Detector vs. Optical Resolution

Most industrial infrared cameras use uncooled microbolometers to measure object temperatures or visualize distant targets. Advances in production methods, particularly wafer-level vacuum packing for thin-layer bolometers, have significantly reduced manufacturing costs, supporting wider adoption in industrial applications. Since the active area determines how many devices can be fabricated on a single silicon wafer—and wafer processing costs remain relatively constant—the active area is a key factor in the cost per device. This has driven a long-standing trend towards smaller detector pitches in bolometer technology.

Smaller detectors allow higher packing density in wafer micromachining, reducing costs compared to larger detector formats. However, pixel size remains critical for performance. Increasing pixel count while reducing pixel pitch decreases the effective absorption area, which can reduce sensitivity. Reducing bolometer size also reduces lens assembly size and cost, which is now comparable to the detector cost and significantly affects overall system pricing.

The pixel pitch—the center-to-center distance between pixels—affects resolution, thermal sensitivity, and image quality. Resolution refers to the number of pixels in the horizontal and vertical directions. Pixel size directly influences sensitivity; larger pixels have higher sensitivity due to their greater surface area, assuming identical material properties.

Why Smaller Pixels Don’t Measure Better

The Airy disk represents the smallest spot to which a beam of light can be focused. Diffraction imposes a fundamental limit on this spot size. As the lens diameter D approaches the wavelength of the light, the Airy disk becomes significantly larger, and the aperture behaves increasingly like a pinhole. Resolution can be improved by using shorter wavelengths or increasing the lens diameter. However, even in lens systems with negligible aberrations, diffraction-induced spreading defines an absolute limit to image quality.

Figure 2 shows a top view of bolometers with different pixel pitches, assuming identical optics. Because the diffraction limit of the optics remains the same, the Airy disk has the same diameter, requiring more pixels to capture the total power of the incident infrared light. In practice, lens design constraints and manufacturing tolerances in lens elements or assemblies further restrict the smallest achievable spot size, reducing both resolution and contrast.

The German standard VDI/VDE 5585-1 defines the minimum field of view for temperature measurement, which is also referred to as the measurement field of view as the field of view MFOV of the smallest circular target at which the radiance-proportional signal has dropped to a specified fraction when an iris diaphragm is progressively closed; the chosen fraction must be stated and is typically 90 %, together with the measuring distance, the lens used, and the reference radiator temperature. Usually, MFOV is defined in mrad. In practical applications, engineers discuss the minimal measurable object size at a given distance, derived from the MFOV value in mrad.

Practical Example: How many pixel do you need to trust the temperature of a thermal camera.

In the following, three infrared cameras with pixel pitches of 34 µm, 17 µm, and 12 µm are compared with each other, each using comparable optics with an f-number close to 1 to isolate detector-related effects. Table 1 compares their characteristics.

Table. 1. Overview of infrared camera parameters

This experiment investigates the encircled energy, calculated by first measuring the total energy on the image plane and then comparing it to the energy captured within smaller image areas. The focal plane array is illuminated by a large blackbody so that each pixel receives full intensity, with the focused image plane significantly larger than a single pixel. An adjustable aperture placed between the infrared camera and the constant infrared source reduces the image size.

The standard’s test method places a focused camera in front of a homogeneous reference radiator with a high signal-to-noise ratio, uses a temperature-regulated adjustable iris opened. This experimental setup uses a blackbody at 100 °C with a variable-diameter aperture mounted in front. When the aperture is sufficiently large, a centered pixel receives the maximum possible energy without obstructing the optical path. Reducing the aperture radius decreases the collected energy, as diffraction causes the point spread function to spread light across multiple pixels. Encircled energy is expressed as the ratio of captured intensity to total power, normalized to a range from zero to one.

In pyrometry, the distance-to-spot size ratio is defined as the diameter of a blackbody at which the detected radiation signal drops by 10% compared to that from a sufficiently large blackbody. This definition accounts for spherical and chromatic aberrations in optics, the detector’s active area, and scattering effects within the optical system. Applying a similar concept to infrared cameras, the image size at 90% encircled energy defines the measurement field of view. This size also indicates how many pixels are needed to fully cover the focused image. When the image size is smaller than this threshold, the encircled energy for a single pixel drops enough that intensity-based temperature measurements—such as those in conventional infrared cameras—underestimate the true value. Figure 6 presents the experimental results graphically, and Table 2 lists the corresponding image sizes at 90% encircled energy.

Table. 2. Image size at 90% encircled energy compared for different pixel pitches. This summary reveals that more pixels must be considered for infrared cameras with small focal plane array pitches for accurate intensity or temperature measurements.

At a blackbody temperature of 100 °C, a ~10% change in detected energy corresponds to a temperature deviation of 7.2 °C. Figures 7 and 8 present the results in the temperature domain using a line scan across a small point source.

In Figure 7, the point source is reduced so that only a single pixel is illuminated. This introduces a measurement deviation for each camera, with the largest deviation occurring in the model with the smallest pixel pitch.

In Figure 8, the object size is increased to the optical minimum, and multiple neighboring pixels are illuminated. In case of a bigger pixel pitch of 34 µm, only 2x pixels need to be illuminated, while for 17 µm, 3 pixels, and for 12 µm, 4x pixels are illuminated, as the complete airy projection needs to be covered by the detecting elements for accurate intensity measurement. Therefore, the smallest measurable feature or point source diameter in the image is determined by the optical characteristics of infrared imaging rather than by pixel resolution.

In a practical scenario where the target is positioned 1 m from the thermal cameras and comparable optics are used to achieve a similar field of view, measurement performance differs noticeably between models. Although the instantaneous field of view corresponds to 0.94 mm at this distance, each camera reports a different temperature for an object matching the IFOV size. Table 3 summarizes the results. Ignoring cost considerations, the 17 µm pixel pitch camera offers the best compromise in this example, as it provides the smallest measurement field of view.

Table. 3. Measurement results of a camera with different pixel pitch.

Beyond Pixel Count: Optical Limits Governing Reliable Infrared Thermometry

Advancements in bolometer performance open new application possibilities, especially where very small measurement points must be positioned flexibly and where thermal contrast is low over large relative distances. The cost of an infrared camera will remain a key factor in determining adoption. The ongoing trend toward smaller pixel sizes in infrared cameras for temperature monitoring, along with the associated optical limitations, challenges the accuracy requirements of small thermal camera concepts. While smaller pixels offer cost benefits by enabling higher packing densities on a wafer, maintaining sensitivity becomes more difficult. From a physics perspective, diffraction-limited behavior of the optical system imposes a fundamental cap on resolution and image quality, even as pixel density increases. With pixel dimensions approaching the detected wavelength, optical resolution is inherently constrained.

Modern systems often apply digital contrast-enhancement techniques to improve perceived image quality, but these do not address the underlying physical limits and can introduce additional noise. In the long-wavelength range commonly used for temperature monitoring, it is critical to recognize that single-pixel temperature readings are unreliable. This is because the diffraction-induced Airy disk is larger than the pixel size, leading to significant loss of encircled energy for individual pixels.

For microscopic applications, the Abbe diffraction limit constrains the ability to resolve fine infrared details. Both detector and optical limitations ultimately determine whether a camera can provide consistent, repeatable temperature measurements regardless of target size or distance.