Effects

Every control in the app, by where it sits on the signal path, with the fault each one models. Generated from the table the panel renders.

Features is the tour of what each stage does. Start there.

The app shows the same text under each control’s ?, and ctrl+k searches it.

Sources#

What either deck can be fed:

Generated — switches instantly: colour bars, a frequency sweep, broadcast snow, blank-tape noise, a video synth
Bundled with the app: a still photo, a test pattern and two public-domain cartoon excerpts
Public archives — fetched live, and never the same twice: a random file from Wikimedia Commons or archive.org, or a search across both
Your own — opens a picker: an image or video file, a typed text card, a video URL, your own clip shelf
Live — asks the browser: a webcam or USB capture device, or a shared window or tab

Source A#

Signal (source A)#

Control	Range	The fault it models
invert (polarity swap)	0–1	The video pair wired backwards, negating the whole composite waveform. Hue inverts too: the subcarrier is on the same wire.
deinterlace	`off` · `on`	Rebuilds each frame from a single field, like a bob deinterlacer: fixes comb teeth at half the vertical detail.
capture luma band (0 off)	0–4.2 MHz	Luma bandwidth the deck passed to the capture card, so the file arrives already soft. VHS manages about 3 MHz at SP.
capture chroma band (0 off)	0–1.5 MHz	Chroma bandwidth from the same deck. Colour-under carries about 0.5 MHz against luma's 3, so colour smears sideways under sharp edges.
capture grain	0–30 IRE	The deck's luma FM noise floor, as fine grain in every frame of the file. It holds still when the deck pauses.
capture chroma noise	0–60 IRE	Noise on the deck's colour-under carrier, which had far less headroom than luma, smeared by the narrow chroma band into blotches of wrong hue.
capture y/c delay	-500–500 ns	The deck's chroma arriving late (+) or early (-) against its luma, displacing colour off its edges by a few hundred nanoseconds on a home deck.
vbi test signals	`off` · `on`	Vertical blanking interval - lines 17-18: VITS multiburst and staircase. - line 19: a VIR reference. - line 21: caption data. Roll the picture or shrink v size to see it in the black bar.

Noise source (static)#

Control	Range	The fault it models
noise bandwidth	0.2–7 MHz	The bandwidth of the path the noise came through, which sets its grain. A tuner's IF stops at 4.2 MHz; a playback head's aperture is tighter.
noise power	0–2	How much noise the detector receives: snow against a black floor on an untuned channel, a swing around mid grey on blank tape.
per-sweep level error	0–1	A gain error lasting one scan line: the AGC hunting, or head contact varying sweep to sweep. It multiplies, so whole lines flicker.
field refresh	1–60 Hz	How often the noise field changes. Below 60 the set draws each field several times and the boil goes chunky.

Video synth (source)#

Control	Range	The fault it models
osc A	0–8000000 Hz	The first oscillator's frequency, read against the raster: one cycle per frame at 60 Hz is a vertical gradient, the 15734 Hz line rate turns it sideways, and a multiple paints standing bars. At 3579545 Hz it lands on the subcarrier and the encoder returns flat colour.
osc B	0–8000000 Hz	The second oscillator, live once the combiner is off "osc A alone". It beats against the first, so a pair a few hertz apart draws a drifting moire.
waveform	`ramp` · `triangle` · `sine` · `pulse`	The waveform for both oscillators. Ramp is a sawtooth, triangle folds it symmetric, sine passes the encoder without ringing, pulse is a comparator output with the most bandwidth to distort.
combiner	`osc A` · `sum` · `ring mod` · `comparator`	How the two oscillators combine: sum is a mixing amplifier driven into its rails, ring mod a balanced multiply leaving only sum and difference, comparator puts B on a slicer's reference.
level	0–4 x	Output contrast around mid-video, ahead of the colorizer. Past 1 the waveform runs into its rails and squares off.
colorizer	0–1	One signal into three guns through three phase shifts 120 degrees apart, which is all a colorizer ever was.
colorizer phase	0–360 deg	Rotates all three phase shifts together, sliding the palette around the wheel while the geometry holds still.
colorizer input filter	0–24 px	The lowpass ahead of a colorizer's slicers. A sharp input crosses threshold on every detail and posterizes into confetti; a soft one lays down slabs.
colorizer input	`oscillator` · `picture`	What the colorizer slices: its own oscillator, or the picture. Pointed at the picture it turns brightness into hue, so colour lands in fields of equal tone.
colorizer type	`phase shifts` · `comparators`	How the box turns level into colour. Phase shifts sweeps hue continuously; comparators is the cheap way, three slicers switching each gun fully on or off into the eight corners of the colour cube.
over picture (A)	0–1	Lays the synth over slot A's picture instead of replacing it. Source A only.
luma → osc A	0–200000 Hz	The picture's brightness into oscillator A's frequency input, so the wave runs faster through bright picture and the image draws itself as contour lines.
that input, patched to	`the deck` · `the loop`	Which picture the frequency input reads. The deck redraws the contours every frame; the loop reads the camera's return, so they trace the last generation's and go round again.

Feed A · deck#

Control	Range	The fault it models
A pause (deck held)	0–1	The pause button on input A's deck. The drum re-reads one track so the frame holds, but the capstan servo is defeated: lines scatter sideways and a mistrack stripe of snow creeps through.
A dropouts	0–400 /frame	Dropout events per frame on input A's tape alone. Shed oxide leaves the head reading nothing, with no compensator to hide the gap.
A dropout len	1–60 us	How long each of A's dropouts lasts, in microseconds of the 63.5 µs line.
A sync suppression	0–1	Head-end scrambling on input A alone: A's sync tips are lifted toward blanking before the mix, so the receiver must choose between A's pulses and B's.
A system	`gated` · `alternate` · `ssavi`	Which scrambling system A's channel uses. Gated suppresses every line, alternate every other line, and SSAVI also inverts the active video.

Feed A · cable#

Control	Range	The fault it models
A loose connector	0–1	How loose the plug in input A's jack is. Bands of lines lose contact, re-rolled every frame.
A bad contact	`pin` · `shield` · `both`	Which of A's two contacts is intermittent. The centre pin breaks the signal path, so bad bands collapse to noise and lose A's sync; the shell breaks the ground, so hum lands on them instead. Both is a wiggled plug.
A ground loop	-40–40 IRE	A ground loop on input A's cable alone, this deck's outlet against the mixer's. It lifts A's sync tips with the picture, so the receiver's hold chases the bar.
A noise	0–150 IRE	Snow on A's feed alone, from a long antenna run or a bad patch cable ahead of the mixer.
A termination (-1 daisy, +1 open)	-1–1	Termination fault on A's cable alone: negative is double-terminated and arrives dim, positive is unterminated and runs hot and ringing.
A polarity (flips sync)	0–1	A signal/ground swap on A's own connector, negating A's waveform and its sync before the mixer. B is untouched, so the receiver may latch onto B.

Mix#

A/B Mixer#

Control	Range	The fault it models
genlock	`dirty sum` · `clean dissolve`	Whether source B is genlocked to house reference. Off, B free-runs and is summed into the composite — a wiring fault — so its detune, roll and skew produce fighting sync and chroma beats. On, B is re-encoded on A's raster and the combine is a clean switcher dissolve.
A gain	-3–3 x	A's level on the summing bus, on the dirty sum only: the clean dissolve sets A to (1 − B gain). Negative inverts A into a difference key that cancels against B.
B gain	-3–3 x	How much of source B reaches the composite line. With genlock off it is the level B is summed in at, and negative inverts B's whole signal, sync tips included. With genlock on it is the crossfade fader.
ring mod	0–1	Multiplies the two composite signals instead of adding them. Two subcarriers multiplied land at their sum and difference, so colour arrives that neither source carries.
bus overload	0–1	How little headroom the summing amplifier has. At 0 the sum is pure arithmetic. Open it and the stage runs out of volts: gain falls away toward the rail, and falling gain multiplies the two signals sharing the bus, so a detuned B beats against A into the chroma band.
line offset	-60–60 Hz	How far B's line rate sits from A's. The two horizontal oscillators are unlocked, so B skews a little more on each line and slides sideways.
sc detune	-3000–3000 Hz	How far B's colour subcarrier sits from A's 3.579545 MHz. The decoder locks to A's burst, so B's hue cycles continuously: the rainbow crawl of a non-genlocked source.
frame roll	-30–30 l/f	B's vertical drift in lines per frame, from its field rate not matching A's.

Character generator (chyron)#

Control	Range	The fault it models
cg over program	0–1	A character generator at the switcher, keying caption text into the picture: the box every lower third and station ident came out of. A CG puts out two wires: a fill, which is video, and a key, which is a matte cut at the characters' own edges. Everything below changes the relationship between them.
key timing	-600–600 ns	The trim every keyer has, because the key path and the video path are different lengths of circuit. Mis-set on a glyph it puts background through one side of every stem and a shadow down the other. One sample is 70 ns.
key clip	0–1	Where the slicer cuts the processed key. On type it is stroke weight: down, thin strokes fuse; up, stems drop out of words.
key bandwidth	0.3–8 MHz	The key-processing amplifier ahead of the slicer, narrower than the video path, which is the only reason a key has a soft edge. Horizontal only: there is no vertical neighbour on the wire.
key invert	`normal` · `inverted`	Which side of the key is cut. Inverted, the box fills the raster and the letters are holes in it.
edge offset x	-24–24 smp	A CG drew its border and drop shadow by delaying the key a sample and a line and OR-ing it under the fill. This is that delay; far out, the shadow detaches from the type.
edge offset y	-24–24 ln	The other half of the drop shadow, in lines.
fill level	0–120 IRE	How bright the characters are laid in, in IRE on the composite. 100 is peak white; past that the box overmodulates and everything downstream reacts.
cg x	0–1	The block's left edge across the picture.
cg y	0–1	The block's top edge down the picture. The stock value is a lower third, clear of the caption decoder's block.
cg size	1–6	Picture samples per font dot. The glyphs are dots on a grid, so the type stays as blocky as the ROM made it.
cg rom address line	0–11	A pin held high on this box's font ROM. Low lines carry the row inside the cell, so every glyph grows a seam; high lines carry the character code, so the whole font is substituted.
cg rom data line	-8–8	The data bus of the same chip: eight dots across one row, so holding one stripes a column down every character.
cg rom crossed lines	0–10	Two adjacent address lines of this box's font ROM transposed. Low in the bus it shuffles the scan lines inside every cell; high in the bus it permutes the font in blocks.
cg rom counter slip	-4–4 /frame	The vertical reset failing on this box's character-address counter, so the whole page is out by the same count and further out every frame. Twelve counts is one whole character.
cg rom slip per line	-0.25–0.25 /line	The horizontal reset failing on the same counter. The count grows on every scan line and the vertical still clears it, so each row comes off a different part of the font.
cg rom cell strap	-11–12 rows	The cell-height jumper in the wrong hole. A glyph's first row is at its code times the cell height and the raster keeps stepping 12 rows whatever the strap says, so the text shears into a diagonal slice of the font.
cg rom bit rot	-1–1	Charge leaked off this box's array. Decayed cells read back as the erased state: positive erases to a lit dot and the letters thicken, negative to a dark one and they crumble. The pattern is in the die.
cg page address line	0–7	A line held on the counter that walks this box's page memory instead of its font. Low lines are the column, high lines the row, so characters keep their shapes and lose their places.
cg page counter slip	-4–4 /frame	The same slip on the page-memory counter, so the block walks diagonally through itself a cell at a time.

Wipe (A/B)#

Control	Range	The fault it models
pattern	`off` · `h` · `v` · `box` · `diamond`	Which switcher wipe pattern decides where B shows. It shapes the picture only — on the dirty path B's sync and burst keep summing across the whole raster.
position	0–1	The wipe lever: where the A/B boundary sits, 0 full A to 1 full B.
softness	0–0.5	Width of the blended border along the wipe edge. 0 is a hard switcher cut.
sweep	0–2 Hz	Drives the wipe lever back and forth at this rate. Locks to MIDI clock with the ♩ icon.

PiP inset (source B)#

Control	Range	The fault it models
inset key	0–1	Squeezes source B into a positionable window over the program, like a switcher DVE. The inset is re-encoded genlocked to the house raster, so it dot-crawls and holds still.
center x	0–1	Horizontal centre of the inset window across the active picture.
center y	0–1	Vertical centre of the inset window down the active picture.
width	0.1–1	Width of the inset window, as a fraction of the active picture.
height	0.1–1	Height of the inset window, as a fraction of the active picture.
border	0–0.03	Thickness of the matte border a switcher draws around a squeezed source.
edge soft	0–0.05	Softness of the inset window edge.
luma key (- inverts)	-1–1	Keys the inset against its own brightness, so a subject drops in without its rectangle. Positive keeps the bright parts of B, negative the dark.
key level	0–1	The brightness the inset key slices at, 0 black to 1 white.
key soft	0.01–0.4	Width of the inset key transition. Narrow cuts a hard matte.

Chroma key (A through B)#

Control	Range	The fault it models
key (- inverts)	-1–1	A chroma keyer across the mixer with B as the foreground: B's backing colour is cut away so A shows through. It slices the chroma the encoder made, so the matte comes out soft across and sharp down, the lopsided edge every composite key had.
backing hue	0–360 deg	Which chroma phase counts as the backing. 241 is a pure green screen, 347 a blue one. These are angles on the wheel the subcarrier carries, so anything sharing a hue with the backing goes too.
acceptance	0–180 deg	How wide a wedge of hue either side of the backing counts as backing. Narrow leaves shadows and folds opaque; wide starts eating skin on a warm-lit subject.
clip	0–0.3	The saturation a sample must reach before the keyer acts on its hue. A demodulator given an unsaturated sample reports an arbitrary phase, so without this greys key out at random.
gain (edge)	0–0.4	How fast the keyer swings between keep and cut, in hue and saturation at once: the gain knob on the front of the box. At 0 the comparator snaps and the composite edge shows as steps.
spill kill	0–1	Cancels the backing colour reflecting off the subject. Luma and chroma are the same wire, so the box reinjects the backing's own subcarrier in antiphase. It nulls fully on the genlocked path; on the dirty path B's carrier drifts, so the cancellation runs late and leaves a residue.
key delay	-1.5–1.5 us	The registration trim, because the key path and the video path are different lengths of circuit. Off zero the matte lies beside the subject: one edge keeps a rim of backing colour, the other removes a rim of subject.
fill	`program A` · `matte` · `loop bus`	What shows through the hole the key cut. Program A is the other input. Matte is the box's own generator, a flat colour encoded on the house carrier, so it dot-crawls like any other colour. Loop bus patches the mixer's last frame in, so feedback regenerates only inside the keyed shape. Genlocked path only.
matte level	0–1	How bright the matte generator sits. This is the luma of a real encoded line, so peak white with saturation up puts the sum past 100 IRE.
matte hue	0–360 deg	The matte colour, as a phase on the subcarrier, off the same wheel as the backing hue. Set near the backing hue it lands inside the acceptance wedge, so a loop keys its own fill away.
matte saturation	0–0.6	How much chroma the matte generator puts on the carrier. At 0 there is no subcarrier at all, which is how to get a black or white fill.

Channel#

Recording (luma & FM)#

Control	Range	The fault it models
luma bandwidth	0.3–6 MHz	How much brightness detail the recording or channel passes. Broadcast is about 4.2 MHz, VHS roughly 3, EP less.
peaking	0–12	The sharpness boost VCRs and TVs apply to fake back the detail the bandwidth limit removed. It overshoots every edge into a ringing outline.
differential gain	-0.5–1	The video amplifier's gain is not flat against the brightness it is amplifying, so a subcarrier riding bright picture comes through smaller and saturation drains out of the highlights. Spec sheets list it as DG%.
differential phase	-60–60 deg	The same amplifier's delay moves with brightness, and a delay at 3.58 MHz is a phase shift, so hue swings with the luma underneath it. The burst sits at blanking level where the shift is zero, so the decoder's reference never moves.
FM over-deviation	0–1	A VHS deck records brightness as pre-emphasized FM, and a white-clip circuit is supposed to stop bright edges overshooting the deviation the tape can carry. Set too hot, the overshoot runs past the response cliff and the discriminator folds back: more frequency out as less video. Colour-under passes through untouched, carrying saturated colour over black.
inversion streak	0.1–0.7 us	How long the demodulator takes to recover from a fold: the deemphasis time constant, which smears the inversion rightward.

Noise & interference#

Control	Range	The fault it models
noise	0–150 IRE	Additive noise on the waveform, in IRE: tape grain and RF snow. It lands on the whole signal, so enough of it disturbs sync and the burst.
noise spectrum (RF ↔ FM)	0–1	Where the noise floor comes from, which decides its colour. At 0 it is the RF path, flat across the video band through the tuner's IF. At 1 it is the deck's FM demodulator, where recovering frequency from phase differentiates the noise, so the floor rises into the chroma bandpass near 3.58 MHz and decodes as crawling coloured speckle.
impulse noise (arcs)	0–24 /frame	Impulse interference: ignition, an arcing thermostat, a dying flyback next door. Each event is a run of signal time at carrier-scale amplitude, and its duration sets its shape: tens of microseconds is a ringing streak, hundreds a stepped diagonal, milliseconds a torn slab of hash. The long ones land on sync tips and the beam-load measurement.
ignition train	0–2000 Hz	A periodic impulse source, such as spark plugs or a commutator motor, firing at this rate. Periodic hits against the 15.734 kHz line rate land a fixed step sideways from the last, so the dashes line up in drifting diagonal lattices.
big strikes	0–20 /s	Millisecond-scale events: lightning, an arcing breaker, a compressor starting. Dozens of full lines of dense hash with a DC lift. A strike spans whole lines, so it lands on sync tips and the beam-load measurement and the whole rig reacts.
impulse strength	20–400 IRE	Peak amplitude of each impulse. Real impulses saturate the front end, so the range is large; past 100 IRE every hit drags the AGC.
dimmer lock	0–1	A triac dimmer fires twice per mains cycle at its set angle, so the random hits concentrate into two bands of hash that roll with the hum bar.

Ghosting & leakage#

Control	Range	The fault it models
ghost delay	0–50 us	Multipath: a reflected copy of the broadcast arriving this many microseconds late, as an echo to the right. The reflection carries its own subcarrier, so every 70 ns of delay turns the ghost's hue 90°.
ghost gain	-2–2	Strength of that reflection. Negative means it arrives phase-inverted, so the echo is a dark outline instead of a bright one.
hum	0–120 IRE	Mains hum riding on the video from a ground loop: 60 Hz on the signal, in IRE. It is not quite locked to the field rate, so the bar drifts up the picture.
hum modulation	0–1	The same mains ripple in the supply of an amplifier the signal passes through, so it moves that stage's gain instead of adding to its output. Sync is scaled too, so its depth varies and the receiver's AGC and hold chase the hum. Mostly 120 Hz, from the rectified supply.
sound carrier	0–40 IRE	The 4.5 MHz intercarrier sound leaking past the trap that should remove it, as a fine herringbone over the picture.
sound buzz	0–1	The same leak, as audio out of your speakers. The sound detector recovers the 4.5 MHz beat between the picture and sound carriers, and a limiter that cannot keep the picture off it passes video through as audio: the vertical interval buzzes at 60 Hz, line structure whines, snow hisses. Silent until the Sound stage is switched to buzz out loud.

Dropouts & dubs#

Control	Range	The fault it models
dropouts	0–400 /frame	How many dropout events happen per frame. Shed oxide or a clogged head leaves the head reading nothing for a moment: white streaks, and on a bad one a line the decoder cannot reconstruct.
dropout compensator	`none` · `1-line` · `2-line`	The circuit that patches a dropout from a delay line holding what played a line or two ago. A line of NTSC is 227.5 subcarrier cycles, so one line back arrives exactly out of phase: invisible in brightness and in the complementary hue. Two lines back is a whole number of cycles, so the hue is right and the patch is two lines stale.
dropout len	1–60 us	How long each dropout lasts. A line is 63.5 µs, so 25 µs is a streak across a third of the picture.
dub generations	1–4 x	Runs the whole tape/channel stage this many times over: a copy of a copy. Each generation adds its own noise, dropouts and timebase wander on top of the last.

RF / Tuner#

Control	Range	The fault it models
adjacent channel	0–1	How much of the next channel up the cable gets through the IF trap. The detector turns the neighbour's carriers into beats: their sound carrier lays a 1.5 MHz weave over everything, and their vision carrier's beat is modulated by their raster, so their blanking crosses as slanted dark bars and their vertical interval as the broad sweeping windshield-wiper band.
fine tuning	-1–4 MHz	The fine-tuning knob pulled off channel. Positive moves the 4.5 MHz sound carrier out of its trap, so the detector multiplies it against the video: chroma comes back at 920 kHz as a coarse beat, and 920 kHz detail comes back at 3.58 MHz as rainbow crawl. Negative slides the picture carrier down the IF's Nyquist slope, so detail softens until the colour killer cuts colour.
weak signal (snow)	0–1	IF noise into the envelope detector, which is what weak-signal snow is. The picture rides a negative-modulation carrier, with sync at peak power and white at 12.5%, so the noise lands unevenly: whites boil first, blacks stay quiet longest, and sync goes last.
CB ingress	0–1	A two-way radio getting into the cable through a cracked shield or corroded fitting. The carrier is unrelated to any NTSC frequency, so its beat draws a herringbone at no fixed angle. It arrives in transmissions, and the program audio stands in for the speech.

Cable / Wiring#

Control	Range	The fault it models
hard polarity (flips sync)	0–1	A signal/ground swap at the connector: the whole composite waveform is negated, sync pulses included. The receiver has to find sync in what used to be peak white.
termination (-1 daisy, +1 open)	-1–1	Composite video expects a single 75 Ω load. Negative is double-terminated, a monitor daisy-chained with its loop-through on, which halves the signal. Positive is unterminated, so the line reflects and the signal runs hot and rings.
chroma-pin only	0–1	S-video miswired into a composite input, so only the chroma pin arrives. With no luma and no sync the receiver free-runs on a bare subcarrier.
loose connector	0–1	How loose the plug is. Bands of lines lose contact, re-rolled every frame, the way a plug hanging on its own cable weight makes and breaks.
bad contact	`pin` · `shield` · `both`	Which contact of the plug is intermittent. - pin: the centre breaks the signal path, so those bands collapse to the input stage's noise floor, sync included, which is why they tear. - shield: the shell breaks the ground reference, so hum lands on the bad bands while the picture and its sync survive. - both: a wiggled plug, with the two faults on independent bands.
sync suppression	0–1	How hard the head-end suppresses sync on a premium channel. The scrambler lifts the carrier during each sync pulse, so a set without a decoder box finds a shallow tip or none. Under about half depth the AGC washes the picture out bright; past that the line oscillator free-runs. The broad field pulses are wider than the line-rate gate, so the frame shears instead of tumbling.
system	`gated` · `alternate` · `ssavi`	Which scrambling system. Gated suppresses every line, so the oscillator free-runs all the way down. Alternate suppresses every other line, so the flywheel is pulled back half the time and the drift shows as a ragged line-pair zigzag. SSAVI is Zenith's: suppression plus inversion of the active video, with the burst untouched in the back porch, so hue survives.
agc pulses (macrovision)	0–1	Macrovision's AGC poisoning, stamped on vertical-interval lines 12-19, exactly the window this receiver averages its sync depth over. A pulse parked on the back porch makes the measured depth balloon, so with agc up the set crushes gain on a signal that was never hot.
colorstripe	0–180 deg	The colorstripe half: bursts on moving bands of lines are rotated off the house phase by this much. The decoder corrects each line's hue by the burst it just gated, so the poisoned bands come out rotated the other way and hue banding crawls down the frame.

VHS colour & tracking#

Control	Range	The fault it models
color-under	0–1	VHS cannot record 3.58 MHz colour, so it heterodynes chroma down to 629 kHz, records it under the luma and converts it back on playback. That path collapses colour bandwidth to a fraction of luma's.
chroma noise	0–120 IRE	Noise on the colour-under carrier, before it is converted back up. The 629 kHz carrier gets a fraction of the luma FM's headroom, which is why VHS colour is blotchy while its luma is merely grainy. Needs colour-under raised.
phase jitter	0–180 deg/line	Per-line phase error in that down/up conversion, so hue wanders line to line and the picture picks up a coloured venetian-blind texture. Needs colour-under raised.
Y/C delay	-3360–3360 ns	The chroma path through a deck or proc amp runs its own filters and delay lines, and mistrimmed against the luma path the colour arrives late or early, bleeding off its edges. The burst travels the same path, so hue stays correct: the colour is displaced, not rotated. Steps are whole samples, about 70 ns.
tracking error	0–1	The head is not following the recorded track. It reads partly off-track, so a band of noise appears where the signal is weakest.
band position	0–1	Where that mistracked band sits vertically, 0 top to 1 bottom. With the servo hunting, it is where the servo is trying to sit.
servo hunt	0–1	The deck's auto-tracking servo searching for the track instead of holding it. It reads the RF envelope and steps until the envelope peaks, with less damping the higher this goes, so every correction overshoots and rings. A scene change, exiting shuttle or a thump through the cabinet knocks it off the peak.
servo kick	0–1	How hard each of those events unseats the servo. Needs servo hunt above 0.
head clog	0–1	Oxide packed into the gap of one of the two spinning heads, so that head reads weak or nothing. The heads take turns, one sweep each, so picture and snow alternate at field rate: a hard 30 Hz flicker.
shuttle (1 = play)	-32–32 x	Tape speed as a multiple of play: cue past 1, pause at 0, review negative. Off play speed the head no longer follows one recorded track — each sweep crosses several, the RF nulls at every crossing, and that many noise bars sweep the frame. Each strip between bars carries its own timing and colour-under phase.

Timebase#

Control	Range	The fault it models
flutter	0–4000 ns	Fast timebase error from capstan flutter: each line starts a slightly different moment late, so edges shimmer. Signal-domain, so the burst moves with it and hue wobbles too.
wow	0–10000 ns	Slow timebase error from tape or capstan wow. Flutter shakes line to line; wow drifts over many lines, so whole regions lean and breathe sideways together.
sticky shed	0–15000 ns	Binder hydrolysis making the tape grab the head drum. Tension builds until the patch breaks free, snaps forward and re-sticks: a relaxation oscillator, chaotic rather than periodic, and the mechanism behind squealing tapes.
head switch	-30–30 us	A helical-scan VCR swaps between two heads a few lines before the bottom of the picture, and the two do not agree on timing: the torn hook at the bottom of every VHS frame.
switch noise	0–1	How much noise hash fills the few lines during the head switch, before the servo settles on the new head. Usually hidden under the overscan.

Enhancer (bent)#

Control	Range	The fault it models
clamp gate	-60–600 us	How far the box's DC-restoration gate has slid off the back porch. A clamp pins one sample per line to blanking and the rest of the line rides on that, so dragged into active video the black level bounces line to line with the picture. Negative puts the gate on the sync tip, lifting the whole line.
clamp droop	0–2000 us	Time constant of the coupling capacitor between the gates. Short enough and the level sags back toward blanking within the line, so bright content drags a dark streak to the right edge. Vertical edges are untouched.
detail freq (0 off)	0–5 MHz	Centre of the peaking stage the detail knob drives. A composite box has no Y/C split, so one knob does two jobs: down around 1-2 MHz it rings on picture detail and lays bars behind every edge, up at 3.58 it boosts the subcarrier.
detail regen (0.75+ howls)	0–1	How much of the peaking stage's output the bend feeds back into it. Low is ordinary edge overshoot; approaching 0.75 the ring lasts most of a line. Past it the stage is regenerative: the sync pulse at the head of every line excites it until it hits the rails.
detail boost	0–16 x	How much of the peaking stage is mixed back into the video. With the regen low it is a sharpness control; past unity it is how loud the howl is.
sync regen	0–1	The stabilizer half of the box: a sync separator slices the signal and stamps a clean 4.7 µs pulse at every crossing it finds. This is how much of that pulse train reaches the output.
sync slice	-40–60 IRE	The level the separator calls sync, in IRE. Blanking is 0 and the real tip is -40, so anything under about -10 only finds real pulses. Raise it into picture territory and dark content produces pulses of its own, so the set is given a line rate the image is writing. It slices a lowpassed copy, so only sustained dark areas trip it.

Receiver#

Sync#

Control	Range	The fault it models
horizontal hold	0.02–2	How hard the receiver's horizontal PLL pulls toward each sync pulse it finds. Low is a loose flywheel that ignores noise but drifts and skews; high snaps to every edge including the false ones. Sync-domain, so the burst gate moves with it.
vertical hold	0–1	How much authority the incoming vertical sync has over the receiver's own field oscillator. At 1 the picture locks solid; as it falls the oscillator wins and the frame rolls.
vertical osc (60 = locked)	10–180 Hz	The free-running frequency of the receiver's vertical oscillator. At 60 Hz it agrees with the signal; detuned, the frame rolls at the difference. Only bites once vertical hold is loose.
retrace flag	0–60 us	A kick to the horizontal PLL at the vertical seam, where the equalizing pulses upset it. The first few lines start late and settle back over the next dozen: the flagging top edge.
horizontal osc detune	-3000–3000 Hz	Free-run drift of the receiver's horizontal oscillator away from 15.734 kHz. The PLL keeps pulling it back, so the picture leans into a skew; past the pull-in range it shears into bars.

Paperclip#

Control	Range	The fault it models
contacts	0–12 /s	How often the metal touches the board, per second. The gaps are drawn at random rather than counted off a clock, so two land together and then nothing happens for a second.
contact point	`sync separator` · `vertical oscillator` · `EHT / beam supply` · `chroma demodulator` · `video output stage`	Which point inside the set the clip is bridging, and which domain it damages. sync separator removes where the line starts, so the picture tears and takes hue with it. vertical oscillator collapses the scan toward a band and lets it spring back, decoded correctly throughout. EHT / beam supply droops the high-tension rail, so the raster swells and the limiter pulls the drive down late. chroma demodulator shorts the reference network, so hue shears without the picture moving. video output stage runs the guns out of headroom.
bite	0–1	How far the short goes while the metal is down: a fingertip on a pin, or a paperclip laid flat across it. The controls the point names travel this far from wherever they rest.
dwell	8–800 ms	How long one contact lasts. How fast the damage arrives and clears is the receiver's, not the clip's: a bite lands over two or three frames and takes five or six to let go, because the flywheel has to find sync again. Under about 40 ms the contact is gone before the picture has reacted.
chatter	0–1	How much the contact breaks up while it is down. Bare metal on a pin bounces and scrapes, and each break takes the contact off entirely. The set takes five or six frames to let go of a short, so a single bounce dips the damage rather than cancelling it.

Deflection#

Control	Range	The fault it models
bend amount	-120–120 us	How far the tube's own scan is displaced sideways, in microseconds of line time. Deflection-domain: the beam bends after decoding, so geometry warps while hue stays put.
shape	`flag` · `skew` · `bow` · `ripple`	How that displacement is distributed down the frame: 0 flag (a hook at the top that decays away), 1 skew (a straight lean), 2 bow (a barrel curve), 3 ripple (a wave down the screen).
decay / ripple period	1–480 lines	How many scan lines the shape takes: the decay length for the flag hook, or the wavelength for the ripple.
v size (underscan)	0.2–4 x	Vertical deflection amplitude, the service knob on the yoke. Below 1 the scan shrinks and the raster comes into view past the picture: the vertical interval, the head-switch band, and beam-off black beyond the retrace. Above 1 is overscan, which is how consumer sets shipped.
HV sag	-100–100 us	A bright picture draws beam current, which loads the high-voltage supply and lets the scan widen, so bright content stretches the geometry around it.
supply ring (0 droop, 1 chaos)	0–1	How well damped that supply is. At 0 it droops smoothly and recovers; toward 1 the tank rings, so a bright edge sets off a decaying wobble down the lines below it.
beam limiter	0–1	The automatic beam limiter. The flyback can only source so much average beam current, so past a threshold the set pulls video drive down, and the sense loop's time constant lands the dimming after the bright content that caused it. The knob undersizes the flyback and strips the damping, so the correction overshoots and the picture pumps.

Decoder#

Control	Range	The fault it models
Y/C comb	`trap` · `2-line` · `3-line`	How the TV separates brightness from colour, which share one wire. - trap: a notch filter. Cheap, and it mistakes fine detail for colour (rainbow fringing on stripes) and colour for detail (dot crawl on edges). - 2-line and 3-line: combs, which use the line-to-line subcarrier alternation and largely remove both artifacts.
S-video bleed	0–1	Chroma crossing into the luma path, as if the Y and C wires were shorted, so the subcarrier appears in the picture as a dense moving dot pattern over anything coloured.
chroma bandwidth	0.05–6 MHz	The colour demodulator's low-pass, which decides how fast colour may change across a line. Real sets are around 0.5 MHz, which is why colour bleeds past its edges while brightness stays crisp. Past about 1.5 it admits luma detail as cross-colour.
chroma trail	0–1	Asymmetric colour smear, trailing to the right only. A lagging chroma path drags colour behind the edge, which is the direction real sets and tapes smear.
chroma upsample error	1–8 px	How coarsely the demodulated colour is sampled before being stretched back up. Coarse sampling lands on the subcarrier lattice at intervals, so moving detail rainbows in blocks.
chroma gain	0–16 x	The colour control on the set: how much the demodulated chroma is amplified. Past 1 saturation blooms and clips against the edge of the gamut.
tint	-180–180 deg	The tint knob, which rotates the demodulator's reference against the incoming colour. Every hue turns together, so flesh goes green one way and magenta the other. At ±180 the reference is backwards and the picture comes out in complementary colour. It sits after burst lock's correction.
burst lock	0–1	How much the decoder trusts the colour burst it measured. At 1 it follows the burst, so phase errors in the signal are corrected out. At 0 it runs on its own crystal, so any subcarrier error shows as drifting hue.
demod axis	0–180 deg	The angle between the set's two synchronous colour demodulators, 90° apart only because the reference network says so. Cheap sets used non-quadrature X/Z axes deliberately. It shears the colour wheel: hues that were opposite stop being opposite, so the picture keeps some of its colours and loses others.
subcarrier detune	-200–200 kHz	The decoder's reference crystal pulled off 3.579545 MHz, the classic circuit-bend. The demodulation axis rotates continuously against the incoming colour, so hue sweeps the whole wheel. Turn burst lock down to let it run.
color killer	0–100 IRE	The burst amplitude below which the set decides the broadcast is monochrome and shuts colour off, in IRE. Raise it and anything that weakens the burst makes colour cut in and out in patches.
chroma AGC lag	0–240 lines	The time constant of the chroma AGC's control voltage, in scan lines of burst memory. At 0 the set corrects colour gain instantly per line, which no real ACC can. Raised, gain and the colour killer respond to burst damage tens of lines late, so colour blooms back after a dropout band and the killer chatters down the frame.
VIR correction	0–1	How far the set trusts the reference stamped on line 19. A VIR receiver decoded that line, compared it against what it knew was sent, and trimmed its own hue and saturation until the two agreed — a closed loop only as accurate as the reference arriving. A weak reference makes the set turn colour up. Needs the VBI test signals on.
VIR lag	1–240 frames	The corrector's time constant, in frames. Short and it chases the reference line by line, so damage that comes and goes makes the picture flicker. A real corrector was slow, responding over a second or more, so a bent reference drags the frame somewhere wrong and leaves it.
output stage clip	0–1	How the RGB output amplifiers run out of headroom. At 0 the matrix is fitted back into gamut without moving the hue. At 1 the three guns hit their rails one at a time, and the first to clip drags the hue toward the two still in range.
agc	0–1	How aggressively the receiver normalizes signal level off the sync tip. At 0 the gain is fixed, so anything that changes signal amplitude changes picture brightness directly.
encoder chroma bw	0.1–4 MHz	Colour bandwidth at the encode end, the camera's own limit. Wide enough and the chroma sidebands spill into the luma band as cross-colour.

Captions#

Control	Range	The fault it models
caption decoder	`off` · `on`	The set's own caption decoder, slicing line 21 off the signal it received. The caption is data, and it has been through everything the picture has: snow, a narrow channel and generation loss arrive as dropped characters, wrong ones, and a solid block wherever parity caught an error. The set paints the page on its own timing, so the picture can roll and tear underneath a caption that sits still. Needs vbi test signals on.
caption box	0–1	How black the box behind the characters is. Broadcast captions sat in a solid one because type keyed over picture is unreadable once the picture is bright.
rom address line	0–11	A pin held high on the character generator's font ROM, a literal circuit bend. Low lines carry the row inside the cell, so holding one makes every glyph repeat a scan line. High lines carry the character code, so holding one substitutes the whole font for its neighbour in the ROM. The font fills the low 1152 bytes of a 2 KiB part, so a line held high enough reads cells nobody programmed: all ones, a solid block.
rom data line	-8–8	The other bus. A font ROM's data lines are the eight dots across one row, so holding one lights or kills the same column of every character on the page.
rom crossed lines	0–10	Two adjacent address lines transposed — a chip seated a pin over, or two traces swapped. Low in the bus they carry the row inside the cell, so every glyph gets its scan lines shuffled. High in the bus they carry the character code, so the font is permuted in blocks.
rom counter slip	-4–4 /frame	The vertical blanking reset arriving late on the character-address counter, so the whole page is out by the same count and further out every frame. Twelve counts is one whole character: a slow rate crawls the font upward, a fast one churns the page through the alphabet.
rom slip per line	-0.25–0.25 /line	The horizontal reset failing instead, so the count grows on every scan line and the vertical still clears it: the damage runs down the block, and the shear grows inside a character as well as between rows. Twelve counts is one whole character and the block is ninety-six scan lines tall, so an eighth of a count a line walks the font one character from the top row to the bottom. The count restarts at the top of every field, so the pattern holds still.
rom cell strap	-11–12 rows	The cell-height strap on the row counter, in the wrong hole: one chip served 7-, 9- and 12-line cells and the board picked by a jumper. A glyph's first row is at its code times the cell height, and the raster keeps stepping 12 rows whatever the strap says, so the text shears into a diagonal slice of the font.
rom bit rot	-1–1	Charge that has leaked off the array over thirty years. A cell that has lost it reads back as the erased state, and which state depends on how the font was masked into the part: positive erases to a lit dot and the letters thicken, negative to a dark one and they crumble. The pattern is in the die.
page address line	0–7	A line held high on the counter that walks the page memory as the raster crosses the block. Low lines are the column, so holding one repeats columns in blocks of two, four, eight; high lines are the row. Every character is spelled correctly, from an undamaged font, and in the wrong place.
page counter slip	-4–4 /frame	The same slip on the counter walking page memory. Every character keeps its shape and the whole page walks diagonally through itself, a cell at a time.

Screen#

Beam#

Control	Range	The fault it models
blanking strobe	0–20 Hz	Holds the beam-blanking gate on, so the guns are cut for most of each cycle and let through in flashes. The gate sits one line above the phosphor, so light already on the glass keeps decaying through the dark. Everything downstream with memory sees the dark frames too, so a feedback loop pumps at the strobe rate. Lock it to the beat with ♩.
flash length	1–200 ms	How long the beam is let through each cycle. An absolute length, so speeding the strobe up does not shorten the flash. Under one frame it is one frame.
beam profile	0–1	The electron beam is a spot of finite height, so it does not quite fill the gap between scan lines. Raise it for a tighter spot and visible dark gaps.
beam bloom	0–1	The spot grows with beam current, so bright lines are fatter than dark ones and scanlines show in the shadows but close up in the highlights.
beam spot	0–12 px	How wide a spot the gun writes on the phosphor. The beam is a smooth blob, so light from one sample lands partly on its neighbours and every edge arrives as a ramp.
phosphor grain	0–1	The coating is a granular deposit of crystallites, so its emission is mottled. Fixed on the glass, and strongest in the mid tones.
reconstruction (bilinear→cubic)	0–1	How the sampled line is reconstructed into continuous light. Toward 0 is linear interpolation, which loses high frequencies; toward 1 is a cubic that stays flat past the subcarrier.
scan velocity mod	-4–4	Consumer sets faked sharpness by patching differentiated luma into an extra deflection coil, slowing the beam through a dark-to-bright transition and speeding it through a bright-to-dark one. Emission follows dwell time, so light is moved across the edge rather than added. Negative wires the coil backwards.
svm aperture	0.25–24 px	How wide a span the differentiator looks across. Narrow gives a tight edge-liner; wide reaches past the detail and shades whole objects.

Phosphor#

Control	Range	The fault it models
phosphors	`sRGB` · `P22` · `1953` · `green`	Which phosphors the tube is coated with, which sets its primaries. - sRGB: no conversion. - P22: SMPTE-C, a normal colour TV. - 1953: the wide NTSC primaries nobody ever built. - green: a long-persistence monochrome monitor.
phosphor persistence	0–0.9995	How long the layer keeps glowing after the beam has passed: afterglow in the glass. The decay is second-order, so the bright core of a trail loses almost all of itself at once and only the dim remainder lingers. A real picture-tube phosphor is gone well inside one field.
trail tint	0–6	The three phosphors do not decay at the same rate: red and blue die faster than green, so trails tint green as they fade.
trail scatter	0–1	Held light does not leave through the grain that emitted it. It scatters sideways through the layer and the glass into phosphor that is still glowing, so the spread compounds along a trail.

Mask & convergence#

Control	Range	The fault it models
aperture grille	0–1	Strength of the shadow mask or aperture grille, the vertical stripes of R, G and B phosphor the beam lands on.
grille pitch	1–48 px	Spacing of those phosphor triads in screen pixels. Pitches near a small whole number of pixels alias into moiré.
convergence error	-12–12 px	Three guns fire through one mask from three different positions, so they can only be registered over part of the screen. Nulled in the middle and worsening toward the corners, which is why an old tube fringes red and blue at the edges.
purity (magnetised patch)	-3–3	A patch of the shadow mask left magnetised by a speaker set too close. The field bends all three beams together, but a triad is three dots 120° apart, so the same nudge over-excites the dot it moves toward and starves the one opposite. Fixed on the glass, so a rolling picture travels through it.
patch x	0–1	Where the magnetised patch sits across the glass.
patch y	0–1	Where the magnetised patch sits down the glass.
patch size	0.02–2 h	Radius of the magnetised patch as a fraction of picture height. Small is a screwdriver left on the cabinet; large is a year next to a loudspeaker.

Source B#

The second deck, with its own cable ahead of the mixer, so a fault can hit one source alone.

Signal (source B)#

Control	Range	The fault it models
B hue	-180–180 deg	Proc-amp hue trim on B before the mix: a static phase offset on its subcarrier.
B video gain	0–6 x	Proc-amp video gain on B: contrast of the B picture before mixing, separate from how much of B is patched in.
B invert	0–1	Inverts B's picture. Mixed against A it reads as a difference key: where the two agree they cancel toward grey.
B deinterlace	`off` · `on`	The bob deinterlacer from source A, on B's own picture. Combing belongs to the source, so a progressive camera in A and an interlaced dongle in B need opposite settings.

Feed B · deck#

Control	Range	The fault it models
B pause (deck held)	0–1	The pause button on the B deck, at how badly the deck copes. The drum re-reads one track so the frame holds, but the capstan servo is defeated: B's timebase scatters line to line, a mistrack stripe of snow creeps down, and B's hue flickers at frame rate. On the clean path, which implies a time-base corrector, it just freezes the frame.
B dropouts	0–400 /frame	Dropout events per frame on B's own tape. The streaks land on B's raster, so they slip and roll with B's picture.
B dropout len	1–60 us	How long each of B's dropouts lasts, in microseconds of the 63.5 µs line.
B sync suppression	0–1	Head-end scrambling on input B alone: B's sync tips are lifted toward blanking before it is summed in, so B stops contributing to the sync contest.
B system	`gated` · `alternate` · `ssavi`	Which scrambling system B's channel uses. Gated suppresses every line, alternate every other line, and SSAVI also inverts B's active video.

Feed B · cable#

Control	Range	The fault it models
B loose connector	0–1	How loose the plug in input B's jack is. Bands of lines lose contact, re-rolled every frame. B is the input the receiver is not locked to, so a break decides whether B's bands lose picture or sync.
B bad contact	`pin` · `shield` · `both`	Which of B's two contacts is intermittent. The centre pin breaks the signal path, so bad bands of B collapse to noise and stop pushing sync. The shell breaks the ground, so hum lands on those bands instead. Both is a wiggled plug.
B ground loop	-40–40 IRE	A ground loop on input B's cable alone. The bar rides B's own raster, so it slips and rolls with B's picture, and it lifts B's sync tips, so how hard B fights for the line start varies at 60 Hz.
B noise	0–150 IRE	Snow on B's feed only. It rides B's own raster, so it tears and rolls with B's picture.
B termination (-1 daisy, +1 open)	-1–1	Termination fault on B's cable alone: negative halves B into a dim ghost under A, positive runs B hot and ringing so its sync pushes harder.
B polarity (flips sync)	0–1	A signal/ground swap on B's own connector, negating B's waveform and sync before the summing bus.

Camera feedback — feedback loop#

an optical loop. A camera points at the tube and the mixer feeds its picture back into the input ahead of the encoder. The camera itself can only do what a lens does — zoom, shift, defocus, black level — but the return re-enters ahead of the encoder, so a lap is a whole encode/decode generation and every fault between there and the glass is inside it, applied once per generation. Above unity gain it builds structure on its own.

Camera feedback (optical)#

Control	Range	The fault it models
mix	0–1	How much of the camera-pointed-at-the-monitor image returns to the input. The fader and the exposure multiply, so raise both until the product passes unity.
zoom	0.2–4 x	How much bigger or smaller the camera frames the screen each lap. Above 1 detail flows outward into tunnels, below 1 it collapses inward.
rotate	-180–180 deg	Camera tilt on the loop. Each pass rotates again, so structures spiral instead of expanding straight. A hundredth of a degree is visible.
shift x	-1–1	Camera aim off-centre horizontally, which moves the loop's fixed point: the tunnel mouth or spiral core.
shift y	-1–1	Camera aim off-centre vertically.
gain	0–3 x	Camera exposure on the loop. Round-trip gain is this times the mix above, so at a mix of 0.6 patterns persist only past 1.67. A collapsing loop holds well above unity; an expanding one goes to white.
auto-iris hunt	0–1	Puts the camera's exposure on an auto-iris servo metering the monitor it feeds. The loop brightens, the iris clamps a beat late, the loop starves, the iris reopens. Turned up it never settles.
defocus	0–12 px	Lens blur radius on the camera. Smoothing each generation keeps the loop on large structures.
vignette	0–1	Lens falloff toward the corners, so loop gain is high in the middle and feedback dies before the border.
black cut	0–0.2	The camera sensor's black level. Anything dimmer reads as pure black, so trails cut off.
cam s-curve	0–1	Sensor highlight compression. Bright areas roll into a shoulder instead of clipping flat.

Tube face (what the camera shoots)#

Control	Range	The fault it models
beam cutoff	0–0.95	The gun bias point. Drive below this emits no light, which sets the floor a feedback pass has to clear.
beam gamma	0.2–6	The gun transfer curve, light out against drive in. High gamma deepens shadows and stretches highlights.
beam saturation	0–6	Colour saturation of the emitted light, after the beam transfer. Feedback multiplies it every pass.
screen bloom	0–6	Light spreading out of bright phosphor cores: a tight halo that fattens highlights.
halation (warm halo)	0–6	Light scattering inside the thick glass faceplate and bouncing back: a wide, warm halo, broader and softer than bloom.
halation ∝ beam current	0–4	How much the halo widens with local beam drive. Real glass scatter grows with beam current, so peak white throws light much further than mid grey.
phosphor glow	0–4	Faceplate haze, the dull sheen a powered tube has even in black. It lifts the black floor.

Mixer feedback — feedback loop#

an electrical loop. The mixer takes the composite signal off the bus into an input and crossfades it against the live signal. The subcarrier travels round with it, so each sample of cable delay rotates fed-back hue by 90° per generation.

Mixer feedback (electrical)#

Control	Range	The fault it models
loop mix	0–1	The crossfader position toward the loop bus, where the mixer patches the previous frame's composite back in. The subcarrier goes round too, so the loop shifts hue as well as geometry. All the way over the loop feeds only on itself.
loop gain	-3–3 x	Proc-amp trim on the loop return. Past ±1 the round trip exceeds unity and the loop builds until it clips. Negative inverts each pass.
loop delay	0–63 us	Delay on the loop return. The subcarrier rides the same waveform, so delay is also a hue rotation: one sample, 70 ns, is a 90° spin.
loop timebase pull	-60–60 us	The loop's delay trimmer replaced by a varactor hanging off the video bus, so the fed-back waveform tunes its own delay. Bright content and sync tips pull opposite ways, and every 70 ns of pull is another 90° of hue.
v offset	-240–240 lines	Vertical offset applied each trip around the loop, so trails stack into ladders.
loop key	-1–1	Keys the loop return, so only part of the picture feeds back and the loop follows a subject. Positive keeps one side of the slice, negative the other. It slices level until the acceptance angle switches it to hue.
key level	0–100 IRE	The brightness the loop key slices at, in IRE (0 blanking, 100 peak white).
key soft	1–30 IRE	How wide the key transition is, in IRE. Narrow cuts a hard edge.
key input	`self` · `program`	Which connector the keyer's key input is on. Self is the loop return, so the trail draws its own boundary a generation late. Program is the live picture, so a moving subject carves its shape out of what the loop has accumulated.
key hue	0–360 deg	Which chroma phase the keyer slices at, once the acceptance angle switches it from level to hue. 241 is a green backing.
key acceptance	0–180 deg	Swaps the loop's luma keyer for a chroma one: a wedge this wide either side of the key hue, and at zero back to slicing level. The loop delay is a hue rotation, so a region regenerates until its own return spins out of the wedge.
strobe hold	0–60 frames	Freezes the loop's frame store for this many frames before it grabs again, like a frame synchronizer stuttering.
trails	0–1	Peak-hold decay in the loop's frame store: bright areas are retained and fade instead of being replaced. A frame synchronizer's smear, ahead of the phosphor.
loop resonance freq (0 off)	0–5 MHz	A resonant filter in the loop, centred here. Around 3.58 MHz it rings on the subcarrier; lower down it turns edges into repeating bars.
loop resonance Q (broad→ringing)	0–1	How selective that resonance is. Narrow rings for a long time after every edge and lays a fixed-frequency pattern across the line.
loop resonance boost	0–16 x	In-band gain added by the resonance. Push the round trip past unity at that frequency and the loop self-oscillates.
loop ring mod	0–1	The loop bus multiplied instead of summed: a ring modulator with one input patched to the machine's own past, so the spectrum folds over itself every lap.
read clock error	-2–2 %	The loop's frame store read out at a clock this far off the one it was written at. The readout re-triggers on each line's sync, so the picture stretches sideways from the line start. The subcarrier is in those samples, so hue turns further from the line start — eighty degrees by the right edge at a thousandth off.
loop noise	0–10 IRE	The loop amplifier's own noise floor, added to the return every lap, so each generation carries every earlier one's noise scaled by the loop gain. Near unity a fraction of an IRE seeds structure.
frame sync on the return	0–1	Whether the return comes back through a frame synchronizer or down a bare cable. A genlocked store writes its own sync and burst on the way out, so only the picture circulates and the loop can be driven hard. On the cable the loop's sync tip goes round one delay late and lands inside a line, so the flywheel free-runs.
return (Y/C split)	`composite` · `chroma` · `luma`	A Y/C separator on the loop return and a recombiner after it, so one wire comes round the loop and the other from the live picture. Composite sends the whole waveform round. Chroma sends the loop's colour over live brightness, so hue accumulates while the picture stays sharp. Luma sends brightness and the sync tip under live colour, so trails stack in grey.
ring carrier	`program` · `oscillator`	Which connector the ring modulator's other input is on. Program is the live picture, so both sides carry the same crystal and chroma against chroma lands at DC and 7.16 MHz, outside the chroma band. Oscillator patches the box's own subcarrier generator there, making the bridge a chroma modulator: brightness translates onto 3.58 MHz and reads as colour, colour down to brightness.
ring carrier detune	-200–200 kHz	How far that oscillator sits off the house 3.579545 MHz. At zero it agrees with the encoder, so brightness comes back as a single hue. Detuned, the phase it writes ramps through the frame, so the manufactured colour turns along every line.

Sound#

Hangs off Receiver — a mic, a track, a clip’s own audio, or whatever this machine is playing, patched into the receiver. It detunes both hold oscillators, loads the HV supply, drives the deflection and turns the colour reference, so the sound disturbs the set and the picture moves with it.

Audio routings#

Control	Range	The fault it models
bass → vertical hold	0–32 Hz	Bass energy detunes the vertical oscillator, so kick drums shove the frame vertically and it settles back. The field rate itself is moving.
level → horizontal hold	-3000–3000 Hz	Overall audio level pulls the horizontal oscillator off frequency, so loud passages skew and tear the picture sideways. Negative leans the tear the other way.
bass → HV sag	0–160 us	Bass loads the high-voltage supply as if the beam were drawing current, so the scan collapses on each hit and springs back. Needs supply ring (in Deflection) above zero.
waveform into deflection	-80–80 us	The audio waveform patched into the horizontal deflection, one sample per scan line. Deflection-domain, so hue stays put while the glass bends.
audio into HV tank	0–12	Drives the audio into the high-voltage tank alongside the beam current, so the supply rings with the music instead of just sagging. Needs bass → HV sag above zero.
audio into video in	0–150 IRE	The audio patched straight into the video input, in IRE. Loud passages land on the sync tips and the burst as well as the picture: the classic wrong-cable result.
waveform into hue	-180–180 deg	The audio waveform driven into the colour demodulator's reference oscillator, one sample per scan line — the same wire the tint control sits on. The reference is in the receiver, so the hue bands stay on the glass while a rolling picture slides through them.
input trim	0–16	Input trim on the waveform routings, into deflection and into video in. The envelope routings (the two hold oscillators and HV sag) normalize against a decaying peak, so this trim does not move them.

View#

Hangs off Screen. Viewing conditions: nothing here is in the signal path.

View#

Control	Range	The fault it models
magnifier	0.25–12 ×	Where your eye is, up against the glass. Everything that lives on the screen rather than in the image magnifies with it: scanline structure, the beam spot, phosphor grain, the grille triads.
magnifier x	0–1	Which part of the glass is under the magnifier, across. Ignored at 1×.
magnifier y	0–1	Which part of the glass is under the magnifier, down. Ignored at 1× and below.
slow motion (1 = realtime)	0–1 x	Steps the whole simulation at a fraction of display rate. Noise, rolls, sweeps, feedback loops and phosphor all slow together, and 0 freezes the frame. Modulation stays live, so an LFO here warps time itself.
frame rate lock	`off` · `1/2 rate` · `1/3 rate` · `1/4 rate` · `auto`	Renders every second, third or fourth display refresh. A path that costs slightly more than a refresh interval otherwise wavers between full rate and half, which reads as stutter. - off: render every refresh. - 1/2 rate, 1/3 rate, 1/4 rate: a fixed rate. The simulation steps once per rendered frame, so rolls and noise move proportionally slower. - auto: sustained missed refreshes engage the half-rate lock, and it retries full rate with a lengthening pause.

Features — the tour · User guide — driving it · How it works — the code