Pixel SNR vs Illuminance

Plot signal-to-noise ratio as a function of photon count (illuminance). Visualize noise breakdown by source and compare actual sensor performance against the ideal shot-noise limit.

SNR vs Illuminance

Analyze SNR across signal levels with noise breakdown. Compare different pixel configurations and identify noise-limited regions.

QE (550nm): 70%

Full Well Capacity: 10,000 e−

Read Noise: 3.0 e−

Dark Current: 10 e−/s

Exposure Time: 33 ms

Pixel Area: 1.00 μm²

SNR_max

40.0 dB

Unity SNR

6 photons

20dB Threshold

156 photons

Saturation

14286 photons

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

Physics Notes

Illuminance-to-SNR Conversion

Plain-English Intuition

How does scene brightness (measured in lux) turn into image quality? This tool walks the whole chain: lux becomes photons per second per pixel, photons become electrons through QE, and electrons then compete with shot, dark, and read noise. The output curve shows where you are read-noise-limited (low light) versus shot-noise-limited (well-lit).

This tool shows how scene illuminance becomes photons, photoelectrons, and finally SNR after adding sensor noise sources.

Assumptions

Lux-to-photon conversion requires an assumed spectrum and optical throughput; lux is not a unique photon count.
Signal is computed from photon count, pixel area, exposure time, and QE, then combined with shot, dark, and read noise.
The model reports scalar SNR and omits denoise, tone mapping, demosaic, scene contrast, and color noise.

Outputs

SNR versus illuminance, photon/electron counts, noise-regime transitions, and read-noise or shot-noise dominance.
A low-light design view showing how QE, pixel pitch, exposure, F-number, read noise, and dark current move the curve.

Validation Example

At high illuminance before saturation, SNR should approach the shot-noise trend $\sqrt{S}$.
At low illuminance, lowering read noise should improve SNR more strongly than increasing full well.

Core Equations

Electron signal

$$S = N_{\text{ph}} \cdot QE$$

$N_{\text{ph}}$: Incident photons

Optical throughput and quantum efficiency turn photons into collected electrons.

Total noise

$$\sigma = \sqrt{S + D \cdot t + \sigma_{\text{read}}^2}$$

$\sigma$: Total RMS noise

The simulator separates shot, dark, and read-noise contributions.

Ideal limit

$$\text{SNR}_{\text{ideal}} = \sqrt{S}$$

$S$: Signal charge

This is the best possible photon shot-noise limit for a given signal.

Model Interpretation

Low-light SNR is usually read-noise and photon-starvation limited.
Large pixels collect more photons at the same illuminance and exposure.
Illuminance-to-photon conversion depends on spectrum, lens f-number, and calibration assumptions.

Lux To Electrons

Illuminance is photopic and human-eye weighted, so converting lux to photons requires an assumed spectrum.
Lens f-number and transmittance set how much scene radiance reaches the pixel.
Pixel area, exposure time, and QE convert the arriving photon flux into signal electrons.

Reading The Curve

At the dim end, the curve is flat because read noise dominates.
In the middle, SNR rises roughly with $\sqrt{S}$ as photon shot noise dominates.
At high illuminance, full well and PRNU can limit further SNR improvement.

Known Missing Physics

The conversion from lux to photons is spectrum dependent and not unique.
The model does not include lens flare, scene contrast, demosaic, denoise, or tone mapping.
Real low-light quality depends on color noise, fixed-pattern noise, and temporal processing as well as scalar SNR.