Bean probe sensitivity to airflow

The bean probe does not only measure the temperature of the beans. It is affected by the metal carcass where it is fixed – and by the air passing by.

Also listen to podcast episode #13 🎧

The bean probe is quicker affected by the airflow than the beans. Metal absorb heat quicker than the cellulose beans. Think of sitting on a bench in the park on a cold day; the difference of a wood bench and a metal bench. Coffee beans are more like wood than metal.

Here I have tested the bean probes sensitivity to airflow.

This test is on my Bullet R1 roaster (a test roast with old beans, not meant to taste good). The bean probe is a NTC precision type.

Look at the graph after 8 minutes:

Peru test ny probe sept2017

In the last 4 minutes of the roast I jumped between a high airflow; F7 and a low; F2 (and once F1). Each time approximatly 30 seconds stretch. I chose F7 because already at F5 the airflow starts to cool rather than increase heat transfer. That is with 400 grams batches – at larger batches it takes higher fansetting.

High airflow made the ROR go negative within 10 seconds.
Low airflow made it go positive within 10 seconds.

During this shifting airflow the Bean temperature (BT) did anyway rise from 187 to 191°C, measured at low airflow. Probably due to heat transfer from the drum, where the heat setting was constant at P3 during these 4 minutes.

Data excelark

Clearly the bean probe is not only affected by the actual bean temperature – but also of the airflow. As we long expected. And as Rob Hoos also wrote in his Roast Magazine article this month (the sep/oct 2017 issue). He also talks about it in this video

Read more about the Bean Temperature under Roasting Basics.

Note: The first 8 minutes of the roast was because I had just changed the bean probe. At first the test was only to see the difference to the old one. But during the roast I got the idea to test the airflow impact on the BT reading.

Different batch sizes

Another indication that airflow is affecting the BT probe … is that the reading at First Crack start depends on batch size.

In general it is established that First Crack (FC) starts around 200ºC. But it is very common that the bean probe on a roaster gives a lower reading. For all types of roasters.

On the Bullet R1 I have seen a range from around 170ºC up to 196ºC for FC start. This is affected by two things: if it’s an old or new probe … and on batch size.

The lower temperature happens when only roasting 400 grams. The Bullet R1 has a capacity of 1 kilo. Below you can see 2 batches Ryan Garnes did with the same bean; 700 grams and 1,000 grams.

700 grams:

700g_Ryan Garnes

1000 gram:

1000g_Ryan Garnes

First crack comes higher at 1 kilo:
At 700 gram First Crack sets in at 191ºC.
At 1,000 grams it sets in at 196ºC.

In both cases airflow was around F4. The air moves from the back of the drum – through the beans – and affects the bean probe placed on the front (read about Bean probe sensitivity to airflow). When the air travels through very hot beans it gets heated up before hitting the bean probe. More hot beans will heat up the air more. Therefor higher BT reading at FC.

At yellowing point the 700 grams batch lies higher:
At 700 gram at 149ºC.
At 1,000 grams at 141ºC.

Here the beans are not as hot than at First Crack. So the beans rather cool down the air than heat it.

Taking the trier out

If you got a trier on your roaster … taking it out also effects the airflow. See


Also listen to podcast episode #13 in my podcast Coffee Roasting Navigated🎧

High heat start

Our roasting community here in Denmark have been exploring a new approach on the Bullet R1 roaster.

15 of us were on a roasting course with Michael de Renouard from The Factory Roast Lab Copenhagen. He is a roasting consultant and travels the world teaching coffee roasting. At home in his own roastery, he is roasting on a Loring. He roast quite light.

Refshaleoen kursus maj2017

Michael got one of our Bullet to try it out. He thought it did well; no burned or smoked taste in the beans. The roaster do well in getting the beans heated quickly. He likes the seperate control of heat and airflow.

He found that you could start out with high heat without the beans got scorched/burned. Preheat 200°C for a 700 grams batch and then P8 – leaving room to step up to P9 at then end of the ROR peak.

Note: This was for high density beans (high grown). Be aware that lower density beans, like brasil, need a lower preheat temperature. Steffen got a Costa Rica bean which gets burned if the preheat is more than 170°C.

The high heat makes a fast start on the roast. But if it doesn’t scorch the beans, then no problem. Michael also focus on the lenght of the phase from yellow until First Crack start. Here he aims for 3 minutes for this phase – and at the same time aim to bring ROR below 5 for the First Crack phase. To do this the heat must be turned down well before the start of FC.


Exploring the difference

Here Thomas Villars used the strategy on a etiopean bean. He thought it tasted better than earlier roasts of the same bean. Despite FC starts as early as 5:23 min.

New strategy:

Villars Homa rist 13majIn general a fast roast like this gives more “clarity” in the taste – and less body. Here the “middle phase” – from yellow to FC start – is 2:15 minutes long. It should give more body to prolong this phase.

Earlier Thomas roasted the same bean like this:

Villars Homa 30apr

Update 2 months later:
Thomas has continued this new strategy. He like his coffees better now.

Comment from Therese
To me this approach gives a too thin and too acidic taste. I prefer more body and sweetness.


Also see Kenyans for a light roasts


Note: There are many approaches in coffee roasting. Don’t take this website as the only way to do it. Just take it as inspiration – and explore for your self what suits your taste.

Repeating a roast profile

If you made a good roast, you likely want to repeat it :-) But thats not easy to do accurately because so much influence a roast.

Steffen from KaffeRist do it often. Here he made 7 repeated roast of the same bean – trying to make the same roast profile. All done in a row.

You can see how the curves doesn’t overlay completely. And First Crack (FC) starts at different times and temperatures.

Steffens 7 gentagne rist april2017

He roast on a Bullet R1 roaster. The software got a PlayBack function that repeats the settings of preheat, power and fan during from the roast you want to repeat.

But the same settings doesn’t give the same profile even though it’s the same bean and the same batch size. So during the roast Steffen overrules the automatic settings. The ROR curve is the best way to see where the Bean temperature (BT) is heading – to adjust in time.

And then he listens for the First Crack (FC) start every time. And aim to make the development from FC start the same every time: regarding time and temperature rise.

Good and bad tasting of the same bean

If you really want to learn how different roasting profiles affect the taste – then roast 2 batches on the same day with only small differences in the profile. And then taste them side by side in the weeks afterwards.


Here is an example of a good tasting and bad tasting coffee of the same bean. Much in the roast profile are alike, but not all the way. What causes the difference in taste ?

The bean is a washed arabica from Uganda. Fairly high grown at the Mount Elgon mountain on the border to Kenya.

The good tasting, 400 grams, by Therese

mt-elgon-great-2016-11-21-05-29-00Taste: no burn flavour, no bitternes, nice round and big aroma.

I roasted it again 9 days later. First the curves were very close. But after 5,5 minutes the ROR levels differed. Here both curves:

Two roast of the same bean Mt Elgon nov2016

The higher ROR level here gave an earlier FC; at 7 minutes in the second roast where as the first had 8:45 min.


This coffee didn’t taste good to me. It had bitterness and burned taste. And a weak aroma = boring coffee.


What could be causing this difference in taste ?

First the alike elements
The development-% was around 24% for both (the time after FC related to total roast time). And weightloss was about the same: 14,3 and 14,5%

Development time (DT) was 2:44 minutes for the good one. The bad with more burned taste you would expect to be roasted longer, but no, its roasted a bit shorter: the DT was only 2:19 minutes.


Possible explanations for the burnt taste
I cant say for sure why they tasted like they did – because several things was different. But here are some thoughts:

(1) A higher temperature rise in the bad one: 12°Celcius (and BT ended on 181°C) … where as the good one only rose 6-8°C (to end-BT at 175°C).

(2) When entering First Crack the bad one had a ROR around 10°C pr min. Whereas the good one was around 5. I have heard a recommandation around 5°C per minute – and not as high as 10°C when entering FC and for the rest of the roast.

(3) The american roaster Rob Hoos talks about the importance of the middle phase: from yellow point to FC start (see his book “Modulating the flavor profile of coffee“). He calls it the Maillard phase. In these two roasts the drying phase up untill yellow point are not that far apart: 4:00 and 3:45. But the lenght of this middle phase is 3 minutes for the bad one, and 5 minutes for the good one.

A roast consultant told me he prefers 3 minutes, so thats not criminal in itself. But maybe it suits this particular bean better with a slower roast; a longer middle phase and lower ROR levels during FC.

(4) They flavours go up and down during the roast – so maybe the bad one here would have been better if dropped earlier or later at than 2:19 minutes from FC start.


Martin Kjeldsen roast of the same bean

Martin Kjeldsen lives in different part of Denmark, but we got the same bean – and the same roaster, the Bullet. The bean is from Uganda from Mount Elgon (on the border to Kenya).

Martin have also roasted the bean several times. This is his best tasting batch:

500 grams, preheat 185°C


Martins bacth was roasted 2 minutes from FC start to end-BT at 192°C. My good one was roasted 2:45 min and to BT 175°C. But our bean probes do not measure alike. Notice the difference in FC starts: Martin at 182°C. Mine at 167°C and 169°C.


Read more under Roast profiles.

Adjusting profile by control points

beans-trier-sh1It is not that easy to repeat a great roast you did some weeks ago, even though you use the exact same settings on the roaster. That is because the roast is affected by other things: how much the roaster is preheated, the batch size, and the surrounding air (see Changing weather).

The next thing is that the roasting process is slow to respond to changes so you have to make changes in advance. It’s like steering a ship.

So, it is a good idea to have control points along the way to see where the roast is headed and to do adjustments at an early stage … to better navigate the roast.

kompas kort kop

Here are the control points that I (Therese) use :

Turning point (see definition in Roasting Basics): use the temperature or time to see how fast the roast got off the ground. I note down the temperature. If it is lower than normal, the preaheating was too low. But then apply more heat.

115ºC: I note the time when the BT reach 115ºC. Again here I can see how fast the roast is going. If it is too fast, decrease the heat (P-setting).

From here on I prefer to use the physical changes: color change and the sound of First crack.  I make notes of the time, BT and ROR.

You can also keep using the BT reading. Just be aware that it is also affected by the airflow (fan setting). I initially used the BT reading as a navigation aid, but I found it more reliable to go by the color changes:

udsnit-regneark-analyseNote: this was with an older BT probe that had a lower temperature reading.

Looking at the ROR level tells you how fast the roast is going. If it is higher than I want at this stage, I have to turn the heat down. From around First Crack, you can also increase the airflow (fan setting) to the point where it cools.

I typically aim for a First Crack start between 7 and 9 minutes. But for some espresso roasts I like it to be later – but no later than 11 minutes. Read more under Taste preferences.

You got to find your own control points and choose your aims.

Also see
How Steffen repeated 7 roasts using the PlayBack function in the Bullet software but still needed to adjust along the way.



ProfileTool for the Bullet roast data

Another of the Danish Bullet owner, Steen Hjelmstrand, has made this additional software in 2016.

The purpose was to be able to do more analysis on the data from the Bullet roaster than the RoastTime software from Aillio. But still you have to log the roast with RoastTime and save it in a json-file.


Steen have name it ProfileTool and you can download it for free from his website

Here are two profiles showed in ProfileTool:

1000grams protocol John Plato

Airflow and chaff filter

Its not only the fan setting that determins the airflow in your roast batch. Also the exit passage of the air makes a difference.

→ If it flows easily the airflow is bigger.

→ If the exit passage is somewhat restricted; the same fan setting will give less airflow.

If the filter in the chaff collector is dirty (full of chaff and coffee oils) the passage is restricted. There for regularly cleaning is a good idea.

Back in 2016 the first filter in Bullet roaster looked like this. It quickly got full. Now the filter is much larger.








Here is the cleaning routine for some of the Bullet owners with that filter type:


Read more about fan setting under Roasting basics in the section Fan/airflow

Roasting small batches

In general, you have to adjust the heat and airflow to batch size. Use less heat and fan for smaller batches.

But how little can you roast in a given roaster? It depends on how much you can turn down the heat. And then you have to be aware of what it take to get a fair reading with the Bean temperature probe. To tell what is going on in the beans the probe has to be covered in beans.

Aillio says minimum batch is 350 grams to have the right BT measurement on the Bullet R1.

With smaller batches the BT probe will not be sufficiently covered. But you can still roast smaller batches. Just ignore the BT reading.

Klaus investigated this with 100 grams batches to do sample roasts. Read about his settings here.


Diego Cano in Colombia uses the Bullet roaster to do a lot of sample roasting. He found that it works the best with 250 grams samples.

Learning to roast on the Bullet R1

bullet-sort-hvidThe Bullet R1 is a 1 kilo coffee roaster from Aillio. This website has no commercial relations to the company.

I have just collected experiences with roasting on the Bullet because it was new to the market (Therese talking). I got it in preorder together with a bunch of other guys here in Denmark in 2016. There was no one to ask, so we had to explore together.

This roaster has more setting possibilities and more data output than we were used to – and more than other roasters in this price range. So there was a lot to explore. As people did experiments, I collected them on this website. And then it grew to be about roasting in general.


Learning to roast on the Bullet R1

On this site I have gathered the knowledge and experiences at:

1. The basics of roasting and control functions

2. Roast profiles on the Bullet

Just bear in mind that you most likely cannot precisely replicate a profile settings with the same result. The outcome depends on:
• your surroundings (temperature and humidity)
• the beans (type of beans and temperature)
• your power supply

So you have to try out for yourself. Roasting is not easy.



The Do’s and Don’ts of Roasting on the Bullet R1 from Aillio

Official manual for the Bullet R1 roaster from Aillio


Talk with other Bullet-roasters

You can also find Facebook groups with Bullet owners, like Aillio Bullet R1 User Group

Outside Facebook you can join the Roast world forum at Aillio