Reading a roast profile

The profile is the path, not the destination

Sourced

Two things describe a roast: where it ended and how it got there. A doctoral thesis on roasting, written in 2000, says the time and temperature profile had been the most discussed issue in coffee roasting for the previous decade. It calls the temperature profile the most crucial parameter in designing the process, because it sets both how flavour forms and what structure the bean ends up with, and its own roasts taken to the same degree by different temperatures differed in volume and porosity.

A profile is therefore a way of holding the path constant so that a result can be repeated, on the same machine, with the same coffee. That is its real use. A roaster comparing today's batch with last week's is comparing like with like.

What can be changed, and what the sources say it does

Sourced

Each note reports one source about one set-up. None of these settings has a correct value; they interact, and a change to one shifts what the others do.

Rate of rise: what the number is

Sourced

Rate of rise is how fast the measured temperature is climbing. In the drum-roaster study it was logged by roasting software as the increase in degrees per thirty seconds. It is calculated from the temperature reading, so it is a second line derived from the first, not a separate measurement.

It is read from a point called the turning point, which the study defines as the moment the beans and the roaster come to the same temperature. The study takes its initial rate of rise as the highest rate immediately after that point.

A rate shows a change of pace more plainly than the temperature line it is derived from, which is the reason for plotting it: a roast that is speeding up or slowing down is easier to see.

What the probe is, and is not, measuring

Sourced

The temperature on a roaster's display is often called bean temperature. It is the temperature of a probe sitting in the moving mass of beans. The thesis measured both that pile temperature and, with a fine thermocouple, the temperature at the core of a bean, and found the pile reading higher than the core in every process it ran. Neither ever reached the temperature of the incoming air.

So the curve is a consistent signal from one probe in one place, and a rate of rise calculated from it inherits whatever that probe does. It follows, as a matter of reasoning and not of a study CoffeeHQ has read, that the same number on two machines need not mean the same thing, and that software settings which smooth the line change how the curve looks without changing the roast.

A model of what stands between a bean and the line on the screen

CoffeeHQ explanation

A model of the chain of delays. Each link is ordinary measurement physics; none of the delays has been measured in a coffee roaster in anything CoffeeHQ has read, so no figure is attached to any of them.

Flick, crash and the baked curve: belief and evidence

Accounts differ

Roasters have names for curve shapes they try to avoid, and a research paper that set out to test them records three. A flick is a sudden rise in the rate of rise some time after first crack. A negative rate of rise, the shape often called a crash, is what happens when heat is lost around first crack and the temperature stops climbing. A profile that hurries to the colour change and then lingers before first crack is one the paper describes as baked.

A university group set out to roast each of these on purpose, on a five-kilogram commercial drum roaster, with seven profiles of the same length. Their paper is direct about the state of the evidence. It says the flick is considered by some roasters to be a common defect with a significant effect on flavour, and of the negative rate of rise it says that some roasters believe it seriously degrades quality, although to the authors' knowledge no published data supports this contention.

What they measured was acidity and colour, not taste. Acidity developed differently from profile to profile, but its peak came at first crack every time and was of similar size. Bean colour followed the same path in every profile and was approximately the same at each milestone. Those results do not show that curve shape is irrelevant: nobody tasted the coffee in these two papers, and the authors say that more detailed chemistry is still needed. They do show that two of the things a roaster might expect a bad curve to change did not depend on it in the way the teaching implies.

Development time: what one group's tasting studies found

Sourced

One group has tested roasting time with trained tasters. Across eight studies, seven on a one-kilogram drum roaster and one on a fifteen-kilogram hot-air machine, with airflow held constant, they varied three things: the colour the roast was taken to, the time it took to reach first crack, and the time from first crack to the end, which roasters call development time. Six of the studies used a trained panel of seven to eleven assessors and two used coffee professionals after a short training.

Colour was the stronger predictor of flavour. Darker roasts and longer roasts moved the same way: more bitterness, less acidity, less fruitiness and less sweetness. But the authors conclude that time mattered at a fixed colour as well, and of the two periods it was the time after first crack, not the time before it, that had the larger effect. Two roasts of the same colour are not the same coffee if one spent longer getting there after first crack.

The results tables show how strong that is and is not. Two of the studies varied colour and timing together. Colour had the larger correlations in both. In the first, where the roasts were made far apart on purpose, development time was related to every attribute reported; in the second, where the differences were subtler, its links were mostly not significant, and the panel found samples that differed only in development time much harder to tell apart. Pooled over seven studies and some 1,900 ratings, time after first crack was related to every attribute scored, and time before it to five of nine. That pooled analysis has no term for colour and includes the second of those studies, in which colour varied, so it is not strictly a comparison at one colour. The authors' own illustration of size is that three minutes of development corresponded to about two centimetres less acidity on the line the assessors marked, while the slopes for sweetness and body were so close to nothing that they call them more likely to be spurious. In the one study that reported it, total roast time, taken alone, was related to no attribute. Longer development also went with a stronger note of roasted bread, which the authors connect with the fault roasters call baked.

One of those studies was published separately, and it is the nearest thing read to a test of how the end of a roast is shaped. One Colombian coffee was roasted four times on the same roaster to the same colour, reaching first crack at about nine and a half minutes each time and then taking 90, 143, 266 or 390 seconds to finish, at end temperatures falling from 204 to 191 degrees Celsius. The slowest, in which the temperature rose very little after first crack, the authors call baked. Forty-six coffee professionals, tasting blind at a trade event, rated the fast roast highest for acidity, fruit, sweetness and clean cup, and the two slow ones highest for bitterness, astringency and roasted and nutty notes. Body did not differ. The study measured description and not liking: its authors say that whether the change is for better or worse is a question for consumer research.

The paper is frank about why it was needed. Views on how a curve should be shaped, it says, are in most cases based on the individual opinions of professional roasters and not on carefully planned experiments with trained panels. It is two roasters and nearly all of it one, washed coffees only, one way of running the machine, and a panel that belongs to the first author's training company; and although it varied how fast a roast finished, it did not look for a crash or a flick. The papers describe heat that was only ever turned down as a roast went on, and report no roast in which the bean temperature fell and recovered, so they do not test the claim that a rate of rise dropping below zero, taken alone, harms a roast. What it supports is that development time is a real variable. It gives no value of it, and no share of the roast, to aim for.

Six numbers used for development, and what kind of thing each is

CoffeeHQ explanation

Development is used for the roast's progress after first crack, and half a dozen numbers are offered as measures of it. They are different kinds of thing, and treating a convention as if it were a measurement is the commonest mistake.

Development time

A process variable: a duration the operator controls, counted from an event judged by ear. The one set of tasting studies read found it affects flavour at a fixed colour.

Development time ratio

Arithmetic on two durations: the time after first crack as a share of the whole roast. A convention. No study read tests any target for it, and the same ratio describes a fast roast and a slow one that the tasting studies would not treat as equivalent.

End temperature

A reading from one probe in one machine. A proxy, repeatable on that machine and not comparable with another's.

Colour

A measured property of the product, read by an instrument that reports reflected light on a scale where smaller numbers are darker. In the tasting studies it was the strongest single predictor. It is usually read on ground coffee, because the outside of a bean is darker than its inside, and by how much depends on how fast it was roasted.

First crack

An event, and a landmark. In one group's roasts the beans were about the same colour at first crack however the roast had been run, which makes it a better anchor than a temperature reading.

Weight loss

A measured property of the batch, most of it water. A check that a roast was repeated, for one green coffee.

Weight loss: the one number anyone can measure

Sourced

Weigh the coffee before roasting and again after cooling, and the difference as a share of the starting weight is the roast loss. The thesis defines it exactly this way and reports the usual range in the literature as 14 to 20 per cent. As arithmetic: 250 grams of green coffee that comes out at 212 grams has lost 38 grams, which is 15.2 per cent.

The thesis says the loss may serve as an indicator of the degree of roast for a given raw material, and the qualification matters. Most of the loss is water, so a green coffee that started wetter loses more for the same roast. Comparing the figure between batches of one coffee tells you whether you repeated yourself. Comparing it between different coffees tells you much less.

Sample roasts, production roasts and why a curve does not travel

CoffeeHQ explanation

A sample roast is a small batch roasted to evaluate a green coffee, usually before buying it. A production roast is the batch that will be sold, and what it has to be above all is repeatable. The drum-roaster study includes a profile it calls production, run at one energy setting from start to finish with no adjustments, and describes it as easily reproduced and therefore typical of large operations that repeat the same roast many times.

Moving a roast from a small machine to a large one is not a matter of copying the curve. The thesis found that its small laboratory roaster behaved in its own way, and puts the difference down to the particular proportions of conduction, convection and radiation that come with a small batch. It used roasts on industrial equipment as the standard its laboratory work was checked against, not the other way round. The reasonable inference is that a profile is rebuilt on each machine by tasting the result, using the old curve as a starting sketch.

CoffeeHQ could not open the published protocols that specify how a sample roast for cupping should be done, so this page does not state them.

A comparison to run if you roast

CoffeeHQ explanation

• Same coffee, same end point, two paces

  Change

  Roast two batches of one green coffee to the same roast loss, one reaching first crack noticeably sooner than the other.

  Keep the same

  The coffee, the batch weight, the roaster, how the batch is cooled and how long both are rested.

  What to notice

  Taste them side by side without knowing which is which. Whether you can tell them apart, and which you prefer, is the question the published studies have not yet answered.

What this page does not give you

CoffeeHQ explanation

It gives no target curve, no rate of rise to aim for, no development time and no percentage of the roast that should follow first crack. Those figures are widely taught, and CoffeeHQ found no published test that supports any of them as a rule. The one group that tasted roasts of different development times found that the time matters, and offers no figure for it.

It does not say that curve shape has no effect on flavour either. The honest position is that it is unproven in both directions, and that a roaster's own blind tasting on their own machine is better evidence for them than anybody's rule.

What to do next

• Logging a roast — What to write down so that two batches can be compared at all.
• Coffee roast defects — What the same research group found when it made baked, scorched and underdeveloped roasts deliberately.
• Coffee roasters and heat transfer — Why the same settings behave differently in a drum and in an air roaster.
• What happens when coffee roasts — The acidity and colour findings in the context of the whole roast.