Overview
Infrared radiation propagating through the atmosphere is attenuated by absorption and scattering from gas molecules, water vapor, and particulate matter such as dust. The introduction of noise sources at varying levels can further increase attenuation during transmission and affect sensor performance.
According to MemsConsulting, a research team at Qingdao Zhiteng Microelectronics Co., Ltd. recently published an article titled "Study on the Influence of Noise on Infrared Sensor Performance" in the journal Sensors and Microsystems.
This work experimentally introduced three different levels of noise into a designed infrared sensor system and compared the collected infrared images using host PC software.
Principles
Theoretical basis
An infrared sensor typically consists of an optical system, a detector, signal conditioning circuitry, and a display unit. An infrared source is any object emitting infrared radiation. An infrared detector converts incident infrared radiation into electrical energy. The detector is the core component of the infrared sensor and, together with an imaging processing chip, provides non-contact temperature information of the target.
Key performance metrics for infrared detectors include responsivity, response time, resolution, and the imaging chip's digital signal processing rate and frame loss behavior.
According to Kirchhoff's law, objects that are strong radiators are also strong absorbers. At a given temperature, a blackbody has the greatest radiative power. A blackbody fully absorbs incident radiation at all wavelengths and emits thermal radiation more strongly than any real object. Idealized models such as blackbodies, gray bodies, and white bodies do not always match real objects, and the introduction of noise sources can alter real radiative characteristics. Therefore, practical characterization often requires experiments and can be complex.
Infrared sensor noise test system
This study focuses on how introduced noise sources affect infrared sensor performance. Under infrared illumination, the detector's sensitive elements generate charge carriers; downstream circuitry collects and amplifies these carriers.
The infrared calibrator used in the experiment served as the calibrated target, providing a reference temperature over a calibration range of -20 to 300 °C. At a calibration distance of 1 m, the calibrator's accuracy reaches 0.3 °C. The central circular area was used as the blackbody target in the experiment.
The infrared sensor is the core of an imaging system, offering high resolution, simple structure, and high frame rate. The test system used in this study employed a 320 K blackbody as the source, a detector sensitive area of 1 cm^2, and a noise-equivalent bandwidth of 1 Hz.
In the system, the radiation source is any object with a temperature-dependent emission; the detector absorbs infrared radiation and changes temperature. Detection is achieved by measuring physical property changes in the sensitive element caused by temperature variation. The small-signal processing stage converts collected analog electrical signals into digital signals. An STM32-based control system converts parallel digital data into image data and outputs the image to an OLED display circuit.
The aim of this study is to investigate the impact of noise on infrared sensor performance. Noise is a random disturbance superimposed on the desired signal. Image noise originates from various components of the imaging chain, including the infrared sensor, readout circuitry, and signal processing. Infrared sensor noise is often the dominant source in the final image. Noise mechanisms are complex and include thermal noise, shot noise, and photon noise.
Analysis and discussion
During the experiment, the infrared detector was well grounded. An oscilloscope probe measured a low noise condition with a peak of 32.5 mV and a frequency of 53.95 MHz. Imaging observed via host PC software showed a clear image with no noticeable artifacts, indicating negligible impact on sensor performance at this noise level.
When the metal enclosure of the infrared sensor was removed, the oscilloscope measured a system noise peak of 200.0 mV. Host PC observation showed blocky artifacts in the image and degraded imaging quality. This indicates that a significant increase in system noise reduces the sensor's electrical performance; the measured disturbance in this case behaved as an interference signal rather than ground noise.
With the sensor enclosure removed and exposed to a noisy external environment, the measured system noise peak reached 310.0 mV. Host PC observation revealed severe image degradation marked by green screen and corrupted-pixel patterns. The experiment shows that once noise interference reaches a certain intensity, the infrared sensor performance is seriously impaired and normal operation is no longer possible.
To investigate the underlying cause, the detector datasheet was used to obtain the standard digital signal waveform before STM32 processing. The standard waveform includes a frame or row signal, CMOS communication data lines, and a clock signal.
After STM32 processing and emulator simulation, the theoretical sensor data waveform was obtained as shown below.
With the system noise peak at 310.0 mV, the oscilloscope captured the detector's actual output waveform. The captured waveform shows the data line and the clock signal.
Figure 1. Actual output waveform of the infrared detector under high noise
Comparing the simulated theoretical waveform and the actual waveform under high noise reveals bit shifts and logic errors in the detector output. Bits that should be 1 become 0 and vice versa, causing corrupted data to be passed to the STM32. This results in frame loss and incorrect data bits. Since the displayed image is formed from RGB channels, errors in received RGB data produce mixed or corrupted pixel colors. As a result, visual artifacts such as corrupted-pixel patterns and green screens appear on the image.
Conclusion
The experiments indicate that weak external noise does not significantly affect infrared sensor performance. However, as noise strength increases beyond a threshold, sensor performance degrades and imaging quality is severely affected. The primary cause identified is that strong noise induces bit shifts or logical state errors in the detector output. These erroneous data are then processed by downstream circuitry and manifest on the image as corrupted-pixel patterns or green screens.
Future research should focus on: 1) methods to separate noise from useful digital signals and mitigate noise impact on sensor performance; and 2) developing noise models through extensive experiments and applying algorithmic optimization to improve infrared sensor performance.
