Level: Apprentice

Read time: 10 minutes

Dumb question alert though. I read online that certain small batch companies seem to apply one temperature and then simply play around with the length of time. When they do that, the ramps will follow your rule that each of the Drying, Development and Finishing phases will each have a higher ramp than the next since ramps I believe slow down as the temperature in the beans start to approach the temperature of its environment. My question is whether you have experienced more intense flavors by starting with a high temperature and then coming down.  As well, why did you choose 232F as the end of the development phase.


This isn’t a dumb question at all.  In fact it is a great observation and turns out to be a bit of deep question.  To be clear, those makers that use one temperature and vary the time are using ovens, be it traditional or convection.  To my knowledge none are using drum roasters.  That is important because in many ways they are very different ways to roast and not convertible from one to another.  It is very easy to talk about the Drying (time to 212 F), Development (212-232 F) and Finishing (232 F to EOR)  phase in a drum roaster but virtually impossible to do that with oven roasting even though people keep trying and mostly failing. 

The reason they fail (including myself) is due to the nature of heat flow and to my mind, what makes for a good roast that in turn creates a dynamically flavored chocolate.  In a drum roaster the majority of the mode of heat transfer is by convection.  This is a fancy way of saying that because the beans are being lofted, each bean is surrounded by a pretty consistent blanket of hot air all the time.  As heat soaks into the beans and the air cools, new warm air is there to take its place. Heat penetration into the bean is very consistent, fast and predictable. This is why convection heat transfer is so great.  What that in turns means is you can have, comparatively speaking, fast roast profiles.  These fast profiles in turn create dynamic flavors.  

Contrary to the name, convection ovens don’t primarily roast by convection.  They are called convection ovens because there is active air flow.  Unfortunately half (or more if you have more than one layer deep) of the beans on a tray are not being blanketed by hot air.  They are setting on a metal tray and the only mode of heat transfer is conduction which is much slower.  Lots of people have recognized that and use perforated sheets thinking the air will magically flow through the holes but for a variety of reasons air doesn’t flow through these hole.  Just try blowing through a perforated plate from 12 inches away and see what you feel.  It is going to be next to nothing.

Why this is important is that a roast profile that will create a dynamically flavored chocolate generally has a Development phase 2-4 minutes long and even 4 minutes is pretty long.   That kind of length on the other hand is hard to achieve in an oven without damaging the beans.  The times are usually closer to 6-10 minutes if not longer.  Why don’t you just turn up the heat in the oven I hear you asking.  Well, this is what I recommend, but it has drawbacks.  When the beans are full of water you can have the oven 350-400 F, but if you don’t turn that heat down in pretty short order then the difference in heating modes will quickly cause problems.  Basically the top of the beans will start to over roast (due to the efficiency of convection) and the bottom of the beans will be under roasted (due to the inefficiency of conduction) and so you end up with an uneven, non-reproducible roast.  And even if you do turn the temperature down and stir a lot to try to minimize the effect of those different heat transfer methods, you are still going to end up with a roast that is hard to reproduce.

By trial and error, people have intuitively worked this out.  To minimize the different ways heat flows over the beans and through the tray, the best solution is to keep the oven temperature pretty low so basically the heat can penetrate roughly at the same speed, from top and bottom.  The results are very long roasts where the Development time is so far outside the windows I use for drum roasting as to be basically not applicable, stretching sometimes over 10 minutes.

With such a long roast you now introduce another issue.  Uncertainty.  There is something called the butterfly effect. https://en.wikipedia.org/wiki/Butterfly_effect  To just quote Wikipedia,

In chaos theory, the butterfly effect is the sensitive dependence on initial conditions in which a small change in one state of a deterministic nonlinear system can result in large differences in a later state.

The term, closely associated with the work of Edward Lorenz, is derived from the metaphorical example of the details of a tornado (the exact time of formation, the exact path taken) being influenced by minor perturbations such as the flapping of the wings of a distant butterfly several weeks earlier. Lorenz discovered the effect when he observed that runs of his weather model with initial condition data that was rounded in a seemingly inconsequential manner would fail to reproduce the results of runs with the unrounded initial condition data. A very small change in initial conditions had created a significantly different outcome.

I think you can see that the longer a roast is, the more chaotic it is going to be.  The best way to deal with that chaos is to try to minimize variables and really the only choice here is to pick one temperature and vary the length of the roast.  And this is what people do. 

Just to take off on a minor tangent, not all beans like the same temperature so if you are starting with a bean, you have two variable and chaos ensues as you have to pick a temperature, try it at a variety of times and hope to find one, and if you don’t, pick a new one, rinse and repeat.  I’ve heard of makers using over 150 lb of beans to dial in a roast because of this.  This is in comparison to drum roasting where if I can’t have a good roast in 2 roasts I know it is beans and not my ability to find a good roast.

To bring this together, THIS is why I like drum roasting.  It is predictable.  It is short (comparably), uses predominating one method of heat transfer and is therefore by its very nature less chaotic and prone go astray by the flapping of butterfly wings.  I’ve never said I don’t like the chocolate made from oven roasted beans, I just really dislike the method.  Now, I did say above I tend to find chocolate made from oven roasted beans less dynamic and generally speaking that is true.  Once dialed in, oven roasted beans can of course make good chocolate but mostly I find them uninspiring, painting with a really broad brush there.

So, after all that, yes, I find starting with a higher ambient temperature makes for a more dynamic chocolate because it shortens the Development phase.

As for 232 F being the end of the Development phase, there is a two part answer.  The first is that it is an empirical number.  After doing thousands of roasts over the span of a decade I worked out rather intuitively that the region between the low 200s and 230-240 F was important for flavor development.  After that I made a hypothesis and tested it.  It started with defining the end of the drying phase.  The boiling point of water seems a great and logical spot for that, i.e. 212.  After that I ran hundreds of calculations taken out of log book trying to relate flavor characteristic to the time spent in different zones after 212 F.  What I found was I probably could have used any temperature between 230 and 240 F.  I chose 232 F because it was 20 degrees above 212 F and made the maths easy when calculating slopes in my head while roasting.  I mean, I could have picked 235 F and there would have been 23 F  but who wants to divide roast times by 23 on the fly to know how fast the roast is going?  I know I didn’t.  20 is just easier.  To that end, when I was in Brazil last year I had to do the conversions in Celsius and 232 F turns into 111 C and or 11 C difference and by the end I had shifted it to 10 C and too worked just fine.