a caption can be found again in the recording. The transcript follows the previewer: the caption that would be on screen at the preview frame is shown in bold and scrolled to as that frame moves, which is what tells you the captions are keeping step with the picture. It keeps accumulating as you step or play forward; jumping to an unrelated position starts it again from the frames preceding the new position.
A Service selector at the top of the dialog chooses which of Line 21's four services to decode. The two bytes a field carries are shared by CC1 and CC2 (captions) and TEXT1 and TEXT2 (pages of text such as schedules or scores), and which one a byte pair belongs to is decided by the control codes before it — so they can only be read one at a time. CC1 is what a captioned programme uses and is the default; the others are worth a look when a recording turns out to carry more than one. Changing the selection starts the decode again from the frames preceding the current position, because nothing decoded under the old selection describes the new one.
The status bar reports the caption mode the service is using — Pop-on (captions prepared off screen and shown complete), Roll-up (lines scrolling up as they arrive), Paint-on (text appearing character by character) or Text service (a page of text that scrolls) — and the byte pair recovered from the current frame, with the odd parity bit restored so it reads as it would in an SCC file. A byte whose parity check failed is marked with a question mark; a pair with neither byte intact is not fed to the caption decoder, which is the same rule the Closed Caption Sink stage applies when exporting. Most frames of a captioned recording carry no caption data at all — the caption data line is only written to while there is something to say — and the readout says so rather than leaving a gap.
The Closed Caption Sink stage exports the same data, one service at a time, as a Scenarist SCC, SubRip, plain text or HTML file.
Colour Frame Phase Observer¶
Purpose: Determines each field's position within the repeating colour subcarrier sequence.
What it measures: - Per-field colour-sequence phase ID (PAL/PAL-M: 1–8, NTSC: 1–4) - Colour-frame sequence index derived from the phase ID (PAL/PAL-M: 1–4, NTSC: 0–1)
How it works: The observer measures the colour burst phase of both fields in a frame and classifies each field's position within the colour subcarrier sequence: an 8-field sequence for PAL and PAL-M (EBU Tech. 3280-E, ITU-R BT.1700-1) or a 4-field sequence for NTSC (SMPTE 244M). Each field's phase ID is then reduced to a frame-level colour-frame index. A value of -1 is recorded when the burst is absent or too weak to classify (for example blank tape or a pre-programme leader).
What it looks like: Two integer values per field in the observation context. The Video Parameters window (Preview window, Observers menu) shows the colour frame index for the currently previewed field or frame. Colour-frame information matters wherever colour sequence continuity does — for example validating source alignment or diagnosing chroma decoding issues.
Field Quality Observer¶
Purpose: Calculates overall quality metrics for each video field.
What it measures: - Field quality score (0.0-1.0) - Dropout count - Phase correctness indicator
How it works: The observer combines multiple quality indicators to produce an overall quality score: - Dropout density: Counts the number of dropouts detected in the field - Phase correctness: Validates that the field's phase information is correct - Signal metrics: Incorporates SNR and other signal quality measurements when available
The quality score is used by the disc mapping policy to select the best duplicate when multiple fields have the same VBI frame number (common when stacking multiple captures).
What it looks like: A normalized quality score from 0.0 (worst) to 1.0 (best), along with individual metrics; the Video Parameters window shows the score and dropout count for each field. Higher scores indicate better-quality fields with fewer dropouts and better signal characteristics.
FM Code Observer¶
Purpose: Extracts FM code from NTSC LaserDisc signals.
What it extracts: - FM code presence indicator - 20-bit FM code data payload - Field indicator flag
How it works: The observer reads line 10 of NTSC video fields, where FM codes are encoded. It decodes the frequency-modulated signal to extract a 20-bit data value and a field flag bit. FM codes can carry various types of metadata specific to certain LaserDisc formats.
What it looks like: A 20-bit data value and a boolean field flag when present. Not all LaserDiscs contain FM codes, so the presence indicator shows whether valid FM code data was decoded. The NTSC Observer window (Preview window, Observers menu, NTSC projects only) shows it per field.
White Flag Observer¶
Purpose: Detects the presence of white flag signals on NTSC LaserDisc video.
What it detects: - White flag presence indicator
How it works: The observer examines line 11 of NTSC video fields for the white flag signal. The white flag is a reference signal that can be used for various test and measurement purposes. It indicates the presence of this test signal for quality analysis.
What it looks like: A simple boolean indicator showing whether a white flag was detected on the current field, shown per field in the NTSC Observer window.