THE PROCESS IN MOTION

How does the drum mix the coffee?

Start the scene or choose a stage. The animation simplifies the mechanism; what follows gives the details and the available references.

▶ Play the process01Too slow: the bed slides02Rolling, then cascading03Paddles: the curtain in the hot air04Too fast: cataracting05Fr = 1: pinned against the wall

Too slow: the bed slides

In a drum that turns too slowly, or whose wall is too smooth to carry the coffee along, the bed does not roll: it slides as a block against the metal. The same beans then keep touching a wall preheated to 250 or 300 °C, and this prolonged contact singes one of their faces, the defect known as scorching.

Read the stages without animation

1. Too slow: the bed slidesIn a drum that turns too slowly, or whose wall is too smooth to carry the coffee along, the bed does not roll: it slides as a block against the metal. The same beans then keep touching a wall preheated to 250 or 300 °C, and this prolonged contact singes one of their faces, the defect known as scorching.
2. Rolling, then cascadingFaster, and the wall carries the bed upward and its surface starts to flow: that is rolling, then cascading when the surface bulges. These regimes are compared with the Froude number, the centrifugal force relative to gravity; rolling and cascading lie between 0.0001 and 0.1. The surface is renewed, but the bean stays in the bed, against the wall and its neighbors.
3. Paddles: the curtain in the hot airThe paddles, vanes fixed inside the drum, lift the coffee above the bed and release it in a curtain. In the bed, the bean moves at less than 0.5 m/s and receives heat from the wall; in flight, it travels at 0.5 to 2 m/s through hot air, which heats it by convection. In a 16 cm drum, 42 and 56 revolutions per minute give a Froude number of 0.16 and 0.28.
4. Too fast: cataractingAt 78 revolutions per minute, the Froude number reaches 0.54: the beans no longer fall back in a curtain, they are thrown across the drum, which is cataracting. A bean that gets lighter as it roasts flies farther at the same speed. With a small charge, more beans are in flight, others stay longer against the wall, and the batch heats less evenly.
5. Fr = 1: pinned against the wallWhen the centrifugal force equals gravity, the coffee no longer falls back: it turns pinned to the wall, with no mixing at all. This critical speed is 42.3 divided by the square root of the diameter in meters, in revolutions per minute: 106 for a 16 cm drum, 60 for 50 cm, 42 for one meter. A large drum must therefore turn more slowly than a small one to mix in the same way.

To remember: Drum speed decides where the bean spends its time, against the metal, in the bed or in the air, and so where its heat comes from.

In history: Evans's cylinder already carried vanes in 1824; the path of a bean inside a drum was only measured in 2023, by following a tagged bean.

Read the full article and its sources ↓

The section on the profile names three levers: heat, airflow and drum speed, “which sets the contact with the wall.” The first two have their literature, their curves and their quarrels. The third passes for an installation setting, fixed once and forgotten. Yet it is the one that decides something the probe cannot see: where the bean is during the roast, and therefore by what route heat reaches it.

A bean in a drum knows only three situations. It is buried in the bed, the mass of coffee lying at the bottom of the cylinder, squeezed between its neighbors. It is against the wall, the metal that the flame heats from below. Or it is in the air, released by a paddle, crossing the hot air that sweeps the drum. Section 05 described the three modes of heat transfer: conduction by contact, convection by the air, radiation from the wall and the flame. Mixing is the link between them: it splits the bean's time among the places where each of these modes dominates.

Plate VII, sections 05 and 11.

What we saw by following a bean

For a long time, nobody could look inside a closed drum. The published studies filmed beans in transparent drums, in two dimensions and with low precision: Cristo and colleagues in 2006, Resende and colleagues in 2017. In 2023, a team from the University of Birmingham, with an engineer from Jacobs Douwe Egberts, followed a bean tagged with a radioactive tracer in a small roaster, an Aillio Bullet with a drum 16 cm in diameter. The technique, positron emission particle tracking, comes from medical imaging: a camera locates the tagged bean continuously, through the steel wall, and reconstructs its path.

The drum was not heated. The authors reproduced the roast by loading three coffees prepared in advance, one green, one at mid-roast and one roasted, all from the same Kenyan arabica, with three charges, 300, 600 and 900 g, and three speeds, 42, 56 and 78 revolutions per minute. The reservation matters, and we will come back to it: what is observed is movement, not roasting.

The result fits in one picture. The drum holds two regions that do not resemble each other. A dense bed, where the beans touch and move slowly, from 0 to 0.5 meters per second, carried along by the wall. And an almost empty region, where the beans released by the paddles fly at 0.5 to 2 meters per second. On each turn, a bean rises about 16 cm, the whole diameter of the drum: the paddles carry it to the top before releasing it.

These two regions do not heat in the same way. In flight, the bean is surrounded by hot air, and convection dominates. In the bed, air passes poorly between the beans; heat enters mostly by conduction, from the wall for the beans that touch it, then from bean to bean. Now the wall is hot: preheated to 250 or 300 °C according to the study the authors cite, when the probe reads 180 to 200 °C at the charge. A bean resting against this metal heats very quickly, and the authors compare this case with that of roasted malts, where most of the heating takes place in the wall zone.

Al-Shemmeri et al. 2023, § 1, 2.2, 2.3, 3.3.1 and 3.4.1; Garcia et al. 2018, Cristo et al. 2006 and Resende et al. 2017, cited by them.

A number for comparing drums

To describe a rotating drum, the engineers of rotary kilns, dryers and mixers use a dimensionless number, the Froude number. It relates the centrifugal force that rotation exerts on a bean resting against the wall to the gravity that makes it fall back: Fr = ω²R/g, where ω is the speed of rotation in radians per second, R the radius of the drum and g the acceleration of gravity, 9.81 m/s². A Froude number of 0.1 means the wall holds the bean with a tenth of its weight; a Froude number of 1, that it pins it with a force equal to its weight, enough that it no longer falls back, even at the top.

The value of this number is that it compares drums of different sizes. At equal fill, two drums with the same Froude number mix their charge in a similar way: this is the scaling rule that Henein, Brimacombe and Watkinson drew in 1983 from their tests on rotary kilns, with a correction for the size of the grains relative to the diameter of the cylinder. It has a consequence the trade knows without stating it: a large drum must turn more slowly than a small one. At the speed where a 16 cm drum gently mixes its coffee, a one-meter drum would already pin it against the wall.

Al-Shemmeri et al. 2023, § 2.4.4, equations 3 and 4; Henein, Brimacombe and Watkinson 1983; Mellmann 2001.

Six regimes, from slowest to fastest

In 2001, Jochen Mellmann classified the movements of a charge of grains in a rotating cylinder without paddles. The classification holds for sand, ore or coffee, and is read with two quantities, the Froude number and the fill ratio.

Slipping comes first: if the wall does not carry the charge along enough, for lack of friction, the bed stays in one piece and the wall slides underneath. The same beans always rub against the metal. Then slumping: the wall lifts the bed, which collapses in jolts. Then rolling, a continuous regime where the surface of the bed, flat and inclined, flows without stopping while the bottom rises with the wall; the bed is renewed, and every bean eventually passes through the surface. Faster, the surface bulges into a bean shape: this is cascading. Faster still, the beans at the top leave the bed and are thrown across the drum: this is cataracting. Finally, when the Froude number reaches 1, the whole charge stays pinned against the wall and turns with it: this is centrifuging, where nothing mixes anymore.

The bounds that the Birmingham study takes from Mellmann are wide: rolling and cascading between 0.0001 and 0.1, transition to cataracting around 0.1. Centrifuging needs no measurement: it begins, by definition, when the centrifugal force equals gravity at the top of the drum, so at a Froude number of 1.

Mellmann 2001, cited by Al-Shemmeri et al. 2023, § 1 and 2.4.4.

Paddles change the game

A roaster is not a smooth cylinder. It carries paddles, vanes fixed to the wall or to the shaft, and the section on technologies says so in one sentence: they “lift the coffee and let it fall back.” The cylinder of Richard Evans, the first English patent for a commercial roaster in 1824, already had its own. In the Bullet studied in Birmingham, three paddles have three roles: lifting the bean at the bottom of the drum, mixing it along its length, and pushing it toward the door, which allows the drum to be emptied at the end. With them, the bean leaves the bed even when the Froude number is low: it is carried upward and released in a curtain through the hot air.

On this 16 cm drum, 42 revolutions per minute give a Froude number of 0.16 at the wall, 56 revolutions 0.28 and 78 revolutions 0.54. With small or medium charges and a low or medium speed, the mid-roast coffee sits between cascading and cataracting. The authors also note that speed weighs most on the share of coffee remaining in the bed between 56 and 78 revolutions, where the regime passes from cascading to cataracting.

For the order of magnitude of a professional machine: a study conducted on a professional roaster, with 38 roast profiles, ran the drum at 40 and at 65 revolutions per minute. A speed cannot be compared without the diameter, and that is what the Froude number allows. Here, drawn from the formula, is the critical speed at which the coffee is pinned against the wall: 42.3 divided by the square root of the diameter in meters, in revolutions per minute. It is 106 revolutions for a 16 cm drum, 60 revolutions for 50 cm and 42 revolutions for one meter. A one-meter drum turning at 42 revolutions per minute would no longer mix anything.

One last figure circulates, and it must be cited with its limits. Steven Abbott's calculation app on drum roasting holds that a Froude number “a little below 1” is optimal, without giving a source. The Birmingham study finds a working regime between 0.16 and 0.54 on a small drum with paddles, and notes that small charges at high speeds, in cataracting, increase the risk of defects. The two do not necessarily contradict each other, since paddles change everything; no general rule can be drawn from either.

Al-Shemmeri et al. 2023, § 2.2, table 3, § 3.3.2 and 3.4.3; Catão et al. 2022; Abbott, Coffee Drum Roasting; plate VII, sections 04 and 08; critical speed calculated for Fr = 1.

The bean gets lighter, the regime changes

The most useful result for a roaster lies elsewhere. At equal speed and charge, the three coffees do not move the same way. The green, dense, is in cascading: the beans fall back rolling over the surface of the bed. The mid-roast coffee sits between cascading and cataracting. The roasted, lighter, passes into cataracting: it is thrown much farther, and its bed is much smaller. Since the bean doubles in volume and loses half its density at first crack (section 05), the same batch changes regime during its roast, without anyone touching the speed.

The authors draw two pieces of advice. Adapt the drum speed to the density of the coffee, and vary it during the roast: they propose working near the capacity of the machine, at moderate speeds that decrease over the course of the batch. And charge by volume rather than by weight: two green coffees of different densities, charged at the same mass, do not occupy the same volume, do not have the same bed and do not heat in the same way. This advice is deduced from a study of movement, not from cup trial results.

Al-Shemmeri et al. 2023, § 3.3.2, 3.4.3 and 3.4.4; plate VII, section 05.

Too slow, too fast: the defects

Too slow, the drum leaves the same bean too long against the hot wall. The section on defects names this case: scorching, one face left pressed against the metal instead of rolling, caused by a bent vane, a drum that turns too slowly or a charge that is too large. The section on technologies already said as much of the direct-heat drum: it scorches the bean when the wall is too hot or the drum too slow, that is, when mixing no longer suffices to keep the bean away from the metal.

Burnt tips, or tipping, have another main cause in the plate: a charge that is too hot at the start, which carbonizes the thinnest part of the bean. Contact plays a role there too, however: according to the Book of Roast, the double drum was introduced against tipping caused by the localized overheating of the perforated cylinder (section 08).

Too fast, the problem changes in nature. In cataracting, with a small charge, more beans are in flight and those lifted by the wall stay there longer; the Birmingham study infers an increased risk of defects, through stronger convection in flight and localized conduction near the wall, which make the batch less homogeneous if the profile is not well held. And at the end, at the critical speed, centrifuging eliminates mixing: the charge turns stuck to the metal.

Plate VII, sections 08 and 13; Al-Shemmeri et al. 2023, § 3.4.4.

What is not yet known

None of the studies cited links drum speed directly to taste. The Birmingham study follows beans in a cold drum. The study on 38 profiles varies the speed among five settings and finds that the time elapsed after the crack weighs most on the measured aromas and on the cup score, to the point that about thirty seconds cost a few points on the Specialty Coffee Association scale; nothing in its abstract places speed first. Mixing therefore remains what section 11 says of it, a lever that sets the contact with the wall, whose mechanism is now better known than its effect in the cup.

Catão et al. 2022; Al-Shemmeri et al. 2023, § 4.

The sources for this sheet

1. Al-Shemmeri M., Windows-Yule K., Lopez-Quiroga E., Fryer P. J., “Coffee bean particle motion in a rotating drum measured using Positron Emission Particle Tracking (PEPT)”, Food Research International, vol. 163, 2023, article 112253, doi:10.1016/j.foodres.2022.112253, CC BY license.
2. Mellmann J., “The transverse motion of solids in rotating cylinders: forms of motion and transition behavior”, Powder Technology, vol. 118, no. 3, 2001, pp. 251-270, doi:10.1016/S0032-5910(00)00402-2; read through the citation made of it by the preceding reference.
3. Henein H., Brimacombe J. K., Watkinson A. P., “Experimental study of transverse bed motion in rotary kilns”, Metallurgical Transactions B, vol. 14, no. 2, 1983, pp. 191-205, doi:10.1007/BF02661016 (abstract).
4. Cristo H. P., Martins M. A., Oliveira L. S., Franca A. S., “Transverse flow of coffee beans in rotating roasters”, Journal of Food Engineering, vol. 75, no. 1, 2006, pp. 142-148; cited by Al-Shemmeri et al.
5. Resende I. A., Machado M. V. C., Duarte C. R., Barrozo M. A. S., “An experimental analysis of coffee beans dynamics in a rotary drum”, Canadian Journal of Chemical Engineering, vol. 95, no. 12, 2017, pp. 2239-2248; cited by Al-Shemmeri et al.
6. Garcia C. C., Netto A. D. P., Da Silva M. C. et al., 2018; cited by Al-Shemmeri et al. for the temperature of the wall and of the charge.
7. Catão A. A., Mateus N. de S., Garcia C. da C., da Silva M. C., Novaes F. J. M., Alves S., Marriott P. J., da Silva A. I., “Coffee-roasting variables associated with volatile organic profiles and sensory evaluation using multivariate analysis”, Applied Food Research, vol. 2, no. 2, 2022, article 100223, doi:10.1016/j.afres.2022.100223 (abstract).
8. Abbott S., “Coffee Drum Roasting”, online app from the Practical Mechanical collection, stevenabbott.co.uk, accessed October 21, 2026.
9. PXL Roasters, plate VII, sections 04, 05, 08, 11 and 13, and the sources they cite, including Scott Rao, The Coffee Roaster's Companion, 2014, and Jens Roelofs in The Book of Roast, 2017.