Shenzhen Link-Card Tech Co., LTDMenu

Неохлаждаемые матричные детекторы LWIR: принцип работы и структура пикселя

Технический блог · Принцип работы детектора

Неохлаждаемые матричные детекторы LWIR: принцип работы и структура пикселя

Опубликовано 2026-09-03 · Link-Card Technology
Текст статьи доступен на английском языке.

The uncooled infrared focal plane array detector is the heart of a thermal imaging core module. It operates near room temperature and requires neither liquid nitrogen nor a Stirling cooler. Following the signal chain, this article explains how it converts 8–14 μm radiation into a digital readout — and where the material, the microbridge, and the readout circuit each run into their limits.

1. What is an uncooled focal plane array detector

An uncooled focal plane array (UFPA) detector operates near room temperature and does not cool its sensing elements down to the 80 K range. Cooled photon detectors offer higher sensitivity and microsecond-scale time constants, but they carry a dewar and a cryocooler. The uncooled approach is an order of magnitude lower in power consumption, size, and cost, which is why handheld, security, and industrial temperature-measurement cores mostly use it. Power-up still does not deliver a stable image immediately: non-uniformity correction is required after power-on, and a shutter-based core has to wait for one shutter cycle first.

The signal chain is a thermal effect, not a photoelectric one: scene radiation passes through the lens onto the focal plane, the pixel absorbs it and warms up, the resistance of the thermosensitive film changes, and the readout circuit converts that resistance change into a voltage, which is then output through an ADC. The four steps, in order:

  1. Infrared radiation is incident on the pixel and absorbed as heat;
  2. the pixel temperature rises, and the resistance of the thermosensitive material changes accordingly;
  3. the readout circuit detects the voltage / current change caused by the resistance change;
  4. the analog signal is digitized by the ADC and passed on to image processing and image formation.

The scene temperature difference and the bridge temperature rise are not the same number. Under f/1 optics, the thermal conversion efficiency of a microbridge is typically around 1%: a 1 °C change in the scene raises the bridge by only about 10 mK, and designs with lower thermal conductance can reach roughly 20 mK. The readout circuit has to resolve the resistance change on top of this small temperature rise and convert it back to a scene temperature, so the noise corresponds to tens of mK. The NETD (noise equivalent temperature difference) on a datasheet is scene-referenced — a rating such as ≤35 mK @25 °C, f/1.0 refers to the scene temperature difference, not to the bridge actually rising 35 mK.

"Uncooled" only means there is no cryocooler. A 1 °C change in substrate temperature typically shifts the output by far more than the NETD. Under direct sunlight outdoors, or as the housing warms up, the image drifts. Temperature measurement and high-stability imaging depend on a TEC to hold the focal plane steady, or on lookup-table compensation over temperature; both approaches have to be designed into the system separately — do not assume the detector stays stable on its own.

2. Mainstream thermosensitive materials compared

The core of an uncooled detector is its thermosensitive material. Three types dominate today: vanadium oxide (VOx), amorphous silicon (a-Si), and thermopiles based on the thermoelectric effect. Each has its own emphasis in temperature coefficient of resistance (TCR), noise behavior, and process maturity, as summarized in the table below.

Table 1. Characteristics of the three mainstream uncooled thermosensitive materials
Material TCR 1/f noise Process maturity Typical applications
Vanadium oxide (VOx) −2% to −3%/K Generally low High; the mainstay of 12 μm volume production in China Security surveillance, industrial temperature measurement, outdoor observation
Amorphous silicon (a-Si) −2% to −3%/K Relatively high in early devices; greatly improved in newer films High; CMOS-compatible Temperature measurement, security, large-format arrays
Thermopile None (thermoelectric effect) Low Medium Spot temperature measurement, gas detection

At the same TCR of −2%/K to −3%/K, early VOx showed lower 1/f noise than a-Si and therefore held a clear NETD advantage. That gap has narrowed in current production parts: Lynred's a-Si device PICO640S rates NETD <30 mK (f/1, 300 K, 30 Hz, with a 12 ms thermal time constant), in the same tier as the 30–50 mK of mainstream VOx. a-Si still stands out for uniformity, CMOS compatibility, and scaling to large-format arrays. When selecting a device, read NETD under matched test conditions — do not rank thermal sensitivity by material name.

A high TCR does not automatically yield a lower NETD. However large the temperature coefficient of resistance, if 1/f noise rises along with it, the signal-to-noise ratio can end up worse. Evaluate TCR, noise spectrum, fill factor, thermal conductance, and readout noise within one and the same chain. Mind the sign as well: a microbolometer's TCR is negative — temperature rises and resistance falls, opposite to metals. This sets the polarity of the output voltage under current bias; wiring the polarity backwards turns white-hot into black-hot.

3. Pixel microstructure

The pixel is the smallest imaging unit of the detector. A typical uncooled pixel uses a microbridge structure, suspended above the silicon substrate on slender legs to achieve thermal isolation. Its layered structure is shown in the figure below.

Uncooled FPA detector pixel structure Infrared radiation Infrared absorber layer Thermosensitive layer (VOx / a-Si) Support leg Support leg Electrode Electrode Readout circuit (ROIC) Silicon substrate Microbridge
Figure 1. Layered structure of an uncooled FPA pixel (microbolometer)

From top to bottom, the pixel structure consists of:

4. Key design parameters and performance trade-offs

Pixel performance is set by several parameters together. Common pixel pitches are 17 μm and 12 μm, with 8 μm already available in volume-production models; the smaller the pitch, the denser the sampling at the same field of view, and the smaller the die at the same resolution. The fill factor is the ratio of absorber area to pixel area — the more area the legs and gaps occupy, the lower the fill factor. The thermal time constant τth = C/G sits mostly in the 8–15 ms range for production devices. Thermal conductance G is determined by leg length, cross-section, and leg material, and it enters both NETD and response speed.

Sensitivity and speed are pinned to the same set of thermal parameters. Reducing G enlarges the temperature rise and improves NETD — but lengthens τth. A 60 Hz frame period is 16.7 ms. When τth approaches or exceeds the frame period, moving targets leave smear between adjacent frames. Pushing the frame rate beyond 60 Hz usually requires shorter or wider legs, and NETD degrades accordingly. This is a structural trade-off, not something backend gain can recover.

Engineering practice adds one more point: mechanical reliability takes priority over ultimate sensitivity. Making the legs excessively thin in pursuit of a high fill factor can cause the microbridge structure to collapse under process stress. It is advisable to verify leg mechanical strength with finite-element simulation already at the layout design stage — for a volume-production device, yield and reliability usually matter more than paper sensitivity.

5. Array formats and readout modes

Individual pixels form an array before they make a complete focal plane. Uncooled detector array sizes range from 80×60, 256×192, 384×288, and 640×512 all the way to 1280×1024 — the higher the resolution, the richer the image detail. There are two main readout modes:

The thermal time constant is already on the order of 10 ms, and the readout mode adds a further time difference on top. Snapshot integrates the whole array at the same time and then reads out row by row, so a moving target stays positionally consistent within one frame. With rolling integration, each row's integration window is offset by one row time, and a heat source crossing the scene quickly ends up displaced from row to row. Most production uncooled detector datasheets specify rolling readout, not full-array simultaneous integration. When you see "rolling", do not estimate motion smear as if the device were a snapshot one.

Conclusion

The core chain of an uncooled infrared focal plane array detector fits in one sentence: infrared radiation heats the pixel, the resistance of the thermosensitive element changes, and the readout circuit detects the change and forms an image. Understanding this chain — and the trade-offs among material, pixel structure, and key parameters — gives a clearer view of the physics and engineering behind a thermal imaging core module's NETD, resolution, and frame rate, and lays the groundwork for drive-timing design, non-uniformity correction (NUC), and temperature compensation.

Назад в технический блог

Нужна помощь с выбором или индивидуальным решением?

Link-Card помогает с выбором модулей, интеграцией и комплексными решениями.

Наш отдел продаж ответит на указанный email или телефон. Также можно написать напрямую на sales@link-card.com.