COMPASS

English

한국어

English

한국어

Appearance

Pixel Scaling Trends ​

Explore how key image sensor performance metrics scale as pixel pitch shrinks, with reference data from commercial sensors.

Pixel Scaling Trends

Explore how key sensor metrics scale with pixel pitch. Compare theoretical scaling laws against published sensor data.

Technology Node:
Full Well Capacity (ke⁻)
0.51.01.52.02.53.002040608010065nm40nm22nmIMX586IMX766IMX890IMX989IMX283Pitch (µm)
Max SNR (dB)
0.51.01.52.02.53.0253035404550IMX586IMX766IMX890IMX989IMX283Pitch (µm)
Nyquist Frequency (lp/mm)
0.51.01.52.02.53.02004006008001000Pitch (µm)
Diffraction QE (%)
0.51.01.52.02.53.0020406080100Pitch (µm)
FWC
10000 e⁻
Max SNR
40.0 dB
Nyquist
500 lp/mm
Diff. QE
28.1%
Dark Current @60°C
5.0 e⁻/s

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

Resolution, MTF, and Pixel Scaling

Physics Notes

Pixel Pitch Scaling Laws

Plain-English Intuition

Making a pixel smaller is not just shrinking — it changes everything at once. Smaller pixels collect fewer photons (lower SNR), hold less charge (lower full-well), are hit harder by diffraction blur, and demand much stronger read-noise reduction. This tool lays out those trade-offs side by side so you can see what scales how.

Shrinking pixels changes photon count, full well capacity, crosstalk, diffraction sensitivity, and read-noise requirements together.

Assumptions

  • Scaling laws are first-order proportionalities; real process nodes add architecture-specific offsets and improvements.
  • Photon count and approximate full well follow area trends unless microlens, BSI, DTI, and capacitance design compensate.
  • Diffraction pressure is evaluated against pitch using a simplified Airy diameter proxy.

Outputs

  • Pitch-dependent area, photon/SNR tendency, full-well tendency, diffraction ratio, and mitigation context.
  • A tradeoff map for when shrinking pitch starts requiring stronger optics, isolation, binning, or read-noise reduction.

Validation Example

  • Halving pitch should reduce area by 4x and reduce shot-limited SNR by roughly 2x at fixed illuminance and exposure.
  • The ratio $D_{\text{Airy}}/p$ should increase as pitch shrinks or f-number grows.

Core Equations

Area scaling
$$A_{\text{px}} = p^2$$
  • \(A_{\text{px}}\): Pixel area

Photon capture and approximate capacitance both scale with pixel area.

Shot-noise limit
$$SNR_{\text{shot}} = \sqrt{N_e}$$
  • \(N_e\): Number of electrons

If the collected electron count falls by 4x, shot-limited SNR falls by 2x.

Diffraction ratio
$$\frac{D_{\text{Airy}}}{p} \approx \frac{2.44 \lambda N}{p}$$
  • \(D_{\text{Airy}}\): Airy disk diameter

This ratio rises as pixels shrink, increasing optical sharing between pixels.

Model Interpretation

  • Modern sub-micron pixels rely on BSI, DTI, microlens shift, and computational binning.
  • Pitch scaling is not a single-variable problem because optics and electronics scale differently.
  • Commercial trend lines should be treated as context, not as a process design rule.

What Scales With Pitch

  • Photon count falls roughly with $p^2$ for the same illuminance and exposure.
  • Diffraction pressure rises through $D_{\text{Airy}}/p$, so the same lens f-number becomes harder for smaller pixels.
  • Read-noise requirements become stricter because fewer electrons are available before SNR collapses.

Mitigation Stack

  • BSI improves optical access; DTI limits crosstalk; microlens shift recovers CRA response; binning recovers photon statistics.
  • None of these fixes is free: each adds process complexity, optical side effects, or signal-processing assumptions.
  • Scaling should be evaluated at the camera-system level, not only as a pixel-layout shrink.

Known Missing Physics

  • Simple scaling laws ignore detailed capacitance, source-follower noise, CFA geometry, and wafer process differences.
  • They do not predict HDR behavior, dual conversion gain, or computational multi-frame recovery.
  • Treat trend curves as first-order pressure indicators, not final performance forecasts.

References