12 µm or 17 µm: Which Pixel Pitch Wins Your Case?

When you spec a thermal camera module, a thermal imaging module, or a complete thermal imager module for a real mission, the 12 µm vs. 17 µm decision is not a fashion choice—it’s a system choice. Pixel pitch changes how much lens you need to achieve a given field of view (and therefore range), what that lens will weigh and cost, how the detector will perform on NETD, and how close you’ll run to the diffraction limit at long-wave infrared. The right answer depends on your optics budget, platform constraints, and target geometry—not on a single marketing bullet.

What “pixel pitch” really changes in system terms

Pixel pitch is the center-to-center spacing of pixels on the microbolometer. Over the last two decades, uncooled LWIR sensors have shrunk from ~45 µm to today’s common 17 µm and 12 µm families, while array sizes increased (e.g., 640×480/512 and beyond). Smaller pitch lets you pack more sampling points into the same focal plane and, for a given lens, yields a smaller instantaneous field of view (IFOV) per pixel—more pixels across the same target, all else equal. 

Two first-principles relationships govern what you get on target:

• • Rectilinear FOV: , where sss is sensor dimension (width/height/diagonal) and fff is focal length. 

• • IFOV (per pixel, radians): approximately . Smaller pitch or longer fff reduces IFOV, increasing angular detail per pixel. (Engineers also estimate range using Johnson-style perceptual thresholds in pixels across the target.) 

From these, a few truths follow:

• Same lens, different pitch. If you keep the same focal length, 12 µm collects ~17/12 ≈ 1.42× more pixels per angle than 17 µm. That directly pushes detection/recognition/identification distances upward in planning models that scale with pixels on target.

• Same FOV, different lens. If you demand the same FOV, a 12 µm sensor reaches it with a shorter focal length (roughly 0.71×), which usually means a smaller-diameter, lighter, cheaper LWIR lens for the same f-number. That’s a real BOM and payload win.

On the noise side, NETD (noise-equivalent temperature) is your practical yardstick for sensitivity. Typical uncooled imagers sit in the tens to hundreds of millikelvin range, and system NETD reflects detector physics, optics f/#, and processing. Smaller pixels can challenge SNR if all else is equal, but modern processes and optics often claw that back. Treat NETD as a measured system property, not a guess from pitch alone.

Finally, at LWIR wavelengths (~8–14 µm), diffraction matters. The diameter of the diffraction blur (Airy disk) in the image plane scales with $2.44\lambda N$ (where $N$ is f/#). At $\lambda \approx 10\mu \text{m}$ and f/1.0, that central lobe is on the order of ~24 µm across—meaning a 12 µm pixel slightly oversamples a good f/1 lens, while 17 µm is closer to the diffraction footprint. This is one reason 12 µm can yield finer detail if the lens and mechanics support it.

Market reality: security vs. UAV

For ground security, camera placement is fixed and wind-shake is your enemy. The 12 µm route often lets you step down a focal length for the same FOV, cutting lens cost and mast load while keeping enough pixels on human-sized targets to meet your response thresholds. For UAV gimbals, weight and stability dominate. A 12 µm core can hold the same swath with a shorter, lighter lens, directly helping endurance and PID tuning. If you try to chase range with a heavy, long 17 µm lens on a small gimbal, you may give back the “spec win” as blur in real wind.

A practical comparison (same array size, 640-wide)

Below we compare what happens when you pick 12 µm vs. 17 µm for a 640-wide sensor, first holding focal length fixed (more pixels on target), then holding FOV fixed (smaller lens).

Case A — Hold focal length constant (e.g., 35 mm)

• IFOV scales with pitch: IFOV12/IFOV17≈12/17\mathrm{IFOV}_{12} / \mathrm{IFOV}_{17} \approx 12/17IFOV12/IFOV17≈12/17.

• Johnson planning (pixels across target) improves by ~1.42× for 12 µm, implying a similar gain in detection distance for a given target width when contrast and atmosphere cooperate.

Case B — Hold FOV constant (same horizontal FOV goal)

• To match the 17 µm sensor’s FOV, the 12 µm option uses roughly 0.71× focal length; for f/1 optics, lens diameter scales with $f$. Net: lighter/cheaper lens for the same view, with the same pixels across the scene (still 640).

Either way, smaller pitch gives you room to optimize: you can take the range boost (Case A) or the lens/BOM savings (Case B)—or split the difference for a balanced spec.

What about NETD and MRTD?

NETD captures how much temperature change equals the sensor’s noise floor; MRTD (minimum resolvable temperature difference) ties that sensitivity to spatial detail using a bar-target test and the system’s MTF. NETD numbers in the tens of millikelvin are typical for modern uncooled systems; MRTD curves show how that sensitivity translates across spatial frequencies. Put simply: lower NETD and stronger MTF make small, low-contrast targets easier to see at distance. Pixel pitch alone does not guarantee lower NETD, but 12 µm can still win on effective range when paired with good f/1 optics and clean processing.

Diffraction and “useful sampling” at LWIR

LWIR’s long wavelength makes diffraction a first-order constraint. Using the Airy expression, f/1.0 at 10 µm yields a ~24 µm central spot; at f/1.2 it’s ~29 µm. That means a 12 µm pixel oversamples the spot (≈2 pixels across the lobe), which helps interpolation, deconvolution, and EIS; 17 µm is closer to critical sampling. If your lens is not close to diffraction-limited, or mechanics blur detail, the extra sampling from 12 µm won’t fully convert to range—but when the optics are good, it does.

Cost and weight implications (the often-ignored win)

Germanium and coatings dominate lens cost. Because focal length and entrance pupil diameter scale together for the same f/#, a 12 µm system meeting the same FOV can step down focal length and front element size. That reduces lens cost, weight, and moment of inertia, easing PTZ stress on masts and extending flight time on drones. For integrators chasing ROI, this is typically the decisive advantage of 12 µm over 17 µm—even before you count the range boost of Case A.

A math-grounded mini-example

Say your mission demands ~18° HFOV on a 640-wide array and you care about human detection at 400–500 m.

• With 17 µm, the sensor width is 640×17 μm=10.88 mm. The 18° HFOV needs f≈ .

• With 12 µm, the width is 7.68 mm. For the same 18° HFOV, f≈24.3 mmf.