With an Eye on Applications
Driven by demands from mobile phone, display and specialty lighting equipment manufacturers, the need for sophisticated and accurate chip-scale color and spectral sensors has become stronger than ever. In this article, ams’ Kevin Jensen describes the types of optical sensors and detectors. He also provides ideas on evaluating the suitability of each type for specific applications.
The market for sophisticated and accurate chip-scale color and spectral sensors has grown significantly in recent years, driven by demand from mobile phone, display and specialty lighting equipment manufacturers. Optical semiconductor manufacturers have responded by developing distinct families of sensors that meet specific types of application requirements. This article describes the types of optical sensors and detectors commonly used today, providing ideas on evaluating the suitability of each type for specific applications.
THE SCIENCE OF COLORIMETRY
Colorimetry—the measurement of color—is growing in importance in certain consumer, medical, industrial and commercial applications. Color sensor ICs play a crucial role in enhancing the performance of a smartphone’s display and camera. These ICs are at the heart of a revolution in horticulture, enabling the output from specialty LED lights to boost yields in intensive “vertical farms” which grow food crops, such as lettuce, in tightly controlled conditions. And new applications for color sensing are continually emerging.
Early applications for color sensors ICs were based on the use of simple RGB (Red/Green/Blue) sensors. Today, the sensor or detector requirements often are more complex. Color measurement requires know-how on the part of the system developer.
This is, in part, because a human’s perception of color is not based solely on an absolute physical value such as electrical current or atmospheric pressure. There is a strong subjective or physiological element to the perception of color. This means that, while it is possible to derive by statistical means a standard for the “average” human’s perception of color, every person’s ocular physiology is different, and outliers in the human population have very different color perception from the average.
The acuity of a human’s perception of color has an impact on the requirement for a color sensor’s measurement accuracy. For this purpose, two standard models have been defined: the CIE1931 industry standard for human eye like color perception as well as separation of light/color into its individual spectrum.
In each case, every colorimetrical application requires the same basic system elements and sensor setup:
• Sensor, filter
• A target for calibration
The choice of light sources, the system’s operation and the characteristics of the filter determine the scope of the sensor module’s detection capabilities. The electronic circuitry in a sensor IC has an important effect on the quality of the sensor signals, and on its speed of operation.
There are differences in the capability and performance of various types of color measurement devices:
• XYZ or True Color sensors
• Multi-spectral sensor ICs
• RGB sensor ICs
COMMON DEVICE TYPES
Two types of device are commonly used for colorimetric applications. The first is the traditional spectrometer as a reference and calibration device and the second is a color sensor IC, which provides a low-cost way to achieve good or even excellent color measurement accuracy. A mini-spectrometer might also be an appropriate choice in some circumstances. To provide a meaningful comparison of the performance of these different device types, ams has performed measurement tests in application-based setups.
True Color sensors—example sensors: AS7261 and AS7221:
True Color sensors may be used for absolute-value color measurements. They use interference filters, which provide a sound technological basis for the measurement of color to color standards. These sensor ICs measure values as accurately as the human eye sees them (Figure 1).
The interference filters allocate specific sensitivity values to each wavelength per color channel. When calibrated, it is possible to render the measured color values as XYZ values (chromaticity co-ordinates), which are used as base values for conversion into other color spaces. XYZ coordinates are based on the CIE 1931 “Standard Observer” characteristics of the average human eye. The use of a True Color sensor IC makes it possible to describe in number values the color of fabrics or print products in the same way as would a human eye.
Multi-spectral sensors—example sensor: AS7341:
Multi-spectral sensors are next-generation sensors that use multiple channels to provide the maximum information output at a low price point. When color coordinates are not enough, the spectral composition of objects is measured. This principle can compensate for metamerism (false color matching). A multi-spectral sensor provides the answer to the question whether an orange color sample is a mix of red and yellow, or a pure orange. Multi-spectral sensors separate the chosen spectrum into spectral channels. The filters are arranged in such a way that their limiting ranges align, leaving almost no gaps in the chosen visible or near infrared (NIR) spectrum (Figure 2).
In the visible range, a multi-spectral sensor’s measurement takes place at the radiometric level rather than the colorimetric level. This means the sensor outputs the spectrum of the sample and determines the color point via these spectral values. In the NIR range, the measured spectrum can also be used to look at specific band-passes and chemical bonds to identify moisture, fats or proteins. The further the detection range in the NIR, even past the silicon range, the easier it is to determine specific substances.
RGB Sensors—example sensor: TCS3400:
Traditional RGB technology can be seen as a subset of spectral sensors. These consist of three band-pass filters in the visible light spectrum (Figure 3). The peaks of the spectral graphs are not set uniformly in relation to particular wavelengths, but are defined during the design process in response to the specification of the measurement task and cost.
This kind of color measurement is not aligned to any standard or model of the human eye’s perception of color. An RGB sensor can, nevertheless, be used in colorimetric tasks, depending on the required accuracy. But even with the application of complex calibration methods, the accuracy of an RGB sensor’s color measurement is limited to the 3 band-passes of information.
Mini-spectrometers are compact and sturdy sensor solutions that measure spectral values and support an interpretation of the color space. Their resolution is limited in comparison to laboratory-grade spectrometers but with fewer spectral scanning points, they are faster.
Comparison color sensor IC types is performed with one or multiple samples, which are measured and used as reference values. A limiting value is set for RGB (relative measurements) or colorimetrical XYZ value, such as ΔEL*a*b* (for absolute measurement).
To compare the various sensor and detector types, ams configured test setups based on actual applications. Certain LED luminaires or LED display backlights require illumination at a tightly specified color temperature, or at a specific color point. In addition, effects such as color shifts due to temperature drift or aging may need to be compensated for.
Color differences of ∆ u’v’ ≤0.005 can be seen by the average human eye. In fact, a trained eye can even perceive color differences as small as a ∆ value of 0.003. The test results in Table 1 describe the measurement results from RGB and True Color sensors when measuring a D65 white light source.
In our testing, two systems using a feedback control loop, one using an RGB sensor and one using a True Color sensor, were set up and calibrated at a temperature of 40°C (104°F). Next, the temperature of the LEDs was changed, producing a color drift. This drift was compensated by the feedback control loop. Table 1 shows that the RGB sensor system has a control accuracy of greater than 0.007 at 20°C (104°F), with further drift at higher temperatures. The color deviation in the feedback loop containing a True Color sensor, however, remains imperceptible to the human eye, at 0.0011.
In medical equipment, it is important that screens of diagnosis devices provide a high contrast ratio for easy viewing of fine detail, which requires display measurement devices offering high precision and sensitivity. Traditionally, display calibration has been performed—at a high cost—by calibration laboratories. Modern color sensors offer the prospect of a cheaper, faster, more convenient and equally effective alternative.
To test this assertion, ams created a second test. Here, it illuminated a diffusor plate with LED light at room temperature and an operating temperature at the LED of 20°C (104°C), and measured the color point. The measurement values produced by a Multi-Spectral sensor IC and by a mini-spectrometer were compared to the reference values provided by a spectrometer (Figure 4).
The measurement results demonstrated that the sensor IC and mini-spectrometer process signals more quickly than the reference spectrometer, but their error and accuracy values vary. The measurement values of the mini-spectrometer show a mean error of the color point measurement ranging from ∆ u’v’ 0.01 to 0.03—perceptible to the human eye. The measurements with the Multi-Spectral sensor show results with a mean error range of ∆u’v’ 0.001 – 0.005, a difference far less than the human eye can perceive (Table 2).
The printing industry has a requirement for spectral measurements during the printing process. Inline measurements can be challenging, since the measurement values are used in controlling the printing process. In a practical test, an X-Rite ColorChecker was used to provide absolute color measurements. Alongside this, ams used a multi-spectral color sensor with a multi-channel transimpedance amplifier and flexible amplification levels to perform spectral measurements. A white LED was used as a standard light source.
The multi-spectral sensor was used to measure the 24 color spaces of the ColorChecker, which themselves were compared with the reference values from a spectrometer. A regression equation for spectral approximation shows that the ColorChecker instrument achieved average accuracies of ∆E00 = 0.72 (Figure 5). A True Color sensor in identical conditions achieved average accuracies of ∆E00 = 1.57.
An advantage of multi-spectral sensors is their greater accuracy, and the scope that they provide to use spectral approximation methods. If the printing colors are known, the results can be improved via calibration of the specific colors. Therefore, it is possible to achieve absolute accuracies of ∆ E00 less than 1, independent of the standard observer and standard light source. The deviation values of the sensor compared to the spectrometer are ∆E00 = 0.3 for Cyan, ∆E00 = 0.9 for Magenta and ∆E00 = 0.3 for Yellow.
In our tests, all measurements were performed within a calibrated system. The light source, target and sensor have been calibrated to a reference spectrometer. These tests demonstrate that True Color sensors are capable of achieving the accuracy of a mini-spectrometer when making color measurements. In some applications, they are even more accurate. When deciding on the color measurement technology to be used, it is necessary to know what kind of color or spectral information is required and how this data is to be processed.
For example, mini-spectrometers are not capable of taking consistent measurements of the color of PWM-controlled LED light and are, therefore, inadequate in this application. RGB and True Color sensors do not provide spectral measurements so they cannot be used in applications that require spectral values. In this case, multi-spectral sensors or a mini-spectrometer should be chosen.
Table 3 summarizes the comparison of sensor and detector types. The table provides a grading of each on a five-point scale. Our testing shows there is a suitable sensor solution for every application. RGB sensors are a perfect match for simple color detection. True Color sensors are ideal for absolute color measurements. And multi-spectral sensors or a mini-spectrometer are suitable for absolute or spectral measurements.
The market for sophisticated and accurate chip-scale color and spectral sensors is growing quickly, driven by demand from mobile phone, display and specialty lighting equipment manufacturers. Optical semiconductor manufacturers have responded with the development of distinct families of sensor products to meet specific types of application requirements. This article described the types of optical sensors and detectors commonly used today, and explained the way to evaluate the suitability of each type for specific applications, and to specify the required characteristics and performance.
PUBLISHED IN CIRCUIT CELLAR MAGAZINE • SEPTEMBER 2019 #350 – Get a PDF of the issueSponsor this Article