sections[].dropouts

Per-section dropout simulation. Optional — omit the block entirely to disable dropouts.

    dropouts:
      random:
        scale: 5
        seed: 42
      scratch:
        scale: 3
        seed: 42

Keys

Key	Type	Required	Description
random	map	No	Poisson random dropouts (below)
scratch	map	No	Persistent radial scratch defects (below)

At least one must be present and non-zero. No other key is accepted.

random

Surface and media degradation, modelled as a Poisson process — short losses scattered across the picture, independent frame to frame.

Key	Type	Required	Range	Description
scale	int	No	0–20	Severity, default 0. 0 disables, equivalent to omitting the block
seed	int	No (required with anchor: disc)	any int64	Fixed RNG seed. Omit for a run-specific seed
flagged	bool	No	true, false	Default true. false applies the dropouts but leaves them out of the sidecar; see flagged
anchor	string	No	section, disc	Default section. What the dropouts are keyed to; see anchor

scratch

A persistent physical defect: it appears, grows to a peak severity and fades away over a run of frames — a triangular amplitude envelope — so it spans many consecutive frames rather than flickering.

Key	Type	Required	Range	Description
scale	int	No	0–20	Severity, default 0. 0 disables
seed	int	No (required with anchor: disc)	any int64	Fixed RNG seed
flagged	bool	No	true, false	Default true. As for random
anchor	string	No	section, disc	Default section. As for random

flagged

A real player flags most dropouts, but not all: a defect the RF detector misses still damages the picture. flagged: false models those.

• The dropouts are applied to the signal exactly as with flagged: true. The flag changes no random draw, so the same project with flagged: true reproduces the same events, and their rows.
• No sidecar row is written for them, so the sidecar holds no record of them at all.
• The sidecar is still created whenever any section has dropouts, so the set of output files does not change when the flag is toggled.
• random and scratch are flagged independently. In a section with one of each flagged, the sidecar records only that type's runs.

anchor

anchor sets whether the dropouts belong to one capture or to the disc.

Value	Keyed by	Models
section (default)	The section and the frame's output position	Dropouts of one playback: a different section, or a later start, draws different ones
disc	The disc frame (picture number − 1 on CAV) and the seed	A defect of the pressing: every capture with the same seed carries it on the same picture, whichever section and output position the picture lands on

With anchor: disc:

• Random dropouts on a frame depend only on seed and the frame's disc position, so a capture that starts later, skips or replays a range shows the same dropouts on the same pictures.
• Scratch events depend on seed and the disc frame the section starts on (its picture_number start_value − 1 on CAV), and each event's lifespan is counted from that frame. A replay section with the same start_value therefore repeats the first play's scratches, while a section starting elsewhere on the disc draws its own.
• seed is required, since every capture must draw the same events.

On a disc without picture numbers, the disc frame is the output position plus the project's first-picture offset.

Severity scale

scale is a severity level, not a physical unit. The mapping is exponential in t = (scale − 1) / 19:

Parameter	Formula	Scale 1	Scale 10	Scale 20
Random events per frame (Poisson mean)	0.5 × 400^t	0.5	~8.5	200
Random minimum duration (samples)	5 × 10^t	5	~15	50
Random maximum duration (samples)	50 × 8^t	50	~134	400
Scratch count	2 × 20^t, at least 2	2	8	40
Scratch maximum lifespan (frames)	2 × 250^t	2	~27	500
Scratch maximum peak width (samples)	5 × 400^t	5	~85	2000

Values are rounded to whole numbers except the event rate. A random dropout's duration is drawn uniformly between its minimum and maximum.

Rules

Rule	Severity
scale outside 0–20	Error
anchor: disc without a seed	Error
anchor other than section or disc, or flagged not a boolean	Parse error
A dropouts: block that enables nothing — dropouts: {}, or both scales 0 or absent	Error — omit the block instead
Derived maximum scratch lifespan longer than an integer duration_frames	Warning: the scratch envelope is truncated

Dropout validation error in section 'Programme': dropouts block has no active
type (both random.scale and scratch.scale are 0). Remove the dropouts block or
set at least one scale to 1–20.

The "disable by omission, not by zeroing" rule exists so that a block that looks like it configures something always does.

The dropout sidecar

When any section has dropouts enabled, the pipeline writes a companion SQLite database:

<basename>.dropouts.meta      schema version 5

with a single table:

dropout_run(cvbs_file_id, frame_id, sample_start, sample_count, severity)

severity is 25 for runs entirely within the non-visible VBI and 75 for runs intersecting the active picture.

This is ground truth for every flagged dropout videosynth introduced, with its exact extent, so a dropout detector or concealer can be scored directly against it. Dropouts with flagged: false are absent from it.