Choosing the Thermal Channel — Uncooled LWIR or Cooled MWIR

Almost every enquiry we receive starts in the wrong place. It starts with a model number. The decision that actually governs the system sits one level above that: which thermal channel the mission needs. Get that right and the rest of the selection — positioner, mast, interface — follows from it.

The two bands, and why the choice is not free

A thermal imager detects emitted radiation, not reflected light. Two atmospheric windows are usable: the long-wave band around 8–14 µm and the mid-wave band around 3–5 µm. Both work in complete darkness. They do not behave the same way.

Uncooled LWIR uses a microbolometer array that operates at ambient temperature. Nothing has to be cooled, so the unit draws less power, starts in seconds, has no cryocooler to wear out, and costs materially less. In the range this is the SEG10, SEG12 and SEG14 class, with 640 × 512 detectors at 12 or 17 µm pitch and NETD figures of 30–50 mK.

Cooled MWIR uses a photon detector held at cryogenic temperature by an integrated cooler. That cooler is the reason the unit costs more, draws more and needs a cool-down period before use. What you buy with it is sensitivity: the SEB20 reaches 1280 × 1024 at 12 µm with NETD ≤ 20 mK, against ≤ 50 mK for the uncooled units. Lower NETD means smaller temperature differences become visible, which is what extends useful range.

Three things thermal imaging does not do

Before comparing bands it is worth removing three assumptions, because they cost more programmes than any specification error.

Thermal does not see through fog. It sees through smoke, dust and light haze, because those particles are far smaller than the infrared wavelength and scatter it only weakly. Fog and cloud are made of water droplets whose diameter is comparable to that wavelength, and they scatter infrared about as effectively as they scatter visible light. In dense fog a thermal imager is not in a fundamentally better position than an eye. The rule worth carrying is that thermal beats the visible band when the obscurant is fine, not when it is wet.

A thermal image is not a temperature map. It is a map of emitted radiation, and emitted radiation is surface temperature multiplied by emissivity, plus whatever the surface reflects from its surroundings. A painted engine cover glows. A polished metal panel at the same temperature reads cold, because its emissivity is low and what it is showing you is a reflection of the sky. This is why wet asphalt can appear colder than the air above it, and why the shiniest surface on a target is often the one that disappears.

Contrast is not constant through the day. Around sunrise, and again after sunset, a target and its background pass through the same apparent temperature on their way up or down. For that period thermal contrast collapses towards zero, and a system that worked all night sees very little. The effect is called thermal crossover. It cannot be engineered away, only planned around, and it is one of the honest reasons a thermal channel is paired with a daylight channel rather than replacing it.

Where each one wins

The honest summary is that LWIR wins on cost of ownership and MWIR wins on range and on discrimination. The exceptions are worth knowing.

• Humid, hazy, smoky air. LWIR generally penetrates better. Water vapour absorption is significant in the mid-wave band, so a coastal or tropical site tends to favour long-wave.
• Long range against a small or cool target. MWIR wins. The combination of higher detector resolution and lower NETD is what separates a person at the edge of a field from the field itself.
• Low sun angles over water, glass or wet metal. LWIR gives the quieter picture. The mid-wave band still carries reflected sunlight, so a cooled imager can pick up solar glint that a long-wave imager never sees. The MWIR advantage is real, but it is not evenly distributed across the day.
• Continuous, unattended operation. LWIR wins. A cryocooler has a finite service life measured in thousands of hours, and it is the part that will eventually need attention on a mast that never comes down.
• Rapid response from cold. LWIR wins. An uncooled unit is producing usable imagery almost immediately; a cooled unit needs its cool-down.
• Hot climates. Note which way the penalty runs. A cooler has to pull the detector down against ambient, so its longest cool-down and its heaviest duty fall on the hottest day, not the coldest. Cool-down time is a summer specification.
• Power-constrained platforms. LWIR wins, and often decisively. On an unmanned vehicle or a battery-backed remote site the cooler’s draw is the argument, not the price.

Read the whole chain, not the detector

The detector is one term in the equation. The others matter as much:

Pixel pitch. A 12 µm pitch packs more pixels behind the same optic than 17 µm. It improves sampling of the scene, but it also means each pixel collects less energy, which the detector design must compensate for. Compare NETD and pitch together; neither on its own tells you what the image looks like.

Optics and field of view. Range is a function of focal length as much as of the detector. A narrow field of view puts more pixels on a distant target and makes it harder to find in the first place. This is why the range pairs a narrow thermal field with a wide-field EO channel at 1920 × 1080 and 30× optical zoom: you search on one and identify on the other.

Which range is being quoted. A range figure means nothing until you know which task it refers to. Deciding that something is there, deciding what class of thing it is, and deciding which particular thing it is are three separate distances against the same target, and they fall in that order by a wide margin. Two datasheets can print the same number and describe performance a generation apart. Ask which of the three, against what target, in what atmosphere.

NETD versus MRTD. NETD is a detector figure measured under laboratory conditions. MRTD is a system figure that includes the optics and the display and correlates better with what an operator can actually resolve. Where both are published, MRTD is the more honest number.

A working method

1. State the target, the range and which of the three tasks you mean. “Recognise a person at 2 km” and “detect a vehicle at 8 km” are different systems. Write it down before looking at products.
2. State the environment. Humidity and typical haze push toward LWIR. Clear, dry, long sightlines push toward MWIR. Fog pushes toward accepting that no band solves it and planning a second means of coverage.
3. State the duty cycle. Continuous unattended watch penalises the cryocooler.
4. State the power available, and the highest ambient temperature at which it has to be available. This constraint decides more selections than performance does.
5. Only then compare models, and compare on NETD, detector resolution, pixel pitch and field of view together.

What we would ask you

When a request reaches us with the target, the range, the environment and the power budget stated, we can usually name the right channel in one reply. When it reaches us as a model number, the first thing we do is ask these questions anyway. Sending them with the enquiry saves a round trip.