Editor’s note: This article was first published in Roast Magazine’s May/June 2021 issue, and is reprinted here with permission.
An Evaluation of Consistency, Cupping Settings, and the Impact of Roast Degree
Experiment design, operation, summary and analysis by Chris Kornman and Elise Becker with contributions from Sandra Elisa Loofbourow
One of the numerous variables identified for altering the strength and flavor of a brewed coffee is the grinder’s setting, which affects the particle size and shape of ground coffee, with dramatic implications for the flavor profile of a coffee beverage. The size and shape of ground coffee particles directly corresponds to the available surface area for interaction with water during brewing, a key consideration in the final total dissolved solids (TDS) in the brew and the extraction percentage of the spent grounds. Grind settings can also impact variables such as the rate of staling or degassing of the coffee.
A well-regarded article published in 2016 in Nature’s Scientific Reports titled “The effect of bean origin and temperature on grinding roasted coffee” uncovered two essential principles in grind science. First, it proved that colder temperatures produce finer and more uniform particles. Second, it indicated that there is no correlation between bean origin and grind discrepancies.
Inspired by this finding, we sought to produce a significant cache of repeatable data that further examined additional principles relevant to the specialty coffee roasting community, using easily accessible equipment. Namely, we sought to confirm how manually adjusting grinder burrs affects precision across multiple grinder units of similar makes and models.
This led us to further investigate the industry’s cupping grind setting across a number of metrics, including the impact of roast degree on grinder precision, variance between surveyed roasters, and differences between the in-house settings at our company, The Crown: Royal Coffee Lab and Tasting Room, and the standards recommended by the Specialty Coffee Association (SCA) and the Coffee Quality Institute (CQI).
The results are published here for the first time, and we are thrilled and astonished by the findings.
MAHLKÖNIG EK43 GRINDERS – FACTORY SETTINGS AND INDUSTRY STANDARDS
The Crown’s fleet of grinders includes four Mahlkönig EK43 grinders: three classic EK43 models fitted with 11-digit faceplates, and one of the “shorty” EK43S models, which is equipped with a 16-digit faceplate. Mahlkönig generously supplied an additional “S” model for this experiment.
One of the first questions we had was related to the factory calibration methods. We checked in with Mahlkönig manufacturer representatives from the Hemro Group, who supplied this response:
“The factory calibration and testing procedure for new EK43S (standard disc setup) ensures grinders reach a fineness of 180 microns or finer to grind espresso while also ensuring enough output power to grind upward of 50 kilos of filter coffee on a daily basis. The factory does not employ variation testing between grinders since each of them is assembled by hand, yielding a certain degree of variation among all grinders. Variation testing and calibration would require a lengthy and repetitive process that would also raise the product’s price.”
Thus, rather than providing users with a baseline for “calibration” to a factory standard, instead Hemro supplies a standard for output capacity and fineness. We clean and align the burrs on a quarterly basis using the recommended calibration steps from the manufacturer and the added step of using a dry-erase marker and aluminum foil to increase the precision of the burr alignment, as popularized by Barista Hustle and extensively analyzed by Thomas Manion for Flight Coffee.
For the uninitiated, the process involves removing the faceplate, cleaning the burrs, marking the outer flat edges and replacing them, then turning the grinder on and screwing the burr set into place so that they lightly touch at the finest setting. With the faceplate removed, any irregularities in the spinning axis of the burrs can be analyzed by observing erasure of the markings and corrected by inserting strategically placed aluminum shims to correct the asymmetries.
We inquired with Hemro spokespeople as to their stance on this additional step. The official response was enlightening:
“We take this method seriously as an indication from the high-end market to optimize burr alignment in new products and have acted upon it in 2018 by optimizing the pre-breaker and machining tolerances around the grinding areas. Users who advocate [for] this method typically employ EK43S for single dose grinding of high-quality espresso beans and are willing to exert significant effort to achieve the optimal extraction result. As a manufacturer, these lead users govern our [research and development] but exceed requirements of the mainstream market that cannot accept the price increase needed to adjust manufacturing processes while complying with global food and safety certifications that were obtained with the original design. The main reason we do not support this method is that it introduces loose metals into a critical food area that may result in bodily harm for coffee consumers.”
We figured we could test our “better-than mainstream” alignment method by grinding coffee and weighing the doses using the popular and easily accessible Kruve sifters, which are conveniently delineated in micron (μm) measurements. Our hypothesis was that well-calibrated grinders should produce highly similar distributions of ground coffee particle size.
Kruve CEO Adam Krupa informed us that Kruve sieves “are manufactured using a chemical milling process used in various industries including the most advanced medical instruments. The actual tolerances on the holes are +/- 20 μm,” which exceeds the standards for laboratory sieves. Krupa confirmed that by weighing grind distribution using Kruve sifters, users could “measure and compare the difference between a grinder setting, burr wear, [and] share or tweak recipes.”
The way coffee grounds fall through sieves will be affected by numerous factors, including degree of agitation and sifting duration. While industrial machines like the popular RoTap are automated, the Kruve is a manual device, susceptible to operator variances. Krupa told us that “30-40 seconds of sifting time generally gets you 90 percent” accuracy. He continued, “Sifting for an additional 60 seconds … will not yield much different results. Maybe a more accurate method is to use a laser particle
Krupa’s response gave us a high degree of confidence about using a consumer product for sufficiently precise measurements, so long as we held ourselves to strict standards.
Methodologically, we adhered to several stringent steps to assure the reliability of our results, outlined here:
- Each grinder was deep cleaned and modified with aluminum shims for accuracy of grind, and thereafter nicknamed after one of the four hobbits in The Lord of the Rings for easy differentiation.
- Each grinder was tested by grinding a 50-gram sample at each full number on the EK43 faceplates, 1-11 on the standard EK43 and 1-16 on the EK43S. (Users may adjust their grind settings along a dial; each faceplate setting represents an incremental adjustment of a burr disc, which moves closer or further away from its corollary static burr, thus creating wider or narrower passages for the coffee as it passes through to produce coarser or finer grounds.)
- We roasted and ground a Colombia Huila Supremo, chosen for its tight green screen size distribution, and roasted in single batches for each round of testing.
- Each 50-gram sample from each setting on each grinder was passed through Kruve screen sizes starting at the largest in the set, stepped down incrementally by 200 microns. The screens used are 1,600 μm, 1,400 μm, 1,200 μm, 1,000 μm, 800 μm, 600 μm, and 400 μm. (Because the Kruve holds only two screens simultaneously, it took four passes for each 50-gram sample to clear all seven screen size settings.) A blank screen at the bottom of the Kruve accounts for the total remainder of particles that passed through (anything less than 400 μm). For later rounds of testing, the 200-μm screen was introduced to confirm Hemro’s claim that the grinders could produce 180-μm fines. (Refer to https://www.mpechicago.com/wp-content/uploads/2019/12/size_conversion_table.pdf for screen size conversions from microns).
- Each 50-gram sample was ground into a 2-quart Cambro for ease of transfer. Each time the sample was loaded into the Kruve sieves, any grounds clinging inside the Cambro were brushed into the Kruve.
- Each sample was shaken lightly by hand for no less than 2 minutes and no more than 5 minutes.
- After each pass through the sifter, grounds passing through to the bottom were reloaded into the same Cambro for the next pass. Grounds sorted by the sifter were set aside into a separate Cambro.
- The sorted grounds from each sieve setting were weighed on a gram scale (sensitive to 0.01 gram). Between each change of screens, material that became trapped in the screens from the previous sift, along with small amounts of fines that collected in the silicon gasket between the sifter basket and the screen, were carefully brushed off and set aside to avoid re-recording them on other screen sizes.
- After each screen change, the screens and sample baskets were brushed clean with a flat cut nylon bristled utility brush. Denatured alcohol wipes were applied to each sieve between every three to six measurements to remove oils (more frequently with darker roasts).
- For the smallest particle size, clumping of the fines (especially for finer grind settings on the faceplate) disrupts accurate sifting. Troubleshooting measures for accuracy included gently breaking up clumps with a finger throughout sifting.
All four grinders produced similar percentages of large “boulders” and small “fines” across a range of grind settings. While all grinders were adept at eliminating coarse particles at their finest settings, each grinder, even at the coarsest setting, produced a small percentage of fines (2.5 percent to 7 percent of the total mass measuring greater than 400 μm), and only the standard EK labeled “Samwise” produced less than 5 percent at its coarsest faceplate setting.
Generally speaking, the EK43S “Frodo” produced finer grounds overall. The faceplate has more settings so one might expect the existing settings to produce a tighter spread, but even at the finest and coarsest grinds, it was still producing finer grinds than the regular EKs at their finest and coarsest settings by small percentages.
The graphs below display measurements of individual grinder settings on the y-axis (finest setting No. 1 at the bottom). Particle size (x-axis) is represented as a percentage of the total mass of each sample at each Kruve micron sieve. Warm colors on the left are finer grinds, while cooler colors on the right are coarser.
Surprisingly, rather than displaying high percentages of coarse particles at the coarse settings, the grinders displayed a wide spread of distribution across all particle sizes.
While the three standard EK43 grinders were each configured using the same method (outlined earlier), each displayed slight idiosyncrasies. Overall, they tended to produce ground coffee in similar percentages in the mid-range settings while showing some discrepancies in the outlying coarse and fine settings. This struck us as interesting, as there is no difference in the grinder motors or burrs from the manufacturer.
While the engineers at Hemro declined to comment on the presence of very fine particles even at coarse settings, they did mention that the EK43S “Frodo” as an outlier was “not expected, since the EK43 Standard and EK43S do not differ apart from their model height.” As a result, they shipped us an extra “shorty” to investigate further, though the results of this comparison are still pending at the time of writing this report.
Having sifted through the entire range of EK43S faceplate settings, we were able to check the average from the cupping setting on the three regular EK43 grinders against the full range on the EK43S and make an accurate prediction of the setting that would produce the closest numbers to the group average on the EK43S “Frodo.” This proved to be 12.5 on the 16-digit “shorty” faceplate.
Looking more granularly at grinder alignment at our in-house cupping grind (8.5 on the 11-digit faceplate), all four of the EK43s were quite close, no more than 0.77 grams (1.54 percent) standard deviation—and at all but two screens within less than 0.3 grams (0.6 percent) standard deviation—at any Kruve sieve size, an impressive precision between grinders.
The prediction was near-perfect, with no more than 2.25 percent maximum discrepancy at any Kruve sieve size. Interestingly, the 1,000 μm sieve proved the most inconsistent at this setting from grinder to grinder, with three of the four grinders (including Frodo, the “shorty”) experiencing their largest divergences at this micron level. Such small variations may be attributable to our manual sifting techniques or even small amounts of coffee oil buildup on the screens.
What this seems to confirm is that well-aligned grinders should be able to produce minimal variation at similar grind settings using the same roast, a comfort to anyone trying to communicate grind settings to another user of the same make and model.
ROAST DEGREE AND IMPACT ON GRIND PARTICLE SIZE AND DISTRIBUTION
However, the effect of roast degree on grind profile may be more significant than many users realize.
In our preliminary research, we were unable to locate published data correlating the impact of roast level on grind profile. This led us to measure the effect of dramatically different roast degrees on the way a grinder interacts with the coffee bean.
We were expecting that, to some degree, darker roasts would produce a higher percentage of fine particles, based on anecdotal observations of lower moisture and higher brittleness. We were astounded at the degree to which this was true. Conversely, we also found that lighter roasts produced significantly more large particles.
For our experiment, we took a light (53.0 ColorTrack, ground sample, approximately 50 Agtron), medium (57.5 ColorTrack / ~45 Agtron), and dark roast (64.5 ColorTrack / ~40 Agtron) of the same coffee, Colombia Huila Supremo, and ground it on two of the EK43 standard faceplate grinders, Samwise and Merry, at the odd-numbered settings 1 (fine) through 11 (coarse). We adhered to the same sifting protocols outlined earlier in the article for all roasts.
On both grinders at every faceplate setting the overall result, with few exceptions, was that the dark roast was finer, with fewer particles present in large micron sieves and, conversely, more in the smallest sieves.
When measuring the results, we noted that especially the light roast, and to some degree the medium roast, produced significantly more chaff than the others, which remained in the largest micron sifters and likely contributed to the increased weight in the 1,400- and 1,600-μm sieves. By contrast, the dark roast grounds were free of chaff but tended to cling and clump more due to the oil content, taking significantly more time and effort to accurately break clumps and tap through the Kruve screens.
The dark roast produced more fines overall, and significantly more at finer grind settings. Additionally, the light roast showed a wider spread of particle sizes at all grind settings.
When the data are grouped into two distinct categories of overall coarse “Boulders & Chaff” (greater than 1,200 μm) and “Fines” (less than 400 μm), expressed as a percentage of the full sample, the differences were stark.
At the six grind settings we looked at, the dark roast consistently showed more fines and fewer boulders/chaff. Conversely, the light roast consistently showed more boulders/chaff and fewer fines at every setting. Roast level appears to be a significant contributing factor to how a coffee shatters when ground.
When comparing just the dark and light roasts, the dark roast produced average differentials of more than 4 percent fewer boulders and chaff, but more than 10 percent more fines across all grind settings on average. Maximum differentials peaked at 13.3 percent fewer boulders and chaff at setting 9, and 23.4 percent more fines at setting 3. These percentages fall well outside the expected variance based on what our inter-grinder calibration tests suggest might be “normal” deviations.
IMPLICATIONS OF ROAST DEGREE AND GRIND SETTING
A commonly repeated aphorism in coffee journals and conversations maintains that darker roasts are more soluble. Data presented in 2018 at ASIC by the 2017 Swiss Barista Champion André Eiermann suggests that this is true (in capsule brewing applications, specifically), though this data is contradicted by some of The Crown’s own informal investigations, published in Daily Coffee News in 2017, which implied that lighter roasts of the same green coffee may tend to contain more soluble material (higher TDS) when brewed using the same grind, dose and water on auto-drip machines.
Our 2017 findings were corroborated by Johnny Randolph on the Fellow Products blog, asserting, “Some people might tell you that a darker roasted coffee is more soluble, and that seems logical, but in reality, a lighter roasted coffee is more soluble! The darker a coffee is roasted, the most compounds have been roasted away and although the cell structure is more broken down, there are less solubles available.”
If the data presented in this report on grinder and roast-degree relationship holds, it’s worth considering that in some circumstances, darker roasts may appear more soluble when ground at identical settings due to their increased percentages of finer particle sizes (which have a higher surface area to volume ratio and therefore may extract at higher percentages than larger particles). We note that our research does not include three-dimensional analysis of grind particle shape anomalies, which may differ depending on a roast level’s shatter profile, and thereby affect extraction.
For those considering the intersection of solubility and roast degree, it seems likely that grind particle distribution will be an important factor to add to the list of variables. Thus, using a similar grind “profile” or sifting to normalize deviation likely reduces variables compared to using a similar grind “setting.” The data here support the use of sifters when brewing to maximize consistency regardless of the roast or coffee in use.
In coffee brewing, the presence of boulders and chaff will theoretically have a more significant detrimental impact at finer grind settings. We observed that roast-related percentages of coarse particles did not become significant until the middle of the faceplate dial, between settings 5 and 7. Thus, there’s likely little impact of a light roast’s additional boulders at finer grind settings (such as those used for espresso).
However, fines at all settings, and the high percentages of fines in the dark roast specifically, may have a significant impact in brewing. Consequently, when grinding at medium and coarse settings (e.g., filter drip or cold brew) the implied impact on extraction will theoretically be significant.
What all this seems to indicate is that there is much more work to be done to understand the convergence between roast degree and grind size, and by extension a coffee’s solubility.
One additional implication may for roasters who offer ground coffee for sale to consumers. If a dark and a light roast from the same roaster are ground on the same setting and packed as, for example, “ground for paper filter,” they may in fact be significantly different grind profiles from each other. As a result, roasters who sell ground coffee to consumers may wish to consider unique grind settings depending on application (such as drip or espresso), which are aligned to the degree of roast.
CUPPING GRIND COMPARISON AND EVALUATION OF THE SCA/CQI STANDARD
CQI and SCA use identical standards for cupping grind, stipulating that “coffee used for cupping shall be ground so that 70-75 percent of the grinds pass through the 20 mesh sieve.” As illustrated in the diagram below, a standard US 20 sieve is equivalent to 841 μm in length and width (a Tyler Mesh 20 is 833 μm), but because standard mesh sieves are composed of crosshatched wire squares, the diagonal maximum measurement is 1,189 μm, making this measure highly imprecise.
Kruve sifters are engineered with round openings, however, meaning that 800 μm means 800 μm no matter which way you look at it. We note that round holes cannot account for irregularities in three-dimensional grind particle shape, but neither can square holes, for that matter. With this in mind, we believe our measurements were able to determine that the SCA’s specs include a potentially wide range of settings on our EK43 grinders.
At the minimum micron size, it appears that somewhere close to the No. 4 setting on the 11-digit EK43 faceplate is a reasonable proxy for the finest possible interpretation of the SCA standard. At the coarsest, about 70-75 percent of ground particles under 1,200 μm (a reasonable proxy for 1,189 μm maximum diagonal measurement) pass through at setting 9.
However, it stands to reason that irregularly shaped coffee grounds with a maximum diameter of 1,189 μm will not necessarily all pass through a square mesh with 841 μm length/width dimensions, so to achieve the SCA grind recommendation on a calibrated EK43 11-digit faceplate, you must definitively grind coarser than 4 and finer than 9.
Exactly where in that range might be optimal is up for debate. We corresponded with SCA Technical Director Mario Fernández, who expressed dissatisfaction with the current language, mentioning that “sieving coffee isn’t a perfect measure since it is very dependent on particle shape, amount of chaff, etc.” Both Fernández and Krupa also mentioned that without a sieve recommendation for a minimum size, the CQI and SCA’s current standards lack sufficient precision to be prescriptive.
We contacted a former member of the SCA’s now-disbanded Standards Committee, who suggested that the recommendations were derived from Ted Lingle’s The Coffee Cuppers’ Handbook, and sure enough, the language is identical. From the third edition (2001), page 27:
“Coffee beans used in this specific infusion [cupping] method are ground into a ‘fine’ grind, in which approximately 70 percent to 75 percent of all the particles pass through a U.S. Standard Size 20 sieve. The purpose of this grind standard is to achieve an 18 percent to 22 percent extraction rate…”
Combined with the intended dose and water weight, Lingle’s formula lands the cupping extraction firmly within the so-called “golden ratio.”
We noted with interest Lingle’s language of “fine,” which conflicted with our recollection of comparisons to “slightly coarse” or “sea salt.” Fernández explained, “If you remember, the original protocol read: ‘Grind particle size should be slightly coarser than typically used for paper filter drip brewing …” This language is no longer included on the SCA’s downloadable “Coffee Standards” PDF, but the phrase was in play for years, and clearly contradicts Lingle’s “‘fine’ grind” recommendation. Based on historical data from Modern Processing Equipment, Inc.’s 2019 “Ground Coffee Specifications Analysis,” it appears that differences in “drip” and “fine” grind settings in the 1940s were negligible. Thus, modern specialty coffee’s drift towards coarser drip settings may account for current ambiguity regarding SCA cupping and drip grind profiles.
We decided to test the outer limits of Lingle’s recommendations. Using Third Wave Water and all other standard SCA practice protocols, we set up a cupping to observe and report TDS and extraction data on two different coffees (a Peru and a Rwanda) over the course of 30 minutes of steeping with a standard break at four minutes followed by a skim. Coffee was agitated lightly on the surface once at 15 minutes to simulate a live cupping, and samples were drawn from the top 10-mm of the cupping bowls to replicate the portion usually sampled.
The results, even at the finest possible setting of 4, synchronized with Lingle’s expectations, and The Crown’s in-house settings were clearly much different based on dose and grind, with lower TDS and extraction percentages. The implication is that to achieve recommended extraction specifications, the cupping grind setting potentially should be far finer than The Crown’s house setting.
This led us to inquire with other local coffee professionals. If our house grind setting was miscalibrated and under-extracting, were other coffee professionals in a similar predicament?
We solicited self-selected, ground samples from five Bay Area specialty coffee roasters at their respective cupping settings, as well as one from the Royal Coffee QC grinder (used in regular cuppings for importing purchases). Participating roasters included Alchemy Collective, Blue Bottle, Café Reveille, Proyecto Diaz and Retrograde Roasters, each of whom agreed to participate with anonymized data.
Coffees submitted were chosen by the roasters as representative of their standard roast profile, and country of origin was not considered an important factor. Grinder brands employed included Bunn, Mahlkönig, Grindmaster and Ditting. Rationale behind roaster decisions for their dose and grind settings ranged from adherence to SCA standards to sensory preference and experience from working with others.
We found that across the board, none of the surveyed roasters conformed with the SCA’s finest grind setting and many were grinding at similarly coarse settings compared with The Crown’s (including three nearly identical grind profiles: two from local roasters—“VIW” and “AYE”—and the other from the Royal QC Lab).
We confirmed, speaking with the roaster of the finest grind, labelled “WAE,” that their protocols were established to achieve a target TDS in their final brew, one which conforms with the Lingle method.
In the chart, the bars at the bottom are the outer bounds for the “SCA” standard, as ground at No. 9 and No. 4 on an 11-digit EK43. Particle size is represented as a percentage of the total weight of each sample at each Kruve micron sieve. Warm colors on the left are finer grinds, while cooler colors on the right are coarser.
Fernández, who holds a doctorate in food science, assured us that the SCA is “… currently working on a standard that will use a more modern approach to measure particle size, and will also take into account new findings in brewing dynamics.” We look forward to the publication of these new standards.
In the meantime, the discrepancy between empirical sieve data and an ambiguous coarseness recommendation leaves most roasters in a place where they must decide for themselves what standards to employ and which factors matter most: extraction data, sensory evaluation, grind size,
dose and water ratio, and others. Clearly, this has led to broad distribution of grind settings across the industry.
At The Crown, our findings have led us to reexamine our in-house settings and consider adjustments to both our coffee dose and grind size to better align with extraction science and the importance of maintaining proximity of calibration with our peers.
CONCLUSIONS AND NEXT STEPS
Our examination of grinder settings led us to better understand how well-calibrated grinders can be relied upon to produce overall similar particle size distributions. It also enabled us to successfully predict settings on similar grinders with differing faceplates (e.g., the 11- and 16-digit EK43 options). We received some excellent feedback from Mahlkönig and Hemro representatives and from Kruve CEO Adam Krupa regarding
production specifications and opinions on grinder calibration techniques and particle measurement precision.
This data led us down a few rabbit holes, and to the surprising revelation that the SCA’s cupping grind recommendation is poorly defined and potentially much finer than we expected. We surveyed local roasters and found that despite some degree of uniformity between some roasters,
overall, the grind and dose settings were generally somewhat different from each other and somewhat divergent from the SCA language.
We also found that grind particle size is affected substantially by the color of roast of the beans. The implications of this may indicate that our assumptions regarding solubility of differing roast levels should account for grind profile. It likely also implies that grinding for any application should consider the important differences in fine and coarse particle size, which will likely be contingent on the degree of roast of the coffee in question.
We felt this was useful enough to share, and we welcome feedback and critique. We also look forward to the work undertaken by Fernández and his team at the SCA to better refine our understanding of cupping standards and extraction theory.
We would hesitate to recommend immediate changes based on our findings, as they have not been peer reviewed, nor is the cache of data definitive. We do believe, however, that these findings can make meaningful differences in practical applications. Using our indications, roasters can choose to align more closely with the SCA’s cupping grind recommendation, gain confidence in their calibration routines, and make decisions regarding how to grind coffees that are roasted to different color degrees.