Chocolate making, professionally

Chocolate Refining: Melangers, Roller Refiners, Ball Mills

By Oliver BennettPublished October 15, 2026Updated October 15, 202615 min read

What refining does

Refining grinds the sugar and cocoa particles in chocolate down until the tongue reads the result as smooth rather than sandy. The machine crushes and shears the solids in the paste, and the particle size it leaves behind shapes how the chocolate looks and how it feels in the mouth, as Tan and Balasubramanian put it in a 2017 study. If the cocoa liquor was ground well beforehand, the main work left is shrinking the sugar, according to a 2019 study by Hinneh and colleagues.

It is also slow and expensive. A 2019 Penn State study by Breen and colleagues calls refining “a common bottleneck in production”, because finer particles take more processing time and more energy. That trade between fineness, time and energy runs through every machine below.

Most refining is done by one of three machine types: stone melangers, roller refiners and ball mills. They reach similar fineness by very different routes, and the route sets what comes before and after them in the process.

How fine is fine enough?

There is no single threshold, and the figure you will see most often has a weak source. Chocolate is widely said to feel sandy or gritty once its particles pass roughly 25 to 35 µm. When the Penn State researchers traced that figure, the original source was missing. The closest they found was a 1967 report by Rostagno, which judged chocolate unacceptably coarse when 20% of its particles were larger than 22 µm.

So they tested it. In melted chocolate, the smallest particle size difference people could detect was about 5 µm, in a range common in commercial chocolate. Then they gave tasters two solid commercial chocolates with D90 particle sizes of 19.70 and 25.20 µm, about 5.5 µm apart. The grittier one was picked out reliably, but only thanks to people with sharper oral sensitivity: 17 of 20 in that group chose correctly, against 12 of 24 in the less sensitive group.

What feels gritty depends on who is eating. Other papers repeat an upper limit of 30 or 35 µm and a target range of 15 to 25 or 15 to 30 µm, mostly by citing earlier work rather than new tasting panels. A 2020 study by Clark and colleagues is one of them; the rest are listed at the end. Use any single number as a starting point for your own tasting.

Going too fine has its own cost. Clark and colleagues, citing earlier work, note that below 4 µm the particles have so much surface for the fat to coat that chocolate can turn slimy and thick. The bean-to-bar maker Dandelion Chocolate puts it in practical terms in a 2015 post on grindometers: if the texture and flavor taste flat, the batch may have stayed in the melanger too long and been refined too fine. They aim for 20 to 30 µm.

When a paper or a spec sheet gives one fineness figure, it is usually the D90: the size that 90% of the particles, by volume, fall below. Measured by laser diffraction, it tracks how people perceive the largest particles, the Penn State team found. Not every fineness figure is a laser D90, though; the section on measuring covers why that matters.

Melangers (stone grinders)

A melanger refines by rolling heavy stones over a stone base inside a turning drum, and in small workshops the same machine often does part of the conching as well. It is the cheapest way into refining and the slowest.

How they work

In the machine studied by Hinneh and colleagues, cylindrical granite stones turn at 135 to 140 rpm on a granite slab, crushing and shearing the cocoa solids. Tightening the tension between stones and slab makes the grinding more intense. Stone shape made little difference in the comparison by Tan and Balasubramanian: melangers with conical and cylindrical stones performed similarly, though cylindrical stones tended to produce thin, flat particles in the first 4 hours.

Melangers are usually described as running at ambient temperature, while conches are generally heated up to 80 °C, Clark and colleagues note. Friction changes that picture. In Hinneh’s runs the chocolate warmed to about 40 °C on its own, and one low-fat batch (27% fat) climbed to 78 °C after more than 3 hours in the small drum. The model they tested had no temperature control, which they named as its main flaw.

Time does most of the work. With a full-fat recipe of about 40% fat in the small drum, Hinneh measured a D90 of 37.87 µm after 1.5 hours, 23.86 µm after 3 hours and 22.89 µm after 4.5 hours. The last 90 minutes bought less than 1 µm. The 27% fat recipe was still at 65.39 µm after 3 hours. In Clark’s study of a small melanger, the majority particle size was generally below 25 µm at every point after 8 hours, read on a grindometer. Dandelion Chocolate leaves each batch in the melanger for three to five days, refining and conching together.

One melanger maker and another advertise particle sizes around 15 µm without saying how that figure is measured. Neither independent study above reached it: Hinneh’s D90 settled between 22.89 and 23.86 µm, and Clark’s grindometer readings stayed in the “below 25 µm” range. Recipes and methods differ, so this isn’t a direct test of the claim. It is a reason to plan around what you measure on your own recipe.

Where they fit and where they stop

Clark and colleagues, citing trade press, describe melangers as now the most common refining tool for small-batch chocolate makers, because they scale easily and cost little up front. Hinneh’s team adds a small footprint: the model they tested took up less than 0.14 m².

Hinneh’s study also found the melanger a workable alternative to a laboratory three-roll refiner, as long as the recipe carried around 40% fat. The difference was time. The roll refiner took 5 to 10 minutes; the melanger took hours.

Usable capacity can be lower than the listed one. Hinneh’s team put the tested model’s large drum at about 3.6 kg and its small drum at about 1 kg. The maker lists the same model at 4.5 kg.

The evidence that refining method changes flavor in a way people notice is thin. In a second 2019 study by Hinneh and colleagues, melanger and three-roll chocolate differed in some groups of aroma compounds but not in total volatiles, and a 112-member consumer panel showed no significant preference between them. Inside a melanger, time had a bigger effect on flavor than temperature, Clark’s team found.

A melanger stops paying off when batch time, batch size, low-fat recipes or the lack of temperature control start to cost more than a bigger machine would.

Roller refiners (five-roll)

A roller refiner crushes and shears chocolate paste between steel rolls that turn at different speeds. In industrial lines a two-roll pre-refiner usually feeds a five-roll refiner, and what comes out is a dry flake that still needs conching.

How they work

Makers describe two forces at work: pressure between the rolls, and shear from the speed difference between them. The paste enters at the lowest feed roll, and each roll after it runs faster than the one before, so the particles get smaller with every pass. A smaller gap gives a finer result, Hinneh and colleagues note.

The two-stage setup splits the job. According to an equipment distributor’s guide, the pre-refiner takes paste made with coarse crystal sugar down to 100 to 200 µm, and the five-roll refiner then takes it to about 20 µm. Paste made with powdered sugar can go straight to the five-roll.

Roll refining is the most common way to reduce particle size in commercial chocolate manufacturing, according to the Penn State team. It runs continuously, and the published capacities of industrial five-roll refiners from two maker groups run from 850 to 1,600 kg per hour, at a quoted fineness of 20 or 22 µm. Smaller three-roll refiners are used in laboratories.

The output is the catch. Refined flake is a dry powder of sugar and cocoa clumps with the fat trapped inside them. Conching breaks those clumps up and releases the fat, as a 2024 study by Guckenbiehl and colleagues describes. A five-roll line therefore needs a separate conche.

Refiner-conche or separate conche

A refiner-conche does mixing, refining and conching in one vessel. One maker describes a shaft with grinding bars and an adjustable pressure mechanism: it turns without pressure to mix, with pressure on the grinding bars to refine, and with low or no pressure to conch. Another calls its version a “3 in 1 machine”.

Published batch sizes run from 45 to 6,000 kg for one maker group and 50 to 1,500 kg for another. The second quotes a fineness of 25 to 28 µm and refining times of 8 to 12 hours depending on size. Price belongs in the comparison too: a 2019 thesis by Clark gave $10,000 to $12,000 for a refiner-conche handling 20 kg batches, against around $2,400 for a melanger of the same volume. Those are 2019 prices.

The trade here is control against simplicity. A separate refiner and conche let you set fineness and conching time independently, and a five-roll refiner forces that split anyway. A refiner-conche means one machine to buy, house and clean, at the cost of running both steps in the same batch.

Ball mills

A ball mill refines by stirring the chocolate mass through a vessel packed with small balls. Arms on a rotating shaft keep the balls moving, they grind the particles by compression and shear, and the mass is circulated back through until it reaches the target fineness.

How they work

In the description by Stožinić and colleagues (2024), the grinding vessel is a vertical or horizontal cylinder with a central shaft and arms. Grinding media fill about 60 to 80% of its volume, as balls of steel, ceramic or another material, and a recirculation system returns the mass to the mill until it is fine enough. Temperature control melts the solid fat at the start, protects the product from heat damage, and can take over part of the work of traditional conching.

Most of the size reduction happens early. In a pilot ball mill making milk chocolate, Lucisano and colleagues saw particle size drop effectively over the first 120 minutes and then level off. A 2016 study on compound chocolate by Toker and colleagues found that a faster shaft and a longer run gave smaller particles and higher viscosity. When the Penn State team needed chocolates finer than 30 µm D90, beyond what their laboratory roll refiner could reach, they used a ball mill.

Where they fit

Ball mills cover a wide range of scales on paper. Makers list laboratory units with 50 kg batches, batch mills from 25 to 1,000 kg, and continuous industrial installations of 500 to 1,800 kg per hour in which two or three mills run in series. Published batch times range from about an hour and twenty minutes to 8 to 12 hours when the machine also conches. Some makers pitch them at spreads, creams and ice cream coatings as well as chocolate.

Settings matter, too. When Alamprese and colleagues retuned the settings on an industrial ball mill, output rose from 109 to 156 kg per hour, and energy consumption was one of the factors they tracked.

Flow is where ball mill and roller chocolate can part ways. Yield value is the push melted chocolate needs before it starts to move, and it depends mainly on how the particles interact, Hinneh and colleagues explain. Two small studies compared a ball mill route with five-roll refining plus conching, and both found higher yield values from the ball mill. In a cocoa topping (a compound product), Stožinić and colleagues measured yield stress twice as high for the milk version and 4.5 times as high for the dark one, while the ball mill took 1.5 hours against 2.5 hours for the roll-and-conche route. In dark chocolate, a 2019 study by Saputro and colleagues found higher Casson yield values in ball-milled samples, with similar Casson viscosity.

Two small studies settle little. One was a compound topping, and the fat content of a recipe can outweigh particle size: in Hinneh’s work, the finest sample at 43% fat also had the lowest viscosity. If flow matters for your molding or enrobing, test your own recipe on the mill you’re considering. For what separates compound from real chocolate, see real vs compound chocolate.

Melangers, roller refiners and ball mills compared

The three technologies differ most in how they run (batch or continuous) and in what has to happen after them. On energy use and cleaning the sources hold only makers’ claims, so those rows stay qualitative.

The melanger is the low-cost route and pays for it in time and batch size. The five-roll line is built for throughput and commits you to a separate conche. The ball mill spans the most scales and deserves a close look if you also make spreads or creams, provided you check how your chocolate flows afterward.

What makers’ published figures say

The figures in this section are makers’ own claims, as published on October 11, 2026. They are not independent measurements and shouldn’t be read as the norm for a category. Four of the brands behind them belong to one equipment group, so they count here as a single maker group. Set side by side, the claims can’t be ranked: the melanger figures come without a measuring method, one refiner figure is given for dark chocolate only, and as the section on measuring shows, a fineness figure without its method and recipe can’t be compared with another.

On roller refiners, one maker’s five-roll refiner page gives no capacity or fineness figure and claims up to 15% less energy consumption. The maker group’s five-roll refiner is listed at 900 to 1,500 kg per hour at 22 µm for dark chocolate, with a fineness range of 16 to 60 µm, and a third maker’s refiner series at 850 to 1,600 kg per hour at 20 µm.

On ball mills, one maker’s 2008 article gives 25 to 6,000 kg per batch and a batch time of normally 8 to 12 hours including conching. A second maker lists a model for up to 25 kg and one for up to 100 kg, both under 20 µm, the larger one in about an hour and twenty minutes. A third maker’s batch ball mill runs 50 to 1,000 kg at 25 to 28 µm in 4 to 6 hours. The maker group lists a continuous ball mill and an all-in-one unit at 500 to 1,800 kg per hour, below 20 µm, plus a laboratory unit with a 50 kg batch.

On melangers, the two makers linked in the melanger section list a model at 4.5 kg and about 15 µm, and one at up to 3 kg and less than 15 µm.

On refiner-conches, the maker group’s refiner-conche page lists 45 to 6,000 kg batches with no fineness or time, and the third ball mill maker’s refiner-conche lists 50 to 1,500 kg, 25 to 28 µm and 8 to 12 hours.

Measuring particle size

Laboratories measure particle size with laser diffraction and report the D90. Small makers often work with a micrometer or a grindometer instead, and those tools read different things.

Laser diffraction gives the full size distribution, but the instruments are expensive and out of reach for most small manufacturers, Tan and Balasubramanian note. A micrometer measures a pinched sample of chocolate, so it only catches the largest particles in it, as Dandelion Chocolate’s grindometer post explains. A grindometer is a stainless steel gauge that is read by eye. Clark’s team used a 0 to 50 µm grindometer and reported a range for the majority particle size.

The tools don’t agree. When Tan and Balasubramanian compared methods on melanger-refined cocoa, the micrometer read larger than every other method after 30 minutes of refining, and smaller than every other method after 2, 4 and 24 hours. They found the micrometer and light microscopy both usable for tracking a batch, while the Hegman gauge (a type of grindometer) failed to give meaningful measurements.

So a number without its method is hard to use. A “20 µm” from a micrometer, a grindometer and a laser D90 are three different readings.

How to choose: questions to ask a vendor

Start with your volume and your recipe, then put the same questions to every vendor so the answers line up.

• What fineness, measured how? Ask whether it’s a laser D90, a micrometer reading or a grindometer reading, and on which recipe. One maker’s own spec sheet notes that its throughput figure refers to dark chocolate and varies with the product.
• What throughput at my target fineness? Published capacities are quoted at a specific fineness. Ask for the figure at yours.
• What has to happen after refining? Find out whether the output needs a separate conche, can go to a liquid conche, or is conched in the same machine.
• How is temperature controlled? Friction heats chocolate during refining, and not every machine controls it.
• Energy per kilogram, against what baseline? With no independent comparison across the three technologies to lean on, “up to X% less” claims need a reference point and a recipe.
• How long is a recipe changeover and a full clean? Ask what comes apart and what has to be flushed out.
• Can we run a trial with our own recipe? Whether you make a dark couverture, a milk chocolate or a compound coating, fat content and ingredients change what a refiner can do.

Where refining sits in the line

Refining comes after mixing and before conching, and tempering comes later, just before molding or enrobing. In a classic industrial line, a mixer feeds a two-roll pre-refiner, then a five-roll refiner, and the flake goes on to a conche. In a ball mill line, the mixed mass goes through the mill and on to a liquid conche, or the mill does part of the conching itself. In a melanger workshop, nibs and sugar go into one machine and often come out refined and partly conched.

Conching then works on what the refiner leaves behind. Its texture work, in Guckenbiehl’s account, is mainly breaking up clumps of particles and releasing the fat trapped inside them. What tempering then does to the cocoa butter is explained in what tempering is and why it matters. The full sequence for a small maker, from incoming beans to packaging, is in the bean-to-bar process.

Sources and open questions

The 30 or 35 µm upper limit and the 15 to 25 or 15 to 30 µm targets also appear in Hinneh et al., 2019a, Homayouni Rad et al., 2019 and Stožinić et al., 2024. The conching account of clump breakup and fat release is also supported by Blanco et al., 2019.

Still open:

• an independent comparison of energy use or cleaning across melangers, roller refiners and ball mills;
• a maker’s own page for a five-roll refiner sized for bean-to-bar volumes (we couldn’t verify one);
• whether the 2008 ball mill maker document still reflects that maker’s current range.