07/08/2026
Documentation Study EP7
From Roast Curve to Cup Quality
The reference roast shown below demonstrates excellent profile repeatability, reaching nearly the same drop temperature and following a very similar roast trajectory. The question I'm trying to answer now is whether that level of repeatability also translates into the same roast development and, ultimately, the same sensory experience in the cup.
This roast serves as my current full automation reference profile. At first glance, the curves appear highly repeatable, with nearly identical end temperatures and a very similar roast trajectory. However, one observation continues to stand out. Reaching the same drop temperature and following the same profile does not automatically guarantee the same roast development or the same cup quality.
From an engineering perspective, repeatability begins with the control system. From a roasting perspective, repeatability is only validated when the coffee delivers the same sensory result in the cup. Those are two very different objectives, and bridging that gap is what this research is trying to understand.
For the current stage of development, my priority is not to achieve full automation at all costs. My focus is first to understand energy input, heat transfer, bean development, and sensory outcome before allowing automation to take over more of the roasting process.
One interesting observation during this stage of testing was how the automation itself behaved. During one of the full automation reference tests, Artisan automatically executed the DROP command earlier than expected after the target profile conditions had already been satisfied. This happened even without an active DROP alarm configured. At this point, I can't say with certainty whether the behavior was related to playback logic, profile completion, PID behavior, or another internal automation sequence. Rather than assuming a cause, I documented the behavior and added it to the list of items that require further investigation.
Observations like this remind me that understanding the software is just as important as understanding the hardware. Sometimes the expected behavior is obvious, and sometimes it raises new questions that deserve a closer look.
This is also why I don't recommend leaving a roaster completely unattended during automated roasting. Automation can execute commands consistently, but it cannot yet replace the judgment of an experienced roaster when unexpected behavior or process variations occur during a roast.
Every modification—whether in firmware, control logic, PID behavior, sensor filtering, software configuration, or roast strategy—is documented, tested, and evaluated before moving to the next iteration. That's simply how engineering works.
Ultimately, the graph is only one part of the equation. The real objective is to produce the same thermal history inside the bean, the same level of development, and ultimately the same cup profile. A roast can look nearly identical on screen yet still express different sweetness, acidity, body, balance, or finish during cupping.
This project is not about proving that automation is superior to manual roasting. It's about understanding how engineering and roasting science can work together to improve consistency while preserving the experience, intuition, and sensory judgment that remain essential to every roast.
The more I test, the more I realize that automation isn't the destination—understanding the coffee is. Every roast, whether successful or not, adds another piece to that understanding.
This research is still a work in progress, and I'm always open to comments, suggestions, constructive criticism, and technical discussions. If you notice something I may have overlooked or have a different approach, I'd genuinely appreciate hearing your perspective. Continuous learning and collaboration are part of what make this journey worthwhile.
