the COLOR of COFFEE Using Color Analysis in Quality Control

by Paul Songer

OODS AND BEVERAGES caramelization and body. A medium quality control, but it is useful to have appeal to consumers through to full city roast has more complex instrumental confirmation to examine appearance, and color is an aromatics and a balance of acidity the roast in greater detail. While roasters Fimportant visual cue. Coffee is not as and body. Slightly darker roasts of the have flavor attributes on their mind bright or varied in color as many foods type used for Northern Italian espresso when they design a coffee, measuring and beverages, but coffee enthusiasts emphasize sweetness and body, and the the roast so these attributes can be do have preferences of roast degree darkest roasts are dominated by “roasty” consistently reproduced is essential for and can observe this by the color of the aromatics and burnt-sugar tastes. The quality control. There are a number of coffee. While offering a variety of coffee character of the green coffee can be either machines on the market that are capable flavor experiences is a key element emphasized or modified through the of performing this task. This article looks in marketing specialty coffee, when timing and final degree of roast, and this at how the color of roasted coffee can be customers find something they like, they can be measured in terms of color. measured and what one can conclude purchase it again. If what they purchase But coffee often looks better than it from the results. doesn’t provide the same experience as tastes. If it is known that the green coffee Instrumental measurement can give the previous purchase, they are likely to is of good quality, the problem is likely in clues as to what has occurred during the think less of the roaster. the roasting. What went wrong? roasting process, but other information is For roasters, the degree of roast is In production roasting, the goal is to needed as well. As with all lab methods, an essential consideration in balancing produce the intended roast degree and sample preparation and instrument flavor attributes. A “Nordic” light-roasted flavor development consistently. Sensory calibration will affect accuracy. coffee has a great deal of acidity but little determination is the primary method of continued on page 40

38 roast January | February 2015 39 THE COLOR OF COFFEE (continued)

PART 1 What one sees or what a machine measures is the interaction between the colors of the illuminant and the color of the subject The Science itself. The capability to perceive this combination depends on the ability of the eye or sensor to “see” the resulting wavelengths of light Behind Color Analysis being reflected. With perception by eye, one’s ability to perceive is limited to the visible spectrum, but some instruments are not so limited. What to Measure? Instruments designed to measure degree of coffee roast differ in what they measure and how the results are reported. Instruments There are two crucial processes that occur during roasting: that measure visible color as perceived by human beings are (1) development of flavor attributes, mainly sugar-browning intended to measure the degree of sugar browning produced as reactions, and (2) degradation of compounds. Both processes occur the result of roast; other instruments measure different roasting as the result of application of heat over time. The final degree of processes, such as the degradation of compounds, using ranges of roast and the time in which it was produced is an indication of how color that cannot be perceived by the human eye. What is being much heat was absorbed by the coffee. From that, one can make a measured depends on the illuminant and sensor used. How the reasonable assumption about the processes that took place. Roast results are reported depends on how the machinery translates those measurement usually seeks to quantify either development or measurements into a standardized scale. degradation. What is observed or measured in terms of visual cues is the result of three phenomena: Measurement of Color

The color of the light illuminating the subject by the light source One perceives different colors because of the innate human ability or emitter. This is the result of all different wavelengths of light to perceive light waves of specific wavelengths. A wave (Fig. 2) emitted by the source. is a vibration with a certain wavelength; different wavelengths result in different colors. Visible color is light in the area of 400 The optical properties of what is being examined, measured or to 700 nanometers (Fig. 3, pg. 42), but there are other non-visible perceived (the subject). The subject will either absorb or reflect wavelengths. While there are many methods of measuring roast, the light being emitted by the illuminant. they all seek to measure reflected wavelengths. The observer or, in the case of an instrumental measurement, the sensor. The observer perceives the light reflected from the subject and, if the wavelengths of the source are quantified, can Figure 2 :::: A wave. The wavelength (λ) is the distance conclude or measure which wavelengths have been absorbed. between two points at the same height (amplitude)

Figure 1 :::: Aspects of color perception

Light Source (illuminant) Observer

Like most sensory perceptions, not everyone sees the same wavelengths; some cannot distinguish between red and green and some cannot perceive different colors at all (color blindness). All wavelengths of light combined together are perceived as white; absence of any wavelength of light is black. Subject (object being observed) continued on page 42

40 roast January | February 2015 41 THE COLOR OF COFFEE | Using Color Analysis in Quality Control (continued)

Figure 4 :::: Spectral curve for natural Figure 5 :::: Spectral curve Figure 3 :::: Visible spectrum from infrared to ultraviolet. Wavelengths to the left A Scale of Different Colors A major challenge in the history of outdoor light, which varies according to for incandescent light and right of the infrared and ultraviolet lines cannot be seen by most humans. The science has been to develop scales that location on Earth from a tungsten bulb spectrum shows color bands according to specific wavelengths, but in the real world are consistent enough for scientific A raw color measurement consists of several 300 colors are made up of different combinations of these wavelengths. quantification, analysis and comparison. individual measurements of the different 100 250 This becomes more difficult when wavelengths of light being reflected. This is 200 the scale also should correspond 90 often illustrated in the form of a histogram in a meaningful way to a sensory 60 150 showing the intensity of reflected light at experience, such as taste, as one’s 40 100 various wavelengths. The different spectra Gamma Rays X-rays Ultraviolet Infrared Microwaves Radio Waves senses do not always operate in a linear 20 50 can then be used to illustrate a “spectral or predictable way. Developers of color 0 0 (µW/10nm/Lumens) Radiant Power curve.” This measurement of the reflection Equal Lumens for Power Relative 400 450 500 550 600 650 700 300 350 400 450 500 550 600 650 700 750 measurement scales sought to simplify depends on the wavelengths of light Wavelength (nm) Wavelength (nm) the complex interaction between light illuminating the sample (Figures 4–7). Figure 6 :::: Spectral curve for a Figure 7 :::: Spectral curve for a and the perceiver into something As color analysis evolved, different cool white fluorescent tube warm white fluorescent tube that was consistently measurable and scales for reporting colors were developed. corresponded to human experience. 300 300 Most of these were directly related to the The discovery of the spectrum of 250 250 graphic arts or areas where precision and light is credited to Sir Isaac Newton 200 200 consistency of visible color were important, Opticks (1643–1727) in his work . He believed 150 150 such as in the clothing industry. However, different colors were the result of subtle 100 100 as the technology developed and became particles that had mass and followed the more precise, color analysis was applied 50 50 laws of physics. J. C. Maxwell (1831–1879) 0 0 to other disciplines in ways that did not (µW/10nm/Lumens) Radiant Power (µW/10nm/Lumens) Radiant Power theorized that color was made up of 300 350 400 450 500 550 600 650 700 750 300 350 400 450 500 550 600 650 700 750 depend only on the visible spectrum. continued on page 44 Wavelength (nm) Wavelength (nm)

42 roast January | February 2015 43 THE COLOR OF COFFEE | Using Color Analysis in Quality Control (continued)

waves. Earlier, Thomas Young (1773–1829) At this point, the two major aspects of could be mapped in three dimensions. The be combined to produce all colors in the Similarly, Agtron developed two widely This scale is used as a reference had proposed a theory that all color was color perception were quantified: Newton three-dimensional representations are visible spectrum. used scales of coffee roast degree: gourmet throughout the coffee industry, regardless a combination of the primary colors red, had discovered that color was light (no called tristimulus color spaces. To get closer to the actual difference and commercial. On the gourmet scale— of the type of instrument. However, not all yellow and blue, and Maxwell’s work also light equals no color) and Helmholtz had However, these models had their in sensory perception, the CIE L*a*b* the one most widely used in specialty instruments measure the Agtron number in focused on this theory. It was Hermann determined that the observer was the final limitations. Each color has the properties system was developed. This scale was coffee—105 to 120 is considered the point at the same way; most work out a correlation von Helmholtz (1821–1894) who concluded arbiter of what could be seen as the result of hue, saturation and brightness. What based on experimental data of human which the organic compounds associated with the particular measurement technique that the eye perceived in terms of red, green of that light. These discoveries and the RGB has been described so far in terms of color matching. This allowed more with roast start to develop, and 0 indicates they are using. In instrument specifications, and blue, and it is now known that the eye (red-green-blue) scale were used to create the wavelength is the hue. This hue might precise definitions of color similarity and total reduction of organic compounds. (The the Agtron gourmet number is sometimes contains individual cones that perceive these first “color wheels,” in which a color could be intensely concentrated (like scarlet especially color difference, referred to in scale on the machine goes from 100 for very referred to as the SCAA number. three colors. be quantified in terms of three aspects that red) or might have more neutral color most equations as ΔE. This L*a*b* system light roast to 0 for complete degradation.) continued on page 46 mixed in (pink); this is saturation. Shiny was adopted by the food industry for surfaces are more reflective while others various applications. absorb light; this is brightness. When All of the tristimulus systems (RGB, perceiving or measuring color, the effect XYZ, L*a*b*, etc.) are mathematically of these properties must be taken into defined in such a way that they can consideration. be converted from one to another. The This led to the development of other CIE L*a*b* system (Fig. 9, pg. 47) is the scales that corresponded well to the most-used color scale other than those measures of wavelengths, but they did not specifically designed for graphics and correlate well with human perception of monitors. color. A group of color scientists formed the Commission Internationale de l’Éclairage Other Measurements (CIE), also known as the International of Roast Degree Commission on Illumination, in 1913. They systematized the analysis of color regarding Measuring the color of roasted coffee the “standard observer” (based on a person in terms of one of the systems above with average visual abilities), “standard corresponds to the degree of browning illuminants” (the source of light under that has taken place. However, one may which the color is observed), and a model have noticed with these systems that the of measurement called the CIE XYZ system. color brown—the one most associated A number of “standard illuminants” with coffee—is not readily apparent. Also, were defined, including incandescent the measurement of brown color does light, daylight and others. The “standard not directly indicate the type of sugar observer” was represented by a set of browning that took place. For example, three curves of primary colors that could development of furanthiol requires a temperature of at least 120 degrees C, and pyrazines also require higher Figure 8 :::: A two-dimensional color temperatures. Both are important coffee wheel. Several systems use such a aromatics. color space to define particular colors Measurement of the visible spectrum by degrees on such a wheel, with pure indicates the intensity of brown (intended red traditionally at 0 degrees. to reasonably correspond to sugar browning). In certain segments of the food industry, methods were devised for specific applications. For example, in the late 1970s a system was designed to determine tomato ripeness, in which a tomato was cut in half and a single “ripeness” value derived from the ratio of green reflectance to red reflectance. This ripeness ratio was referred to as the Agtron number, named after the machine that used a red neon lamp and green mercury vapor lamp to determine the green and red reflectance.

44 roast January | February 2015 45 THE COLOR OF COFFEE | Using Color Analysis in Quality Control (continued)

More recently, research performed by the quality customers expect when so many Measurement of degree of roast of both indicate how close any batch is to the ideal. If making comparisons, the sample must be appear and measure lighter than if the same other companies has led to different scales variables affect production. Formal quality whole bean and ground coffee. The whole measurements are too far from the standard, prepared the same way each time. sample is measured several hours or days based on observed changes in coffee roast control procedures can ensure consistency. bean is darker than the ground and the dif- one can use the descriptive sensory aspects to Timing also is a factor. If the coffee is later. Part of the standard should include an degree. Most of these are based on the red or In reviewing the different instruments ference between the measurements will in- investigate the closeness to the intended flavor measured fresh out of the roaster, it will approximate timing guide. near-infrared area of the spectrum. available for measurement of color, one dicate the rate of penetration of the heat and balance. continued on page 48 will notice they do not guarantee the the amount and type of sugar browning that To get accurate color analysis results, PART 2 intended flavor attributes were attained took place. The specification should include sample preparation procedures must be Figure 9 :::: The three dimensional L*a*b* space during roast. They can, however, be used the target and the distance from the target rigorous and consistent. It is common Applying as reasonably precise indicators, along measurement that is allowed. for different operators to get different with other measurements, to ensure the measurements from the same instrument Process measurements, or the “roast log” at- Color Analysis to intended degree of roast took place. and sample because of sample preparation tributes of time and temperature, especially White (L=100) Quality control programs start with variability. Coffee is not homogenous in color. Coffee Roasting at what point gold, first crack, second crack a standard—in this case, the ideal coffee Small beans are darker than larger beans, and (if present) and finish occur. flavor profile, which is the result of the when the coffee is ground, the particles from -b Using an Instrument -a green coffee and roast parameters. To A description of the sensory aspects that the outside of the bean are usually darker than for Quality Control -b +a establish a standard, it is suggested that should be present and their relative intensi- the particles from inside due to the progression “a” dimension several roasted prototypes of the green ties. For example, the intensity and type of heat from the outside inward. Grind size, (red to green) “b” dimension As coffee roasting businesses grow, the coffee be roasted to the approximate desired packing, density and distance from the sensor “L” dimension of acidity, mouthfeel and finish should be (yellow to blue) (black to white) founding roaster typically delegates some degree and cupped. One can either select detailed. This doesn’t need to be overly spe- all will affect the measurement. of the responsibility for roasting to others, one of the prototypes or produce another cific, but it should indicate what makes this For most instruments, one is directed to operates multiple roasting facilities, prototype based on what is found at the particular coffee unique. smooth the surface without packing it, with Black (L=0) and/or uses a variety of roasters for cupping table. an absence of “dimples” or low spots. If one can production. This creates a critical challenge Once the standard coffee is determined, With these specifications, one can use the observe a difference, the machine likely will for a growing business—how to maintain the following specifications are created: roast degree and process measurements to measure a difference. Especially when one is

46 roast January | February 2015 47 THE COLOR OF COFFEE | Using Color Analysis in Quality Control (continued)

Key Elements of Color Analysis Instruments

With color, there is a chain of perception that begins with the eye sensing the wavelength, then sending a message to the brain, where one becomes conscious of and interprets the stimulus. Measuring devices have their own mechanical chain of perception, first sensing the wavelength, processing the stimulus, then converting it into a scale for output. Instruments have been built to perceive multiple areas of the wavelength spectrum. Most instruments measure the sample in a closed, light-proof area. The sample is illuminated with a certain wavelength of light before measuring. Sensors often can be adjusted or Sample preparation is critical to accurate color analysis. Note the differences between the samples based on how they’re ground and whether or not the sample has been calibrated to perceive a specific combination of wavelength bands. smoothed/flattened. | photo by Paul Songer When considering a machine for color measurement, one should examine the following:

What is the light source? If the measurement is to be taken under can be used for the evaluation of green coffee. Another is that, because ambient light conditions, one should plan to use a consistent light the standards of color measurement are well established and precisely source to minimize variability. If the machine provides its own light quantified, inter-instrument agreement can be assumed. source but not in a sealed area, light pollution from external sources In general, color meters are based on two technologies, the must be considered. If there is a sealed area, one must consider how colorimeter and the spectrophotometer. A colorimeter measures the source is to be calibrated or maintained so measurements are the sample by passing the reflected light through a red, green and consistent and accurate. blue filter to try to duplicate the standard observer’s perception. A photodetector quantifies the amount of light passing through the How does the sensor work? One consideration is how large an area filters and reports the results in terms of one of the tristimulus scales. is being measured. For most color meters, the actual area of mea- These instruments are less expensive but may not be as accurate as surement is very small. However, this small area can be amplified more recently designed machines. through the use of lenses or by taking multiple measurements and Spectrophotometers break the reflected light from the sample averaging them. Other considerations include the sensitivity of the into a range of different colors on the spectrum. The resulting spectra sensor and in which spectra it can sense. then pass to a diode array that measures that wavelength of light. This

What scale is used to report the measurement? Some use the Agtron method is more accurate than a colorimeter, but the instruments are scale or another widely used color scale. Knowledge of these scales is more expensive. necessary to accurately interpret the measurement. Examples of color meters include the HunterLab ColorFlex EZ Coffee spectrophotometer and the Konica Minolta CR-410C colorimeter. How are samples prepared? What kind of sample preparation is Both units are capable of reporting results in terms of the SCAA necessary and how will (inevitable) mistakes in sample preparation number. The HunterLab ColorFlex uses an optical geometry to increase affect the accuracy of the data reported? the area of measurement. The Konica Minolta has a small aperture that requires multiple measurements for accuracy. Both instruments If there are several roasting facilities involved, one also should consider measure the sample from the bottom of the sample dish, so the optics inter-instrument agreement and calibration. The measurements made of the sample dish must be considered; scratches will affect the results at each facility should reflect the same level of roast as those taken at and the dishes can get contaminated. other facilities. The HunterLab ColorFlex EZ Coffee instrument uses a xenon white Color Meters light source for consistency and to approximate normal daylight color. The illumination is focused in a circle at a 45-degree incidence to the Color instruments measure the color that corresponds to the amount sample, allowing the light to be measured from different angles, of sugar browning that has taken place during the roast. However, as correcting any effect resulting from sample preparation. Its diode array noted previously, the degree of sugar browning will not necessarily measures the intensity of color at every 10 nanometers (nm) within guarantee the intended flavor attributes are present, as the actual the visible 400 to 700 nm range, but it focuses mainly on the 640 nm sugar-browning chemicals may vary according to heat applied and measurement, which the company’s research shows is the optimal timing of the roast. If using a color meter for evaluation of roast, the measurement for coffee roast. The instrument reports the results in standards must include the process aspects (time and temperature, or four ways: SCAA number, roast classification (“light”, “medium”, etc.), “roast curve”) to ensure the intended chemical changes are produced. CIE L*a*b* scale, and the company’s own HunterLab Coffee Color Index

One advantage to using color meters is that the same instrument (HCCI). continued on page 50

48 roast January | February 2015 49 THE COLOR OF COFFEE | Using Color Analysis in Quality Control (continued)

Coffee-specific Instruments (Using such a sample ensures that changes in color are no longer taking A handheld version is available as well as a laboratory bench-top model. What's Next? place over time.) The machine should be serviced and the infrared lamp ColorTrack uses a laser for the light source, which rarely needs

Because of the obvious need for measurement of roast degree and its replaced annually. to be replaced and never needs to be recalibrated. The bench-top HE GOAL OF MODERN QUALITY control programs is to prevent out-of- importance in quality control, several instruments have been designed Sample preparation for the Agtron is of major importance in model has a carousel that turns the coffee sample so approximately Tspec products from being produced rather than simply rejecting poorly especially for coffee. As noted previously, the “brown-ness” or darkness attaining an accurate reading. The surface of the sample must be 200,000 measurements are taken (10,000 per second over a 20-second produced products, but measurement of color in all of the cases examined in of coffee can correspond to the degree of roast, but the brown portion reasonably uniform without “dimples.” The particle distribution as measurement period). Because coffee color is not homogenous, this article occurs after the roast takes place—too late to make adjustments. of the spectrum is small. One also needs to consider the process that the result of grind also is important; it was observed by one lab that the reports include standard deviation over the surface as well as Out-of-spec coffee must be donated, used in another product or disposed of, produced the roast degree. eliminating the larger and smaller particles by measuring only the the average of all measurements. Reports also include graphic often representing a loss for the roaster. Future technology could be capable The Agtron is one of the first instruments to employ methodology particles in the No. 16 screen of a standard set of particle analysis representations, similar to the histograms and spectral curves of measuring color changes as the roast progresses. (There are infrared specifically developed for coffee. Near-infrared (non-visible) light screens (particles of approximately 1.18 mm) resulted in the most discussed earlier. The ColorTrack measures the sample directly measurements for temperature, but at present these are difficult to use.) illuminates the coffee to sense a selected group of quinones (byproducts replicable measurements. Whole bean coffee can be measured, but (without glass or other reflective surfaces) and the laser technology is This type of measurement is technically difficult because one would of the degradation of chlorogenic acids that occur during roasting). The the likelihood of voids between the beans may cause the reading to be not “surface sensitive” (does not depend upon distance, angle incidence need to have a dependable light source and a quick way of reading what is sensor measures certain (proprietary) wavelengths to determine the darker than the actual sample. or smoothness of surface), so sample preparation has a minimal effect reflected. However, instruments for measuring color during roasting have been amount of quinones developed, which directly correlates to the degree Madison Instruments produces Javalytics, which uses near-infrared on results. The measurement focuses on the 780 nm wavelength to investigated and currently Fresh Roast Systems (makers of the ColorTrack) is exploring the manufacture of such a system. of roast. As a direct measure of the chemistry, the results correspond illuminants to measure degree of roast. The sample is prepared similar report results in terms of the Agtron number. A handheld version also well with the intended flavor attributes produced. to the Agtron, leveling across the sample plate or filling a glass dish is available. There are two types of Agtron machines (with some differences before measurement. Ambient light does not affect the result due to While the primary method of evaluating coffee quality is sensory in individual models within these types): a closed system (the M-20) the use of filters. The sensor integrates the reflectance of the entire analysis, the use of color measurement technology can assist in and a bottom-measuring system (the M-15). With the M-20 series, area of the sample (about 3.5 inches in diameter) and results are improving quality and consistency. To optimize the use of these PAUL SONGER has been in the specialty coffee business for more than 30 years, the machine is calibrated each time the drawer is opened. A disk is reported in terms of the SCAA number, with electronic temperature technologies, roasters must understand the technical aspects of color first at Allegro Coffee Co. in Boulder, Colorado, then Coffee Analysts in Burlington, provided for calibration of the M-15. Inter-instrument agreement can compensation. It has a calibration disc that should be used daily, or measurement and integrate these measurements into a quality control Vermont. He holds a certificate in sensory analysis from the University of California be attained with adjustments available; one must have a dead stale more often, depending on the environment. Administrator functions program that also takes into account the roasting process and sensory at Davis. He currently works as a consultant on sensory and technical issues regarding sample—one that has been ground and oxidized for at least 30 days include the built-in RoastMatch wizard, which allows one quickly to parameters for each coffee roasted. coffee and serves as head judge for the Cup of Excellence coffee competition program. without significant moisture added—to measure in each machine. create and store various calibrations for inter-instrument agreement.

50 roast January | February 2015 51