Introduction: When Display Capabilities Surpass Measurement Instruments#
In the past decade, display technology has undergone a profound dynamic range revolution. OLED panels have achieved theoretically infinite contrast through pixel-level self-luminescence and True Black; Mini-LED backlighting has pushed LCD contrast to the magnitude of hundreds of thousands to one through thousands of independent Local Dimming Zones. These technical advances have brought a leap in visual experience for consumers, but simultaneously pose severe challenges to the field of display metrology.
Traditional imaging colorimeters and photometers were originally designed for LCD panels with contrast in the range of 1000:1. When the dynamic range of the object being measured leaps to 100,000:1 or even 1,000,000:1, the core conflict facing the measurement system becomes acute: how to accurately capture optical information from both extremely bright areas (peak luminance can exceed 1000 nits) and extremely dark areas (black levels can be below 0.005 nits) in the same frame?
This article will analyze the specific challenges posed by High Dynamic Range (HDR) displays to measurement, explain the principles of Multi-Exposure technology, and discuss the critical impact of sensor noise on low-luminance measurement.
I. Measurement Requirements of HDR Display Technology#
1.1 Contrast Characteristics of OLED and Mini-LED#
Understanding the measurement challenges requires understanding the characteristics of the measured objects.
OLED (Organic Light Emitting Diode) displays have pixels that emit light independently. When a pixel needs to display black, its current is completely cut off, and the pixel emits zero photons. Under ideal darkroom conditions, the luminance of black areas on an OLED is near zero, making its Dark Room Contrast Ratio theoretically approach infinity. In actual measurement, the On/Off Contrast Ratio of OLED panels can easily exceed 1,000,000:1.
Mini-LED backlit LCDs achieve fine local dimming by arranging hundreds or even tens of thousands of miniature LEDs into a backlight matrix. When an area needs to display dark content, the corresponding Mini-LED backlight area reduces brightness or turns off completely. This raises the contrast of Mini-LED LCDs from approximately 1000:1 for traditional edge-lit LCDs to over 100,000:1.
The common feature of these two technologies is that areas with luminance differing by five or six orders of magnitude may exist simultaneously in the same display frame.
1.2 Fundamental Conflict Faced by Measurement Systems#
The core sensor of an imaging colorimeter—whether CCD or CMOS—has a limited Dynamic Range. The dynamic range of a sensor is determined by the ratio of its Full Well Capacity to its Read Noise, usually expressed in decibels (dB) or stops/f-stops.
A typical scientific-grade CCD sensor has a single-exposure dynamic range of approximately 60-72 dB (about 1000:1 to 4000:1). High-performance scientific-grade CMOS sensors can reach about 75-80 dB. In contrast, HDR display panels require a measurement system to cover a dynamic range that may reach 100-120 dB (100,000:1 to 1,000,000:1).
This means that any single exposure cannot simultaneously and accurately record the luminance values of the brightest white and the darkest black on an OLED panel. If the exposure time is set to suit bright areas (e.g., 1000 nits), the signal in dark areas (e.g., 0.005 nits) will be submerged in sensor noise and cannot be effectively resolved. If the exposure time is extended enough to capture signals in dark areas, pixels in bright areas will be severely saturated, outputting truncated and invalid data.
II. Multi-Exposure (HDR Capture) Technology#
2.1 Basic Principle#
Multi-exposure technology is the primary engineering means currently used to resolve the aforementioned conflict. The core idea is to capture multiple images of the same scene with different exposure times and then intelligently synthesize these images in software, thereby constructing a measurement image with an effective dynamic range far exceeding the single-exposure limit.
A typical multi-exposure workflow is as follows:
Short Exposure Capture: Capture an image with a short exposure time (e.g., 1ms-10ms). At this point, pixels in high-brightness areas are within the sensor’s linear response range and can be accurately quantified; however, signals from pixels in dark areas are weak, the Signal-to-Noise Ratio (SNR) is extremely low, and measurement results are unreliable.
Medium Exposure Capture: Capture another image with a medium exposure time (e.g., 100ms-500ms). Medium-brightness areas obtain an ideal SNR at this exposure; high-brightness areas may start to approach saturation, and dark-area signals improve but remain insufficient.
Long Exposure Capture: Capture a third image with a long exposure time (e.g., 1s-10s or even longer). Pixels in dark areas accumulate enough photon signals to be distinguished from noise; however, pixels in bright areas are severely saturated.
Image Synthesis: Software algorithms synthesize three (or more) images into one HDR measurement image. The basic logic of the synthesis strategy is: for each pixel position in the frame, select the data from the exposure image that has the optimal SNR and is not saturated at that position, then perform weighted synthesis or direct replacement.
2.2 Key Considerations in Synthesis Algorithms#
Multi-exposure synthesis is not simple image stitching; it involves several technical details:
Precise Calibration of Exposure Times: The quantitative relationship between images of different exposure times depends on the accuracy of the exposure time ratio. If there is a deviation between actual exposure time and the nominal value (such as shutter timing error), the synthesized absolute luminance values will have systematic errors.
Sensor Linearity Correction: The photoelectric response of CCD/CMOS sensors is not perfectly linear across the entire dynamic range. Near the saturation region, the response curve usually compresses (i.e., the actual number of photons increases but the output signal increment decreases). Synthesis algorithms need to pre-calibrate the sensor’s linear response curve and exclude or correct data from non-linear regions.
Alignment and Registration: If a slight displacement (such as mechanical vibration) occurs in the measured object or the imaging system during multiple exposures, pixel positions between different exposure images will shift. Sub-pixel level image registration is required before synthesis. For display panel measurement, this problem is relatively controllable since the object is usually fixed and the measurement system is mounted on a stable stand.
Identification and Exclusion of Saturated Regions: The algorithm needs to reliably identify saturated pixel regions in each image and exclude these regions during synthesis. The setting of the saturation threshold needs to consider the sensor’s full well capacity and the bit depth of the A/D converter.
2.3 The Cost of Multi-Exposure#
Multi-exposure technology is highly effective in extending dynamic range, but it comes with practical costs:
Increased Measurement Time: Capturing multiple images with different exposure times, especially long-exposure images (which may take several to dozens of seconds), significantly increases the total duration of a single measurement. This can be a bottleneck in scenarios with strict Takt Time requirements.
Increased Data Processing Load: Synthesis of multiple high-resolution images requires significant computational resources. For sensors in the 29MP or 61MP range, the data volume per image is already substantial, and the computational burden of synthesis is even more significant.
Limitations for Moving Targets: If the displayed content changes during the multiple exposures (such as dynamic content), the content across different exposure images will be inconsistent, resulting in artifacts in the synthesized result. Therefore, multi-exposure technology typically requires the measured content to remain static throughout the capture process.
III. Signal-to-Noise Ratio (SNR) and Low-Luminance Measurement#
3.1 Sources of Noise#
Before discussing low-luminance measurement capability, it is necessary to understand the main sources of noise in imaging sensors:
Photon Shot Noise: This is noise inherent to the randomness of photons arriving at the sensor (Poisson statistics). Its magnitude is proportional to the square root of the signal: N_shot = sqrt(S), where S is the number of signal photons. This is a fundamental physical limitation that cannot be eliminated by sensor design.
Dark Current Noise: Even without light, silicon material in the sensor generates electrons due to thermal excitation. The statistical fluctuations of these thermally generated electrons constitute dark current noise. The magnitude of dark current depends strongly on sensor temperature—for every 5 to 9 degrees Celsius decrease in temperature, the dark current is halved (this characteristic is called the “Doubling Temperature”).
Read Noise: Electronic noise introduced by the readout circuit itself each time a pixel signal is read from the sensor. Read noise is an inherent characteristic parameter of the sensor, independent of signal strength and exposure time.
Fixed Pattern Noise (FPN): Spatial fixed noise caused by differences in sensitivity and dark current non-uniformity between sensor pixels. FPN can be eliminated to some extent through Dark Field Subtraction and Flat Field Correction.
3.2 Bottleneck in Low-Luminance Measurement#
When measuring low-luminance areas (such as the black areas of an OLED panel with luminance < 0.005 nits), the signal is extremely weak. At this point, the Signal-to-Noise Ratio (SNR = Signal / Noise) depends on the magnitude of the noise.
Under low-signal conditions, photon shot noise is also small (because the signal itself is small). At this point, dark current noise and read noise become the dominant noise sources. If the dark current noise is greater than the signal itself, the measurement results will be completely submerged in noise, and meaningful information cannot be extracted.
To understand more intuitively: suppose a pixel collects 5 signal electrons from a measured dark area during a long exposure (e.g., 10 seconds), but 20 thermal electrons are generated by dark current during the same time—then the proportion of the total signal (25 electrons) that truly comes from the measured object is only 20%, and due to the presence of shot noise, a single measurement result could deviate far from the true value.
3.3 Cooled Sensors: The Core Means of Reducing Dark Current#
Based on the strong dependence of dark current on temperature, cooled CCD/CMOS sensors have become a key technology for high-precision low-luminance measurement.
Thermoelectric Cooling (TEC): Active cooling of the sensor using the Peltier Effect. Typical cooled imaging colorimeters stabilize the sensor temperature within the range of -10°C to -30°C. Compared to operating a sensor at room temperature (about 25°C), cooling to -10°C can reduce dark current by approximately one order of magnitude; cooling to -25°C can reduce it by about two orders of magnitude.
Another important effect of cooling is temperature stability. During continuous operation, the temperature of an uncooled sensor gradually rises due to heat from electronic components, causing dark current to increase accordingly—this leads to drift in measurement data within a work shift. Cooling systems lock the sensor temperature at a constant value, eliminating this temperature drift effect and ensuring the stability and repeatability of measurement data during long-term operation.
3.4 Quantitative Metrics for Dark Field Noise#
When evaluating the low-luminance measurement capability of an imaging colorimeter, the following metrics are valuable references:
Noise Equivalent Luminance: The equivalent luminance value corresponding to sensor noise under specific exposure conditions. When the measured luminance is lower than the noise equivalent luminance, the signal cannot be distinguished from noise. The lower this value, the stronger the instrument’s low-luminance measurement capability.
Dark Field Uniformity: The spatial distribution uniformity of dark current across pixels in images captured under no-light conditions. Residuals of dark field non-uniformity after correction directly affect the spatial accuracy of low-luminance measurements.
IV. Measuring the Extremes Simultaneously: A Complete HDR Measurement Workflow#
4.1 Typical Application Scenarios#
Taking production line inspection of OLED TV panels as an example. A typical OLED panel displaying HDR content may simultaneously contain in the same frame:
- Peak highlight areas: luminance exceeding 800-1000 nits
- Mid-tone areas: luminance between 10-200 nits
- Dark detail areas: luminance between 0.01-1 nit
- Pure black areas: luminance < 0.005 nits
The requirement for production line inspection is to accurately measure luminance and color coordinates across all the aforementioned luminance ranges to detect luminance uniformity, chromaticity uniformity, Mura defects, and bright dots on dark backgrounds.
4.2 System Configuration Requirements for HDR Measurement#
To achieve accurate measurement covering the aforementioned complete luminance range, a measurement system needs to satisfy the following conditions:
Sensor: Cooled CCD or scientific-grade CMOS, with operating temperature stabilized at a low state to ensure controllable dark current during long exposures. The sensor’s full well capacity should be as large as possible (e.g., > 100,000 e-) to obtain as wide a linear dynamic range as possible in a single exposure.
Optical System: The stray light (Stray Light/Veiling Glare) characteristics of the lens are critical. When measuring high-contrast scenes, bright-area light that scatters inside the lens and reaches pixel positions in dark areas will artificially raise the dark-area measurement values, leading to underestimated contrast measurement results. High-quality measurement lenses need multi-layer coatings and internal light-absorbing designs to suppress stray light.
Multi-Exposure Strategy: Design a reasonable exposure sequence based on the luminance range of the measured panel. The number of exposure steps (usually 3-5) and the exposure time for each step need to be optimized based on sensor characteristics and measured object characteristics.
Software Algorithms: Equipped with complete HDR synthesis algorithms, sensor linearity correction, Dark Frame Subtraction, and Flat Field Correction capabilities.
4.3 Validation of Measurement Results#
Validation of HDR measurement results also needs to focus on both ends of the dynamic range. At the bright end, comparisons can be made using standard light sources at known luminance levels; at the dark end, since reference instruments (such as spectroradiometers) also face SNR challenges at extremely low luminance, validation needs to be more cautious.
A practical validation method is to use Neutral Density Filters (ND Filters) with known transmittance to attenuate high-luminance light sources, creating light source pairs with known luminance ratios to verify the accuracy and linearity of the measurement system across different luminance levels.
V. Correct Interpretation of Dynamic Range Metrics#
5.1 Nominal vs. Effective Dynamic Range#
“Dynamic range” values specified by imaging colorimeter manufacturers in specification sheets should be interpreted with caution. It is necessary to distinguish between:
Sensor-level Dynamic Range: The ratio of sensor full well capacity to read noise under single-exposure conditions. This represents the theoretical limit of the hardware.
System-level Dynamic Range: The luminance range that the entire measurement system can effectively cover after processing such as multi-exposure and dark field correction. System-level dynamic range is usually far greater than sensor-level dynamic range.
Effective Measurement Dynamic Range: The luminance range that the system can actually cover provided that specific accuracy requirements (e.g., luminance measurement accuracy better than ±2%, color coordinate accuracy better than ±0.002) are met. Due to SNR limitations at the low-luminance end, effective measurement dynamic range is usually smaller than system-level dynamic range.
5.2 The Impact of Stray Light in Contrast Measurement#
It is worth emphasizing that in high-contrast measurement, lens stray light (Veiling Glare) often becomes a limitation for measurement accuracy earlier than sensor noise.
The effect of stray light is to superimpose a “Light Fog” proportional to the brightness of the scene’s bright areas across the entire image plane. Suppose an OLED panel has a luminance of 500 nits in a full white state and a true luminance of 0.001 nit in a full black state. If the lens stray light ratio is 0.1%, the equivalent luminance superimposed by stray light in the black area is 500 x 0.001 = 0.5 nit—this is 500 times higher than the true black level luminance. At this point, the measured contrast is only 1000:1, rather than the true 500,000:1.
Therefore, in the HDR era, the stray light characteristics of a measurement lens may be more critical than the noise level of the sensor. This also explains why high-end imaging colorimeters invest heavily in lens design—including the use of low-scatter glass materials, optimization of coating processes, and addition of internal light-blocking structures.
Conclusion#
The popularity of High Dynamic Range display technology has posed unprecedented challenges to the measurement capabilities of imaging colorimeters. The luminance range that a single exposure cannot cover needs multi-exposure technology to compensate; accurate measurement of low-luminance areas relies on cooled sensors for the suppression of dark current noise; and accurate characterization of high-contrast scenes is limited by the stray light levels of optical systems.
For quality engineers and measurement engineers in the display panel industry, understanding these challenges and the principles of their countermeasures is fundamental for correctly selecting measurement equipment, reasonably setting measurement parameters, and accurately interpreting measurement results in the HDR era. When evaluating the HDR measurement capability of an imaging colorimeter, one should not only focus on the nominal dynamic range digits but also gain a deep understanding of its sensor cooling capability, the reliability of its multi-exposure synthesis algorithms, and the stray light control level of its lens.
FAQ#
Q1: Why can’t a single exposure simultaneously measure bright and dark areas of an OLED panel?#
Because the single-exposure dynamic range of sensors is limited (typically 60-80 dB), while OLED panel contrast can reach 100-120 dB. If exposure time is set to suit bright areas (e.g., 1000 nits), dark-area signals will be submerged in sensor noise; if exposure time is extended to capture dark-area signals, bright-area pixels will be severely saturated. Multi-exposure (HDR) technology must be used to capture multiple images with different exposure times and synthesize them intelligently to cover the full luminance range.
Q2: How much does a cooled sensor help in low-luminance measurement?#
The help is very significant. Dark current has an exponential relationship with temperature—for every 6-8°C decrease in temperature, dark current is approximately halved. Cooling to -10°C can reduce dark current by about one order of magnitude, and cooling to -25°C can reduce it by about two orders of magnitude. Furthermore, cooling locks the sensor temperature at a constant value, eliminating dark current drift caused by operational heating and ensuring the stability and repeatability of measurement data over long durations.
Q3: In high-contrast measurement, which has a greater impact: lens stray light or sensor noise?#
In HDR display measurement, lens stray light often becomes a limiting factor earlier than sensor noise. For example, if the lens stray light ratio is 0.1%, when the bright area is 500 nits, the equivalent luminance superimposed by stray light in the dark area is 0.5 nit—this is 500 times higher than the true OLED black level (0.001 nit). The measured contrast would then be only about 1000:1, far below the true 500,000:1. Therefore, in the HDR era, the stray light control level of measurement lenses may be more critical than the sensor noise level.
This article is part of the Imaging Colorimeter Technology Knowledge Base series.