Sample roasters
Small roasters that turn green-coffee samples into cupping roasts under conditions kept as equal as possible, so the cup shows the coffee rather than the roast. Drum and hot-air designs; no standard fixes their size.
Across the chain: Equipment
In brief
- A sample roaster roasts small lots of green coffee for cupping and grading; ISO 6668 specifies a method for roasting green coffee for sensory analysis.4
- Specialty Coffee Association (SCA) Standard 102-2024 sets the cupping roast at ground-coffee L* 26–29, or Agtron “Gourmet” 63, and allows any technology that reaches it without roasting problems.1
- The superseded protocol of the SCA’s predecessor (SCAA) required 8–12 minutes per roast; SCA 102-2024 sets no time but rejects too-fast and too-slow roasts.1,3
- Samples are air-cooled without water and rest 8–24 hours before cupping.1
- The Coffee Value Assessment asks for the same roaster, sample size, load temperature, power, cooling and resting time across samples, so the roast adds as little variation as possible.2
Contents
What a sample roaster is for#
Green coffee cannot be tasted as it is. A buyer or a grader has to roast it, grind it and brew it first.2 A sample roaster does the first step on a small scale. It turns a sample of a lot into a roast that can be cupped. A lot is a batch of coffee from one origin and process, traded as a unit.
ISO 6668 specifies a method for roasting green coffee, grinding it and preparing a brew for sensory analysis.4 Its result may decide whether a shipment is accepted, if the parties agree.4 The same procedure, repeated identically, also serves to compare several samples.4
The aim differs from production roasting. According to the Specialty Coffee Association (SCA), sample roasting is not about bringing out the best flavour of each coffee.2 It should show the green coffee as it is. Differences in the cup should come from the coffees, not from the roast.1,2
Designs and heat sources#
No standard prescribes a design. SCA 102-2024 allows different roasting technologies and conditions, as long as the target roast is reached without roasting problems.1 The sources consulted for this entry describe two families.
Drum sample roasters. A small drum turns over a heat source. One 2023 study roasted 200 g batches in a gas-fired drum sample roaster; the beans dropped from the drum and a fan cooled them.6 Another 2023 study, at Colombia’s national coffee research centre, used an electric sample roaster; the paper does not describe its design.7 How a drum moves and heats beans is explained in Drum roaster.
Hot-air sample roasters. Here a stream of hot air moves and heats the beans. A 2000 doctoral thesis describes a laboratory fluid-bed roaster with electric heaters and a probe just before the chamber.9 Roasting and cooling took place in separate sections.9 The physics is set out in Fluid-bed roasters.
A 2017 US patent, assigned to IKAWA Ltd, describes a small hot-air roaster.10 An electric filament heats the air, and a fan blows it tangentially into a chamber that widens towards its outlet.10 The beans circulate around the chamber wall.10 A cyclone separates the chaff into a collector, and cooler air cools the beans at the end.10 The patent describes a domestic roaster and does not mention sample roasting.10 No source consulted for this entry links its design to a particular sample roaster.
- Drum. A small turning drum holds the batch; one study used 200 g.6
- Heat under the drum. Gas in one study.6
- Probe and logger. Time and temperature are logged during the roast.6
- Colour check. In a second study, on an electric sample roaster, beans were compared by eye with a colour chart during the roast.7
- Cooling by fan. The beans drop from the drum and are cooled with air.6
- Fan and electric heater. Air is driven by a fan and heated by an electric filament.10
- Tangential inlet. Hot air enters along the chamber wall.10
- Conical chamber. It widens towards the outlet; the beans circulate around its periphery.10
- Cyclone and chaff collector. The collector sits just below the cyclone outlet.10
- Cooling air. After the roast, cooler air is blown through the beans.10
No source consulted for this entry describes an infrared sample roaster, so none is described here.
How a sample roast is run#
SCA 102-2024 sets the rules for cupping roasts in the Coffee Value Assessment (CVA). The target is a “medium” roast.1 In colour terms this is a ground-coffee lightness (L*) of 26–29, close to tile #65 of the Agtron/SCA colour discs.1 On a roast meter, the target is 63 on the Agtron “Gourmet” scale or its equivalent.1
The standard sets no roast time and no batch size. Instead it asks that roasts avoid “roasting problems”: too fast, too slow, underdeveloped or burnt.1 A batch outside the target range, or with such faults, is not cupped. It is roasted again.1
After the roast, the steps are fixed:
- Cool the sample at once with air, without water.1
- At room temperature (about 20 °C), store it in an airtight container or a non-permeable bag, in a cool, dark place.1
- Let it rest 8–24 hours before cupping.1
The full cupping protocol that follows is in Cupping.
Keeping conditions equal. The CVA document asks for the same roaster, sample size, load temperature, power, final roast level, cooling time and resting time across samples.2 It advises against changing the power during a roast.2 If samples differ too much in density or moisture, they can be sorted into two or three groups, each roasted under its own conditions.2 If the roaster is too small for the coffee needed, several small batches may be roasted and combined.2
Checking the roast#
Colour. SCA 102-2024 names colorimetry, such as CIELAB or the Agtron/SCA discs, and infrared spectrometry, such as an Agtron device.1 When a roast meter is used, grind level and coffee temperature should be controlled.1 Samples with much chaff give less reliable readings.1 Meter scales are arbitrary, and different brands do not always agree even on the same scale.2 Colour scales are covered in Roast degree and colour.
Roast loss and volume. Dry-matter loss or bean volume increase may replace colour, once their relation to colour or spectrometry is established.1 Roast loss is explained in its own entry.
Eye and cup. The CVA document knows no simple measurement that detects too-fast or too-slow roasts across technologies.2 Cuppers can be trained to find them from bean structure, colour gradient and flavour.2 The International Trade Centre’s guide adds that a light roast shows immature beans as yellowish pale rather than brown.5 Such pale beans, called quakers, are covered in Roast defects.
Reference values#
Values specified by standards, then values reported for particular machines. Study values are not targets.
| Parameter | Value | Source |
|---|---|---|
| Cupping roast, colour | Ground coffee L* 26–29 (≈ Agtron/SCA tile #65) | SCA 102-20241 |
| Cupping roast, roast meter | Agtron “Gourmet” 63; Agtron “Commercial” 48; Probat Colorette 3b 96; Colortrack 62 | SCA 102-2024; CVA document1,2 |
| Same targets, superseded SCAA protocol | Same four readings, tolerance ± 1.0 units (Agtron “Gourmet” 62–64); measured on ground coffee at room temperature, 30 min–4 h after roasting | SCAA protocol, rev. 20153 |
| Roast time | Not set (2024); 8–12 min (superseded SCAA protocol, rev. 2015) | SCA 102-2024; SCAA protocol1,3 |
| Roasting technology | Any, if the target is reached without roasting problems | SCA 102-20241 |
| Alternative roast-level control | Dry-matter loss or bean volume increase, once correlated with colour or spectrometry | SCA 102-20241 |
| Cooling | Immediately, with air; no water quench | SCA 102-20241 |
| Storage | Airtight container or non-permeable bag from ≈20 °C; cool, dark place | SCA 102-20241 |
| Rest before cupping | 8–24 h (2024); roasted within 24 h of cupping, rest at least 8 h (superseded SCAA protocol) | SCA 102-2024; SCAA protocol1,3 |
| Held constant across samples | Roaster unit, sample size, load temperature, power, final roast level, cooling time, resting time | CVA document2 |
| Samples too different | Sort into 2 or 3 groups, same conditions within each group | CVA document2 |
| Roast degree by purpose | Light for defects; medium for flavour and colour | ISO 6668 (catalogue abstract)4 |
| Gas drum sample roaster, one study | 200 g batches; charge 177.9–181.1 °C per roast in Table 1 (the methods state 170 ± 1 °C); drop times 410–517 s in Table 1 (the text states 400–501 s and 420–501 s); colour a mean of 65 on the Agtron SCA scale in the text, 75–95 “Colour (SCA)” per roast in Table 1; fan cooling, 180 s on average in the methods and 120 s in the results | Campo-Ceballos & Gaviria-López6 |
| Electric sample roaster, one study | Initial temperature 200 °C; 8–12 min to the required colour; colour checked by eye against the Agtron/SCA chart; 85, 55 and 35 on that scale | Gómez Parra et al.7 |
| Earlier studies of titratable acidity (as summarised in a 2024 paper) | Mostly isothermal roasting in small-scale roasters, 50–250 g batches | Anokye-Bempah et al.8 |
From fixed time to fixed outcome#
The SCAA protocol, last revised in December 2015 and superseded by SCA 102-2024, required every cupping roast to take 8–12 minutes.3 It also asked that scorching or tipping should not be visible.3 Its roast-meter targets carried a tolerance of ± 1.0 units.3
SCA 102-2024 dropped the time window. It defines roasting problems relative to the “optimum” rate of a given roasting technology instead.1 According to the CVA document, a roast that is too fast leaves the centre of the bean behind the surface, and a roast that is too slow gives a “baked” bean with a brittle structure, subdued flavour and often grain-like notes.2 The two documents therefore disagree on whether one time window fits all machines (see Contested).
Data table
| min | Source | |
|---|---|---|
| SCAA protocol (rev. 2015, superseded), required | 8–12 min | 3 |
| Electric sample roaster, 300 roasts (Cenicafé, 2023) | 8–12 min | 7 |
| Gas drum sample roaster, 200 g, 13 roasts (2023) | 410–517 s (Table 1) | 6 |
| Small drum, 0.8–1.2 kg capacity, light to dark (2022) | 8.67–11.25 min | 11 |
| Small drum, 2.0 kg capacity, light to dark (2022) | 12.98–17.75 min | 11 |
Reported times sit inside or near the old window. A study at Colombia’s national coffee research centre (Cenicafé) roasted 100 coffees to three degrees each on an electric sample roaster. Each roast took 8–12 minutes, timed by the colour needed.7 In 13 roasts of 200 g on a gas sample drum, the logged drop times were 410–517 s (about 6.8–8.6 min).6 The paper’s text gives 400–501 s in one place and 420–501 s in another; this entry uses the times logged in its Table 1, because they are what the roasts took.6 Two small drum roasters heated with liquefied petroleum gas (LPG), with capacities of 0.8–1.2 kg and 2.0 kg, needed different times for the same named profiles: 8.67 and 12.98 min for “light”, 11.25 and 17.75 min for “dark”.11 Time to a given degree thus belongs to the machine.
Roast degree for cupping and for research#
The cupping target is one point on the colour scale: L* 26–29 on ground coffee, or 63 on the Agtron “Gourmet” meter scale.1 ISO 6668, according to its catalogue abstract, generally asks for a light roast to assess defects and a medium roast for flavour and colour.4 The trade guide gives the reason: lighter roasts make quality and off-flavours easier to detect, and they show immature beans.5
Research uses the same machines for other purposes. The Cenicafé study roasted each of 100 coffees to three degrees, 85, 55 and 35 on the Agtron/SCA chart, judged by eye.7 The 200 g gas-drum study reports a mean of 65 on the Agtron SCA scale, measured with a colour meter.6 Its Table 1, however, lists per-roast values of 75–95 in a column headed “Colour (SCA)”, so the roast degree of that study is unclear and it is not shown in the figure below.6
Data table
| Point | Agtron-type number (higher = lighter) | Source | |
|---|---|---|---|
| SCA 102-2024 cupping target (Agtron “Gourmet” meter) | 63 | 63 | 1 |
| SCAA protocol (rev. 2015, superseded): 63.0 ± 1.0 | 62–64 | 62–64 | 3 |
| Electric sample roaster, three degrees (chart, by eye) | 35 | 35 | 7 |
| Electric sample roaster, three degrees (chart, by eye) | 55 | 55 | 7 |
| Electric sample roaster, three degrees (chart, by eye) | 85 | 85 | 7 |
Sample roasts and production roasts#
Sample roasts and production roasts differ in scale and in control. A 2024 paper notes that most earlier studies on titratable acidity used small-scale roasters with 50–250 g batches at constant temperature.8 Its authors state that such roasts do not necessarily reflect commercial-scale roasting, where the temperature curve over time is the key control.8 They name batches above 1 kg as commercial scale.8
The purposes differ too. A production roast is developed for the flavour a roaster wants to sell. A cupping roast keeps conditions equal across samples on purpose.2 A coffee may therefore taste different in a sample roast than in a production profile (our reading of the two aims).
Direct evidence is thin. In a 2022 study, two small drum roasters gave cupping scores that did not differ significantly. Infrared spectra still showed a separation trend between the two roasters.11 Neither machine was described as a sample roaster.11 No study consulted for this entry compared cupping scores from a sample roaster with scores from a production roaster for the same lots.
Risks and defects#
- Roasting problems. Too-fast, too-slow, underdeveloped or burnt batches add “noise” to the test; the standard says to repeat them.1,2 Their appearance and causes are described in Roast defects.
- Changing settings between samples. Small changes in roasting parameters may have a large effect on flavour, according to the CVA document; it advises constant power.2
- Uneven samples. Small, broken and light beans roast faster than whole and solid beans, so uneven bean size gives uneven roasts.5
- Water quenching. It is not allowed for cupping samples.1 Its effects on beans are covered in Cooling and quenching.
- Colour readings. Chaff, grind level and coffee temperature affect roast-meter readings.1 Meters of different brands do not always correlate.2
- Mixed batches. When several small batches are combined, the descriptive and the affective assessments should use the same batch or the same combination.2
Open questions#
- Sample roast and production roast. No study consulted for this entry compared cupping results for the same lots roasted in a sample roaster and in a production roaster.
- Repeatability. No study consulted for this entry reported batch-to-batch colour or roast-loss variation of a sample roaster, or the variation between the drums of one machine.
- Design and cup. No peer-reviewed comparison of drum and hot-air sample roasters on cupping scores was found.
- Meter correlation. The SCA states that it is researching how its roast-level parameters correlate.2
- Infrared and other designs. No source consulted for this entry describes infrared sample roasters or measures their performance.
Key studies#
Campo-Ceballos & Gaviria-López, 2023 – Thirteen profiles on a gas sample drum. Roasts 200 g batches of one Cauca coffee on a gas-fired drum sample roaster with logged time and temperature, in a response-surface design. Table 1 logs drop times of 410–517 s, while the text gives 400–501 s and 420–501 s; the text reports a mean colour of 65 on the Agtron SCA scale, Table 1 per-roast values of 75–95. The roasts were fan-cooled; consumers rated acidity and body.
Gómez Parra et al., 2023–300 sample roasts for near-infrared calibration. Roasts 100 Colombian coffees to three degrees on an electric sample roaster, starting at 200 °C. Each roast lasted 8–12 minutes, stopped by comparing the beans by eye with the Agtron/SCA colour chart.
Anokye-Bempah et al., 2024 – Small roasters versus commercial profiles. Studies titratable acidity on a 5 kg gas drum loaded with 4 kg. The introduction notes that most earlier work used small-scale roasters with 50–250 g batches at constant temperature, which may not reflect commercial roasting.
Debona et al., 2022 – Two small drums, same profiles. Roasts one arabica to baked, light, medium and dark profiles on two small LPG-heated drum roasters of different size. The two machines needed different times; Q grader scores did not differ significantly, while infrared spectra separated the profiles and showed a separation trend between the roasters.
- Fixed roast time or a rate that suits the machine: the superseded SCAA protocol (revision of 2015) required every cupping roast to take 8–12 minutes3; SCA 102-2024 sets no time and judges “too fast” and “too slow” against the optimum rate of the specific roasting technology, and the CVA document states that no simple measurement controls these problems across technologies and coffee types.1,2
Coffearium (2026). Sample roasters. Coffearium – The open coffee compendium. Retrieved 17 November 2026, from https://coffearium.org/entries/sample-roasters/
@misc{coffearium2026-sample-roasters,
author = {{Coffearium}},
title = {{Sample roasters}},
year = {2026},
howpublished = {Coffearium -- The open coffee compendium},
url = {https://coffearium.org/entries/sample-roasters/},
urldate = {2026-11-17},
note = {Retrieved 17 November 2026}
}Entries are updated when the evidence changes; the retrieval date tells your readers which version you used. Changes are listed in the Change log.
Change log
- – Removed the source on a Swiss university laboratory's equipment list and the sentence resting on it; the patent is now introduced by its assignee alone; sources renumbered (Debona et al. from 12 to 11). Reason: Editor decision; the patent sentence had rested on that source to link the patent to a sample-roaster maker.
- – Campo-Ceballos and Gaviria-López 2023: roast times changed from 420–501 s to the drop times logged in Table 1, 410–517 s (text, chart, table, study card); charge temperature now given as 177.9–181.1 °C (Table 1) beside the 170 ± 1 °C of the methods; the mean colour of 65 is now set against Table 1's per-roast values of 75–95 and removed from the colour chart; both cooling times (180 s, 120 s) named. Reason: The paper is internally inconsistent (400–501 s and 420–501 s in the text; 403–517 s planned and 410–517 s logged in Table 1). The logged times are what the roasts took; turning-point.md already uses the Table 1 charge temperatures.
- – The Cenicafé roaster is no longer called a drum roaster (text, legend, charts, table, study card); the electric heat source is no longer shown under the drum in the figure legend. Reason: Gómez Parra et al. 2023 describe it only as an electric sample roaster.
- – The two roasters of Debona et al. 2022 are called small drum roasters, not shop drums, with their capacities; author list completed (13 authors). Reason: The paper does not call them shop roasters; three authors were missing.
- – The SCAA protocol is marked as superseded wherever its 8–12 min rule or targets appear (contested, reference table, deep-dive text, both charts). Reason: Editor requirement; the protocol was replaced by SCA 102–104-2024.
- – Three quotes corrected to the source wording (SCA 102-2024 “nonpermeable”; Anokye-Bempah et al. “(> 1 kg batch size)”; US 9,545,169 chaff-collector sentence ends without a full stop); the patent tierNote and text now say it describes a domestic roaster. Reason: Quotes must match the source word for word; the patent text speaks of a domestic roasting apparatus.
Sources
- A/BSpecialty Coffee Association (2024). SCA Standard 102-2024 Coffee Value Assessment: Sample Preparation and Tasting Mechanics. Specialty Coffee Association.Quoted passage: a recommendation; SCA standard for sample preparation in the Coffee Value Assessment, 2024; section 5.1 Roasting (roast level, roasting problems, consistency, cooling, storage, rest)..
- A/BSpecialty Coffee Association (writing team Fernández-Alduenda, M.; Ionescu, K. E.; Giuliano, P.; Rugolo, J.; Delrue, R.) (2024). A System to Assess Coffee Value: Understanding the Specialty Coffee Association’s Coffee Value Assessment (June 2024). Specialty Coffee Association.Quoted passage: a recommendation; SCA explanatory document for the Coffee Value Assessment, June 2024; sample-roasting section and Table 4 (roast-meter targets)..
- A/BSpecialty Coffee Association of America (2015). SCAA Protocols: Cupping Specialty Coffee. Specialty Coffee Association of America (revision of 16 December 2015).Quoted passage: a recommendation; SCAA cupping protocol, revision of 16 December 2015 (superseded in 2024); sample-preparation section on roasting..
- AInternational Organization for Standardization (ISO/TC 34/SC 15) (2008). ISO 6668:2008 Green coffee — Preparation of samples for use in sensory analysis. ISO.Quoted passage: a description; ISO catalogue abstract of ISO 6668:2008..
- AInternational Trade Centre (ITC) (2011). The Coffee Exporter's Guide, Third Edition. International Trade Centre (UNCTAD/WTO), Geneva.Quoted passage: a recommendation; Trade guide for exporters in producing countries, 2011; liquoring section on the roast..
- ACampo-Ceballos, D. A.; Gaviria-López, C. A. (2023). Optimising consumer acceptability of Cauca specialty coffee through roasting profiles on acidity and body sensory attributes. Czech Journal of Food Sciences 41(1), 64–72. doi:10.17221/181/2022-CJFS · other copiesQuoted passage: data; One Cauca (Colombia) coffee, 13 roasts of 200 g, gas sample drum, central composite design. Text gives 400–501 s and 420–501 s; Table 1 logs drop times of 410–517 s (used here). Cooling 180 s (methods) or 120 s (results)..
- AGómez Parra, C. R.; Gallego Agudelo, C. P.; Echeverri Giraldo, L. F.; Pabón Usaquén, J.; Ortiz, A.; Osorio Pérez, V. (2023). Determinación de compuestos químicos del café tostado por Espectroscopia de Infrarrojo Cercano (NIRS). Revista Cenicafé 74(1), e74104. doi:10.38141/10778/74104 · other copiesQuoted passage: data; 100 Colombian green-coffee samples, each roasted to three degrees on an electric sample roaster at Cenicafé, for near-infrared calibration..
- AAnokye-Bempah, L.; Styczynski, T.; de Andrade Teixeira Fernandes, N.; Gervay-Hague, J.; Ristenpart, W. D.; Donis-González, I. R. (2024). The effect of roast profiles on the dynamics of titratable acidity during coffee roasting. Scientific Reports 14, 8237. doi:10.1038/s41598-024-57256-y · other copiesQuoted passage: a result cited from earlier work; The authors' summary of earlier titratable-acidity studies, in the introduction to a study on a 5 kg gas drum roaster loaded with 4 kg..
- BSchenker, S. (2000). Investigations on the Hot Air Roasting of Coffee Beans (Diss. ETH No. 13620). ETH Zürich (doctoral thesis).Quoted passage: a description; Description of the laboratory fluid-bed roaster used for the thesis experiments, ETH Zürich, 2000..
- A/BStordy, A.; Frieling, R. (assignee IKAWA Ltd) (2017). Apparatus for roasting coffee beans (US Patent 9,545,169 B2). United States Patent and Trademark Office.Quoted passage: a description; US patent granted 17 January 2017 (PCT filed 1 November 2011); construction of a small hot-air roaster..
- ADebona, D. G.; Louvem, R. F.; da Luz, J. M. R.; Nariyoshi, Y. N.; de Castro, E. V. R.; Oliveira, E. C. S.; Guarçoni, R. C.; de Castro, M. G.; de Oliveira, G. F.; Partelli, F. L.; Soares da Silva, M. C.; Bellon, A. P.; Pereira, L. L. (2022). Heat and Mass Transfer Kinetics on the Chemical and Sensory Quality of Arabica Coffee Beans. Agronomy 12(11), 2880. doi:10.3390/agronomy12112880 · other copiesQuoted passage: the authors’ own result; Two small LPG-heated drum roasters (capacity 0.8–1.2 kg and 2.0 kg; batches 1.2 or 2.0 kg), four profiles of one arabica from Espírito Santo, Brazil; five Q graders, 24 h after roasting. Neither called a sample roaster. Tables 1–2 (PDF)..
Tier A · authoritative · B · trade and expert · C · general. Supporting quotes for every source are in the data version of this entry. If a publisher’s site refuses access, try the doi link or “other copies”.
The literature for this entry was last searched on 16 Nov 2026. A statement that no source consulted for this entry shows something refers to the sources above; newer or missing work may exist.
Discussion
A space for readers to share experience from the farm, the roastery or the bar, ask each other and offer other views. The Coffearium team does not reply here. Please keep to the community rules; to change the entry itself, use Suggest a change.