When evaluating an IR thermal camera, engineers fix on one key metric: NETD. Among the many "culprits" that determine NETD, photon noise is special — not from device defects or process shortfalls, but a limit set by nature itself. This is photon noise, also called shot noise. It is the quantum limit that LWIR detection cannot escape. Understanding it is the cornerstone of detector physics and the key to distinguishing MWIR and LWIR design philosophies.

01 Quantum essence: the never-silent photon "rain"

The root of photon noise lies in quantum mechanics. Even an ideal, perfectly stable IR source emits not a smooth "water flow" but a random "photon rain". Photon emission is a Poisson process; the number arriving per unit time fluctuates randomly around a mean. That fluctuation is photon noise. The math is simple yet deep: mean-square noise current = 2eIΔf, where e is electron charge, I the mean photocurrent, Δf the bandwidth. Photon-noise power is proportional to the square root of signal strength — stronger signal means larger absolute noise, but SNR actually improves with signal (SNR ∝ √I). The core conclusion: what limits ultimate sensitivity is often the photon noise from the background radiation itself, not the signal source.

02 The band debate: why LWIR is photon noise's home turf

Photon-noise intensity is directly proportional to the incident photon flux rate. Besides signal photons from the target, the background filling the whole field of view is the stronger source. By Planck's law, a room-temperature object (~300 K) radiates with a peak near 9.7 μm — right in the LWIR band. So an LWIR camera viewing ambient scenes is immersed in intense background radiation from walls, air and ground. MWIR receives 1–2 orders of magnitude less ambient background. Thus in typical room-temperature scenes: LWIR systems have extremely strong background and background-limited photon noise often dominates NETD; MWIR has weaker background and its noise may be dominated by detector thermal or readout noise. This is the physical origin of "LWIR limited by background, MWIR by the device", dictating the two routes' design strategy and ceiling.

03 Technical countermeasures: how to "suppress" background photon noise

  • Cold shield & cold stop: the core of cooled IR systems — the chip is packaged in a vacuum dewar and cooled to 77 K or below, with a cooled cold stop in front that only lets light from the lens field of view reach the detector, blocking thermal radiation from other directions. Can reduce background by orders of magnitude, the fundamental guarantee of mK-level or lower NETD.
  • Spectral filtering & band optimization: a narrowband filter in front of the detector passes only the target's characteristic band, rejecting out-of-band background — extremely effective in gas-leak detection (e.g. methane 3.3 μm) or high-temperature measurement.
  • Aperture & integration time: larger aperture (lower F-number) scales both signal and background photons, but SNR is proportional to aperture diameter — a direct boost; longer integration accumulates more signal photons (linear) while noise grows as the square root, so SNR improves as the square root of integration time — the classic frame-rate-for-sensitivity trade.
  • Lower detector temperature (cooled photodetectors): cooling MCT etc. not only suppresses background but exponentially reduces dark current (leakage with no light), whose shot noise can then nearly vanish, bringing the system truly to the background limit.

04 Engineering insights & trade-offs

  • LWIR or MWIR? For ambient targets seeking high sensitivity/low NETD, the background-dominated LWIR band demands a cooled detector with large aperture and high-performance cold stop. For high-temperature targets (engines, furnaces) or hot backgrounds, the target's MWIR radiance rises while background stays relatively weak, so MWIR may show better SNR.
  • Why uncooled cameras have higher NETD? Uncooled microbolometers are thermal detectors, responding to total heat not photon energy; their dominant noise is thermal and 1/f, not photon. The non-zero thermal conductance of their isolation structure limits responsivity and stability, so NETD is typically an order of magnitude higher than cooled photon detectors.
  • Performance insight: when a cooled LWIR camera claims extremely low NETD (<20 mK), it implies its optics (F-number), cold-stop efficiency and detector quantum efficiency are already near the background-limit theoretical ceiling — further big gains may need a revolutionary breakthrough.

Photon noise, the faint "hiss" from the universe's quantum nature, draws a soft yet hard boundary on IR sensitivity. Fighting it has driven cold-stop design, deep cryogenic cooling and adaptive filtering. Seeing through parameters to the physical essence behind them — like photon noise — is the foundation of rational choices among countless technical options.