Understanding Coffee Roasting Curves


The coffee roasting curve is an important tool that can be used to understand the coffee roasting process. A typical curve displays the roast time along the x axis (usually in minutes) and the roaster temperature on the y axis (in °C or °F). The curves may typically also display rate of rise on a secondary axis for reasons that will become apparent later. The graphs below may look like a bowl of spaghetti to the uninitiated, but when you understand each line and its relationship to the others the curves will make sense and will tell an experienced operator a lot about what is happening inside the machine. Each curve is a graphical representation of what happened during the roast.  

Temperature Curves

Figure 1 shows an ideal roast curve. In a perfect world most roasts will following along this bending line. The curve is defined by a few key variables:

  • The total roast time usually takes 10-12 minutes, although this can vary depending on the desired final temperature.
  • The endpoint, or the final temperature the roast will end at, determines roast degree. This is on a scale of light to dark.
  • The intersection point dictates the where the turning point for the roaster occurs, which will be discussed later.
Figure 1 - Ideal Roast Curve

Figure 1 - Ideal Roast Curve

The most important thing to know about this curve is that it doesn’t increase linearly, it curves. It almost becomes flat towards the end of the roast, but it never completely does. At no point is it straight either, it always decreases. An ideal roast curve will have a declining Rate of Rise and a reason for this will be discussed later.

Figure           SEQ Figure \* ARABIC      2       - Temperature Sensor Positions

Figure 2 - Temperature Sensor Positions

Figure 3 shows this same ‘ideal’ curve with temperatures from the roaster added. These temperatures are taken using 3, k-type, ungrounded 3mm thermocouples. The orange line at the bottom is read using a 10k thermistor. Figure 2 shows the positions of these sensors on the roaster. In relation to Figure 3, The blue line is the environment temperature, the green line is the bean temperature and the top orange line is the exhaust temperature. These probes are placed in strategic positions inside the roaster in order to return specific information

Figure           SEQ Figure \* ARABIC      3       - Roast Curves with Temperatures

Figure 3 - Roast Curves with Temperatures

The bottom orange line displays data from the exhaust stack. This will be ignored for now. The orange line at the top is the Exhaust Temperature. This one is always a fair bit cooler than the other curves. It is the temperature of the exhaust gases as they leave the roaster. At the start of the roast the temperature drops as the cold beans are introduced. It then steadily increases throughout the roast. The exhaust gas does not increase at the same rate as the other curves. This is because a most of the energy is being taken up by the roast rather than being lost to the air.

The next pair of lines are the environment and bean mass temperature probes. They monitor the temperature inside the drum at two points. The probe associated with the green curve is located inside the mass of beans the blue curve is located outside this bean mass towards the centre of the roaster. These lines will drop together as beans are introduced to the hot roaster. They will then level out and then converge. This is because the cold beans will initially resist taking up energy. They will be colder relative to their surroundings. Throughout the roast the beans will become more conductive and increase in temperature. They will therefore approach the temperature of their surroundings. The bean mass curve will level out at the end of the roast and approximate the environment temperature. The mass of the beans will be close to the air temperature inside the roaster. They may even exceed it in darker roasts as the beans undergo an exothermic reaction post first crack.

The last line is the stack temperature. It is largely constant as in the large diameter chimney the air is moving more slowly. This means the hot air stream increases in volume and losses most of its heat to the environment around it. Most energy is being taken up by the beans if the process is efficient so the temperature of the stack should not be expected to increase dramatically.

Rate of Rise

Figure 4 - Rate of Rise

Figure 4 - Rate of Rise

Figure 4 shows the Rate of Rise (ROR) of a roast. It is the derivative, or rate of change in the bean temperature. Different roasters measure this number differently, which is important to consider when comparing curves. It may be in degrees per minute, per 30 seconds or any number in between. Some roasters will also take averages over the period in an attempt to ‘smooth’ their curves. This can be detrimental to the roast because in order to do this, the roaster will have to compare new values with historical readings. The ROR could be bouncing up and down at the roaster will not report this, because it can only look at an average. This may mean in order to correct a crashing or rising ROR, the controller will act too late to save the batch. It is important that this ROR decrease steadily throughout the roast. There are many theories as to why, but at the time of writing, no consensus. There are possible explanations however. One is a declining rate of rise means that the difference in temperature between the outside of the bean and the inside will be less and less as the roast progresses. The beans throughout the roast will change their ability to transfer heat. The beans use a lot of the energy to increase in temperature initially and are not very thermally conductive. The roaster will be set on maximum heating power in order to raise the temperature of the water inside the bean. It is also important to start the roaster at a hot temperature so that some of this excess energy can help increase the temperature from room temperature to over 100°C as quickly as possible. This causes the curve to ‘turn’ usually at the 1:30-2:00 minute mark. As this water vapourises the bean changes size, the internal cells open up and energy is more easily transmitted into the coffee. If this gradient was too high it will cook the outside much faster than the inside, resulting in both over roasting and underoasting. If the rate of rise was zero, or flat the coffee will be held at relatively constant temperature and will lose it moisture to the environment. So although the roast will otherwise appear fine, the bean will taste over roasted, or baked. This is not to say it will necessarily taste ashy or smokey, but it will have a hollow taste.  But if the rate of rise is decreasing it can have the advantage of quickly raising the coffees temperature at the beginning, saving water content then evenly raising the temperature of the coffee later in the roast. I will stress again that there is no consensus on this, this is just opinion. A declining rate of rise seems to give a much larger window to hit between optimum mass loss, and end point of the roast. Drying is a function of airspeed and air temperature. A coffee that spends a large amount of time at high temperature, without that temperature changing, will dry more than a coffee with a declining ROR. Hence the reactions associated with coffee roasting will progress for longer that leads to flavours associated with longer roasting, despite the bean colour being quite light. In extreme cases this can result in a hollow flavour, or a light roast lacking its origin characteristics that it should have.

Heating and Airflow

Figure 5 - Heating and Airflow Settings

Figure 5 - Heating and Airflow Settings

There are two variables the can be manipulated by the roaster after the batch size and charge temperature have been decided. These variables are airflow and heating and are illustrated in Figure 5. This particular roaster uses electric heaters to supply heat to the roaster. Most roasters will not do this. An electric method of heating requires a lot more planning as it takes a long time to change the heat input of the element. It is also much less efficient than gas burners, especially when larger roasters are considered. However the principles are the same. The white line shows the relative output of the element between full power and no power. At the beginning of the roast the computer allows the machine to heat completely and the element is held at 100%. This is because during the beginning of the roast the change in temperature is an artefact of actual temperature readings and much less critical. To elaborate, the cold beans are charged into the roaster at room temperature. The roaster begins the roast a very hot temperature comes down in temperature as thermal energy is absorbed by cold beans. What we see on the graph is a hot probe being cooled by beans. Eventually the probe reaches an equilibrium point where the probe and is the same temperature at the beans around it. This is referred to as the turning point. They then climb away together at the same temperature or similar depending on the thickness of the probe. From the beans perspective they start at room temperature and climb to this equilibrium point and keep going. So a controller is told to ignore this period of the roast. If it was not told to ignore it, the controller would think it was doing something wrong, because despite having the element on full the roaster was decreasing in temperature. With industrial controllers this can cause problems which can cause them act erratically. So sometimes it is better to drop the coffee in manual mode so that the controller doesn’t suffer from issues.  If the controller is programmed correctly it will be looking at the bean temperature throughout the roast and making sure it is rising at the correct rate. For example, if the controller senses that temperature is climbing too quickly, it will begin to throttle back the heating element. This is important because the coffee transitions between an endothermic (heat absorbing) process and an exothermic process (heat releasing). The roaster will not need the amount of heat available to it throughout the whole roast. A good analogy for this is a car on cruise control going over a hill. To go up the hill the controller applies more accelerator to get car up the hill. Once the car is at the top of the hill the computer releases the accelerator to maintain the set speed. This is because the car doesn’t have to climb anymore, so that extra power is no longer required. When the car starts moving down the hill, the accelerator is completely released. The controller is relying on the friction between the road and the car to slow the speed of the vehicle. Most experienced drivers will know they might need to help the car decelerate at this point by applying the brake. The brake on a coffee roaster is the fan. It is used to pull cold air from outside the roaster, past the element, through the roaster and out the stack. This will result in changes in air temperature as hot air is pushed away in favour of cold. The fan also serves the purpose of pulling smoke and chaff outside of the unit. In Figure 5, the red line is representative of the airspeed inside the main stack. For the first part of the roast it is relatively constant and low because it is being told to ignore the first drop (as discussed previously). When the roast is too hot or climbing too quickly, it turns itself up in an attempt to throttle the roast back. When the roast is too hot it throttles itself up, and when it is too cold it turns itself down. Towards the end of the roast the fan should be at maximum to ensure that smoke is being rejected from the roaster and the rate of rise is not increasing. Again this can be done manually but computer control is preferred. One thing changing the airflow in the roaster will do is change the relative humidity in the stack.

Humidity and Mass Loss

Figure 6 - Humidity and Mass Loss

Figure 6 - Humidity and Mass Loss

Figure 6 shows curves related to Relative Humidity (RH), and the total mass of water in the coffee. The temperature at which water will condense or become a liquid changes based on its temperature. So air will hold a certain amount of water within it, if that air is rapidly cooled, the water will condense. Hot, wet air or air that contains a lot of water will condense at relatively high temperature, think steam condensing out of a kettle. Even on a warm day you will see some “steam” coming out of it as it boils. This isn’t steam however, its liquid water condensing in the air. Simply just blowing on a cold bit of metal will cause some of the water in your breath to fog up the metal. Now, if that same metal is hot, you will not see this same effect because the metal will be too hot for water to condense on its surface, even very warm, humid air may not be able to condense on a hot surface. So if we know the temperature of a gas and its relative humidity we can use a series of equations to work out the total mass of water in the air. If we know how much water is in the ambient air per cubic metre, and how much air is going through the roaster we can cancel out volume based terms. This means at any given moment the result of this calculation will tell you how many grams of water are present in the stack at that specific point. If you start adding those numbers up, you will be able to tell how many grams of water has left the coffee. This is what blue line in the centre represents. It’s the amount of water that is leaving the coffee. At first, before the water hits its boiling point, not much leaves the system. But at the roast progresses and this water starts to boil this line starts declining at a steady rate. The really important thing to know is this rate is dependent on how much water there is in the coffee and how quickly the energy can be transferred to it so it can be boiled. It largely will not matter how hot the roast gases are (within reason) if the energy cannot be transferred into the coffee. So the rate will be constant as the roast progresses. However at first crack this changes. As the bean puffs up and most of the water has been converted to steam this number slightly, but noticeably changes. It picks up speed and water is much more rapidly transferred out of the coffee. The thermal properties of the coffee have changed. This heat sink of water is largely gone the heat can penetrate the bean much more rapidly. This is important when determining final mass of the coffee because once the coffee is past first crack the drying progresses faster. In many areas of food processing there are optimum levels of water content that contribute to flavour. Although we are drying the coffee by roasting it, a coffee roaster is not trying to desiccate the beans. The roaster will want the beans to undergo reactions associated with roasting and will want the roast to be even through coffee. But also a roaster needs to hit an ideal moisture content at the end.  The roaster has largely determined how far the reaction will progress by setting the end temperature.  The roaster must also maintain the flavour of the coffee by ensuring the coffee does not become over roasted. Over roasting can be thought of disproportionately large mass loss despite a low roast end point.

Putting it all together

Figure 7 - Final Graph

Figure 7 - Final Graph

Figure 7 shows every line that has been discussed on one chart. All curves are related to one another. One cannot hope to see the whole exact picture of how heat is being transferred in the coffee and how the reactions are being developed. That is an impossible task because coffee is simply far too complicated. However clues as to what is happening inside the coffee are left by things such as small changes in temperature over time. For a long time, coffee roasters were reluctant to even add probes to their roasters and monitor the roasting process. Temperature probes are a simple low cost way to get some of the picture. The relative humidity of the exhaust gases can also be monitored to further fill out what is happening inside the drum. By adhering to a few simple rules, like charging correctly and controlling for a declining rate of rise and making sure the coffee does not over roast a coffee roaster can dramatically increase their ability to get the most out of their coffee. This is particularly important when roasting specialty coffee where a roaster will try and preserve the origin characteristics of the coffee. What is great and interesting about this method is that it quantifies variables that were otherwise completely qualitative. This is because a computer cannot smell a roast to make sure it is correct. A computer cannot take part in a cupping. But what a computer can do is use this information to make informed decisions about how a roast is progressing, knowing when to add heat and knowing when to remove it. By analysing water leaving the roaster it can even know when a coffee has finished roasting.

It might be foolish to think that a computer program could replace people in coffee roasting entirely. There may still need to be a coffee roaster who can qualitatively assess the coffee involved in the process. However, it is hoped that program or a program like this could be used to virtually eliminate roast defects associated with machine error. It could also be used to improve consistency in specialty coffee roasting. It could be used to help a specialty roaster best showcase all the farmer’s hard work with the coffee.