PRNU / DSNU Visualizer

Visualize fixed pattern noise (FPN) in image sensors — both photo response non-uniformity (PRNU) and dark signal non-uniformity (DSNU). Observe spatial noise patterns and statistical distributions.

PRNU / DSNU Visualizer

Visualize Photo Response Non-Uniformity (PRNU) and Dark Signal Non-Uniformity (DSNU) as fixed pattern noise on a 2D pixel array.

PRNU: 1.0%

DSNU: 5 DN

Signal Level: 1000 DN

View

FPN RMS

11.18 DN

PRNU RMS

10.00 DN

DSNU RMS

5.0 DN

Peak-to-Peak

74.6 DN

Pixel Array Heatmap (32x32)

Distribution

Model scope

Use this browser tool for intuition, relative trends, and design-space exploration. Its local simplified model is not a substitute for RCWA/FDTD sign-off, silicon calibration, or vendor process data.

Learn more

Noise, SNR, and Dynamic Range · EMVA 1288 Characterization

Physics Notes

Fixed-Pattern Non-Uniformity

Plain-English Intuition

No two pixels are exactly identical — tiny manufacturing variations make some slightly more sensitive than others. PRNU is the fingerprint of those per-pixel sensitivity differences (visible in bright frames), and DSNU is the equivalent pattern in the dark baseline. Unlike random noise, these patterns are fixed in place from frame to frame, so they can largely be calibrated out.

PRNU and DSNU model pixel-to-pixel response variation that remains spatially fixed across frames.

Assumptions

• PRNU and DSNU fields are synthetic spatial patterns used to explain fixed-pattern noise behavior.
• PRNU is multiplicative with signal; DSNU is additive and visible in dark or low-signal frames.
• The model does not represent actual wafer maps, column circuits, hot pixels, or temperature-dependent drift.

Outputs

• Synthetic flat-field and dark-field patterns, PRNU/DSNU amplitudes, corrected residuals, and visual artifact examples.
• A calibration intuition for separating multiplicative gain variation from additive dark offset.

Validation Example

• Increasing signal should make PRNU more visible while DSNU remains an additive baseline pattern.
• Dark-frame subtraction should reduce DSNU-like offsets but should not remove multiplicative PRNU from bright flat fields.

Core Equations

Pixel signal model

$$Y_{ij} = (1 + g_{ij})S + d_{ij} + n_{ij}$$

• \(g_{ij}\): Local gain deviation (PRNU)
• \(d_{ij}\): Local dark offset (DSNU)

g_ij is PRNU and d_ij is DSNU/dark offset.

PRNU

$$PRNU = \frac{\sigma_{\text{flat}}}{\mu_{\text{flat}}}$$

• \(\sigma_{\text{flat}}\): Std dev of flat field

PRNU is measured from illuminated flat-field frames after offset correction.

DSNU

$$DSNU = \sigma_{\text{dark}}$$

• \(\sigma_{\text{dark}}\): Std dev in the dark

DSNU describes spatial variation of dark signal.

Model Interpretation

• PRNU grows with signal; DSNU is present even in darkness.
• Fixed-pattern noise can often be calibrated, but residuals remain after temperature and aging changes.
• The visualizer uses synthetic spatial fields, not measured wafer maps.

Pattern Components

• PRNU is multiplicative, so its visible amplitude grows with signal level.
• DSNU is additive, so it is most visible in dark or low-signal frames.
• Random noise averages down over frames, but fixed-pattern structure stays locked to pixel coordinates.

Calibration Strategy

• Use dark frames for offset maps and flat fields for gain maps; update them when temperature or analog gain changes.
• Normalize flat fields carefully so lens shading is not mistaken for pixel PRNU.
• Residual fixed pattern after correction often reveals unmodeled temperature, exposure, or column effects.

Known Missing Physics

• The synthetic field does not represent wafer reticle patterns, column circuits, or per-color CFA mismatch.
• It does not model temporal instability such as RTS pixels or temperature-dependent offsets.
• Real correction should be evaluated after black-level, lens-shading, and demosaic stages.

References

• EMVA 1288 Standard, Release 4.0