Sunday, November 1, 2015

Die Erste Deutsche SCAE Röstmeisterschaft 2015

The first German SCAE coffee roasting championship took place last week on 26-28 October @Probat in Emmerich. The competition followed a slightly simplified, but authorized, version of the 2016 rules and regulations of the World Coffee Roasting Championship (WCRC) to select the candidates to compete next year in Shanghai for the world mastership.

The previous, more informal, roasting competition dates back to 2009 and took place at the Kaffeeolympiade. This year, the German chapter of the SCAE decided to take the challenge and selected a competitor for the WCRC 2016.

While all other German SCAE competitions, like the Barista Championship, will take place this November at the Kaffeeolympiade in Munich, the roasting competition was relocated for organizational reasons to the Probat showroom in Emmerich.

Sponsors & Supporters

Probat not only provided travel support for the winner, the rooms, the catering, the roasting machine and other equipment, but also staff to help on technical issues, the cupping, and all those little things. It was nice to work with you Marcel Knitter, Jörg Vogel, Daniela Nowitzki, and Jens Roelofs!

The green coffee, one from Brazil and a single origin from Panama, was sponsored by List und Beisler GmbH and InterAmerican Coffee.

Together with Daniela and Marcel from Probat, Thomas Schweiger (German barista champion 2010 and 2012), Marco Cremonese (SCAE Italy) and Tim Albrecht (J.J. Darboven GmbH) made up the jury.

Farmcomp donated a Wile Coffee moisture meter as second prize. I donated a Tonino roast color meter as third prize and supported the event with the Artisan roast logging software.

Many thanks go to my two volunteers Josef Gander (Kuntrawant) and Bono Gargolov that made it all the way from the south to help me out.


It was a close call. The winner and his runner ups are as follows.

  1. (Alec) Alexander Pfuhl (Rehm & Co., Hamburg)
  2. Hannes Fendrich (Five Elephant, Berlin)
  3. Benjamin Pozsgai (moxxa.caffè, Köln)

Technical Setup

The sample roasting was done on a two barrel Probat sample roaster. Green coffees moisture was measured using a Wile Coffee by Farmcomp. Green density was calculated by taking the mass of coffee (in grams) divided by the volume of coffee (in liters), measured in a container of known volume.

Roasting Machines

All roasting took place on a 1kg Probat Probatino as well as a 5kg Probat Probatone. Both machines were equipped with dual bean temperature (BT) probes as well as a second probe in the exhaust (ET) connected to the roast logging system Artisan. Additionally, the network connection of the 5kg machine was hooked up to Artisan to log and control the burner setting.

Both machines have fully modulated burners that can be set from 0% to 100% power. The burner setting of the Probatino controlled via a mechanical knob only, while the one of the Probatone can be set via its touch panel or from within Artisan via a slider or direct numerical keyboard short cuts.

Both machines were connected to separate chimies to avoid cross influences and were running at constant air flow, verified to be at an optimal value before the event.

Roast Logging

Artisan was used as roast logging system configured with a 3sec sampling interval (oversampling turned on). The autosave function was used to avoid the loss of data. The automatic CHARGE and DROP functionality was not deployed and instead the start and end of roasts was manually indicated by an assistant.

The candidates could choose to see the measured temperatures in Celsius or Fahrenheit, with all but one deciding for Celsius.

Artisan was configured to show besides the ET and BT signals, the computed bean temperature rate-of-rise and the ET/BT projection lines. The phasesLCDs were turned on and could be configured to show the development ratio during the last phase of the roast. The turn point (TP) event was automatically computed by Artisan and the yellow point (DRY) was automatically set at 150C/300F. Some candidates manually entered the first-crack start/end events, thus this was not mandatory and kind of difficult to hear on those well isolated machines.


According to the simplified rules, the candidates roasted just one single origin coffee from Panama Boquete (SHB) of 755g/l density and 10.6% moisture. Each candidate had 30min for the competition roasting. All of them did two roasts from which they had to select one to be evaluated.

Here is the winning profile as well as the one of the two runner ups (click to enlarge). The bean temperature (BT) is drawn in dark blue, the environmental/exhaust temperature (ET) in red. Below the temperature curves, there is a brown/yellow line indicating burner changes (0-100%). The burner level is also drawn as light dotted line. The light blue line gives the computed BT rate-of-rise. The line below the graphs indicate the date and time of the roast, the coffee roasted, the batch size, the weight loss as well as the resulting ground color (Tonino scale).

The competition roast profiles (the selected as well as the discarded ones) of all candidates recorded on day 2 are available on GoogleDrive as *.alog files, to be opened from within Artisan, and as PDFs. The ones with a file name ending with C are in Celcius, the ones ending with F are in Fahrenheit.

A sortable table (click on the column heads) allowing to compare the key data points of all 10 selected competition profiles is available online in Celsius and Fahrenheit.


  • First-crack start (FCs) was manually added post-roast in the selected profiles of Alec, Benjamin, Lutz, Matthias and Ralf at 198.54C (the average of the FCs temperatures of those that indicated it) to get an idea of the roast development.
  • All candidates roasted w.r.t. temperatures in Celsius, except for for Hannes who worked in Fahrenheit. While all profiles shown here are presented in Celsius, the repository contains also the profiles converted to Fahrenheit.
  • Where the burner change events were not keyed into Artisan using its slider or keyboard mechanism during the recording, those events were reconstructed from the recorded burner level signal using the Artisan quantifier mechanism.
  • All batches were renumbered and the layout was streamlined for optimal readability.
  • Most candidates selected their second roast and discarded the first one, except for Jochen and Lutz who selected their first roast. The discarded profiles are stored in a subfolder "Discarded".


  • The penalties on deviations from the roasting plan should be reconsidered. The 2016 rules states "(to deduce) one point for every two degrees to either side of desired temperature". Variations in sampling intervals between the roast logging system and the roasting machines may already account for a delta of several degrees, especially on CHARGE, where the temperature changes rapidly (you can see this in Alec's winning profile, where the actual CHARGE temperature might have been 3sec earlier at 204C instead of the logged 197.6; however I doubt this makes any difference). During charge the temperature readings already vary largely from the begin of CHARGE to the end. Besides, the rules only talk about degrees and not about the temperature units. However, the absolute temperature difference of n-degrees is considerably different if taken in Celsius or Fahrenheit. In retrospect, a faster sampling interval, say 1sec or even faster, than the 3sec configured for this competition would be advisable. Also the rules should fix the type and the dimension of the probes to be used as this also influences their reactivity.
  • It should be defined if any automatic CHARGE/DROP algorithm should be used to determine the start and end of the roasts. The advantage of using such a system, that takes its information directly from the bean temperature signal, is that one would not rely on manual logging here and thus all candidates would be treated equally. However, such systems tend to fail on lower charging temperature with a slower decline of temperature (some systems, not Artisan, fail even on declining temperatures before CHARGE like in the selected profile of Hannes).
  • The roast color deviations depend on the color scale in use. The rules read "For every one point past the allowed two points variance, one point of deduction will be made". Two points variance is usually already the measuring tolerance of the common devices (but also this depends heavily on the scale used). It cannot be assumed that the candidates learn the variations on the color scale in use (which might even be non-linear) by just measuring a few sample roasts. This makes the prediction of a color value for the production roast a pure gamble.
  • The above observation were not critical in this competition as it can see from the summary tables linked above that the results were dominated by the cupping results. Thus cupping not even a day out of the roast is questionable too.

Thanks to all for the good time we had while woking together!

Oh, and Alec, the winner,  came up with an almost perfect Natural Roast up to FCs. Nice!