PAL/NTSC burst amplitude error

Visible effect

A wrong reference changes the recovered colour strength

A colour burst is a short reference signal sent outside the visible picture. Here its amplitude changes while active-picture chroma—the signal carrying the picture’s colour—stays unchanged.

Colours look washed out when the transmitted burst is too strong, or excessively colourful when it is too weak, if the receiver regulates chroma gain from that reference. Neutral grey remains neutral and edges stay in place. Before display clipping, the colour vectors grow or shrink without rotating: this is an amplitude error, not a hue error.

The picture alone cannot identify the cause. A saturation control or an error in active-picture chroma amplitude can look similar. Compare the burst on the back porch with active chroma: here only the reference is wrong. With fixed chroma gain and maintained phase lock, the same burst error leaves the picture unchanged.

Try the strong-burst preset, then switch Receiver chroma gain to Fixed. The waveform still shows the faulty burst, but the picture returns to its original colour strength. That unchanged picture is the intended comparison: the visible error depends on the receiver’s automatic adjustment.

Physics

The receiver measures a reference outside the visible picture

PAL and NTSC transmit colour differences on an oscillating colour subcarrier. A short known sample, the colour burst, sits on the back porch: the blank interval after horizontal sync and before active picture. It gives the decoder a phase reference and, in a burst-referenced receiver, an amplitude reference. Analogue television colour systems explains the encoding and PAL’s alternating reference; How analogue television carries an image locates burst within the line.

The receiver’s automatic chroma control (ACC) adjusts chroma gain to bring the measured burst toward its expected level. That same gain applies to picture chroma. This normally compensates attenuation shared by burst and picture chroma. If the encoder instead changes burst alone, the receiver corrects a loss or excess that the picture does not have.

The fault belongs to transmitter video formation: a wrong burst-insertion level or reference-path gain can change the relationship before RF modulation. The visible consequence is produced by the receiver. The experiment assumes correct phase and frequency lock, correct PAL line switching, ideal chroma separation, and settled gain control. PAL’s phase-error averaging does not remove this amplitude-reference error.

Mathematics

Inverse reference amplitude sets the chroma gain

B is the transmitted burst amplitude and B₀ is its correct value, measured with the same peak or peak-to-peak convention. Their dimensionless ratio b avoids assigning one voltage to every PAL/NTSC variant. The cited digital decoder measures burst power, proportional to B²; restoring nominal amplitude gives g = √(B₀²/B²) = 1/b.

The example limits gain to 0.5–2×. These are teaching bounds, not an ADV7842 specification. The burst slider covers 25–200% of nominal, so below 50% the gain stops rising. Fixed mode sets g = 1. A zero burst is excluded: without a dependable reference, colour lock and automatic colour suppression require a different model.

Y′ is encoded luma, using 0.299R′ + 0.587G′ + 0.114B′. C denotes both colour-difference coordinates. Scaling both by one positive gain preserves their angle and therefore hue in the unclipped signal. With burst at 160%, g = 0.625; at 60%, g ≈ 1.67. At nominal burst, g = 1 and the mapping is the identity.

More generally, if active chroma is multiplied by a and burst by b, the ideal unclamped response is a/b. This experiment fixes a = 1 to isolate burst error. Final RGB display clipping can change apparent hue and luma when amplified colours leave the display gamut.

Shader

Scale colour differences around unchanged encoded luma

The shared GLSL function takes one gamma-coded RGB sample, nominal-relative burst amplitude, and a numeric receiver mode (1 automatic, 0 fixed). It returns unclipped RGB. Writing RGB as a neutral luma vector plus RGB-minus-luma scales both colour differences without a second colour-space matrix or spatial sampling. The wrapper makes one aspect-fill texture read, calls this exact function, blends by Effect mix, and lets the normalized display framebuffer clip RGB.

// WHAT: Recover colour with a wrong burst-amplitude reference in PAL or NTSC.
// HOW: A locked automatic chroma control uses inverse burst amplitude;
//      fixed gain ignores it. Scale both colour differences around unchanged Y′.
// WHY: Burst-only error changes the chroma/reference ratio, not hue or timing.
// Inputs: gamma-coded RGB, positive nominal-relative burst, ACC mode (0 or 1).
// Output: unclipped RGB. The display framebuffer alone clips out-of-gamut values.
float burstChromaGain(float burstRatio, float automaticGain) {
  return automaticGain > 0.5
    ? clamp(1.0 / max(0.25, burstRatio), 0.5, 2.0)
    : 1.0;
}

vec3 applyBurstAmplitudeError(vec3 rgb, float burstRatio, float automaticGain) {
  float y = dot(rgb, vec3(0.299, 0.587, 0.114));
  float gain = burstChromaGain(burstRatio, automaticGain);
  // Preserve exact identity at nominal burst or fixed receiver gain.
  if (gain == 1.0) return rgb;
  return vec3(y) + gain * (rgb - vec3(y));
}

Why these steps are here

1. Use the burst ratio to derive receiver gain, rather than directly multiplying picture saturation by burst amplitude.
2. Bound the inverse gain and explicitly separate fixed gain from automatic control.
3. Keep the encoded luma vector and scale only its colour residual; neutral samples are invariant.
4. Leave phase, edge timing, and spatial bandwidth untouched.
5. Apply display clipping at the output, so the model’s conserved quantities are distinguished from display limitations.

Notes

• This is a steady-state decoder response to a transmitter burst-only fault, not a sampled composite encoder/decoder or RF simulation. The source picture stands in for ideally separated components.
• The waveform shows eight illustrative cycles on a fixed nominal-relative scale. It is a burst-interval view, not a calibrated full line or a specification of burst duration. PAL’s alternating phase and standard-dependent subcarrier frequencies are omitted because this experiment holds them correct.
• Gain limits, settling, reference detection, colour-killer thresholds, noise, overload protection, and peak-colour limiting differ between receivers. No universal weak-burst threshold or loss-of-colour sequence is claimed.
• PAL and NTSC share this amplitude mechanism under the stated assumptions. SECAM uses line-sequential FM colour and is outside the article.
• The source is treated as encoded video RGB, not linear light. The transform preserves encoded luma and hue before output clipping; it does not preserve physical luminance or promise unchanged perceived hue at gamut limits.
• The colour-killer is a receiver circuit that suppresses unreliable colour when the burst reference is too poor. It is deliberately excluded here, including in Fixed mode; fixed chroma gain alone does not guarantee that a real receiver keeps showing colour.
• Effect mix is a comparison aid. It changes the shown gain to 1 + mix·(g − 1), while the waveform continues to show the transmitted burst error.

References