Excuse me if this has already been suggested-- I didn’t see it, so:
Coming from a Kaffelogic to a Bullet, it would be nice to have the ability to design a profile in Roastime, and have the roaster automatically adjust its settings to target that profile.
This is a bit different than a recipe, where the operator has mapped out the sequence of setting changes. Instead, the operator maps out the desired roast profile, and the machine compensates to achieve that profile.
Artisan has the ability to design a profile, which you then can use as an overlay, but it is up to the operator to adjust the settings.
I’m sure this is not a trivial problem to solve, but it would be a killer feature!
The reason why true automated profile tracking doesn’t work reliably on the Aillio Bullet comes down to a combination of its step-based control logic and the technical limitations of induction heating.
Unlike roasters such as the Loring or Kaffelogic, which support smooth, continuous power modulation, the Bullet adjusts heat, fan, and drum speed in discrete steps (P1–P9 or P1-P14 on the PRO, F1–F9, D1–D9). This means it cannot make the fine, real-time corrections needed to follow a target curve precisely. Even when using Artisan’s PID mode, the system can only jump between fixed levels—causing overshooting, delayed reactions, or instability in the roast curve. Of cause I tried that but failed miserably
This limitation is especially pronounced due to the nature of induction heating. Induction doesn’t allow for linear, continuous adjustment of output like a gas valve or resistive heater might. Instead, it relies on controlled electromagnetic pulses to generate heat within the drum, which are typically implemented as fixed power stages to maintain system stability and hardware safety. Achieving high-resolution, real-time modulation with induction would require complex power electronics far beyond what the Bullet is designed for.
In contrast, systems like the Loring use modulating gas burners that respond smoothly to control inputs, while the Kaffelogic employs high-frequency PWM and low thermal mass, enabling rapid, precise adjustments. Both were built from the ground up to support automated curve tracking.
In summary, the Bullet’s combination of discrete power control and the inherent response limitations of induction heating make precise, automated curve following technically infeasible—at least without significant external control logic and even then, only within narrow tolerances.
Hello,
thank you for your explanation, I thought that the recipe mode could read a curve and could do repeatability automatically, like a stronghold s7 since I thought that the R2 had the same kind of probe. So this raises a lot of questions for me about my future purchases if there is no real repeatability mode.
Well since I designed this thing I would like to clarify.
Induction heating in the Bullet is created by a continuous high frequency switching magnetic field in the order of 20000 times per second.
If you wanted to you could regulate the heat at a much higher frequency than any gas valve, heating element etc. So to say that this is an inherent property of induction heating is not correct.
It was a design decision to use discrete power steps, as it make the Bullet easier to use.
@morgan it is not a trivial problem to solve if you want to make a very stable system that adapts to different environments and batch sizes. We chose not to go down that path and instead focus on recipes.
I can imagine that replicating exact profiles is indeed “not a trivial” problem to solve (and the more I think about this, the more I love the understatement in that expression!).
Yet, I wonder, if this is something that could either be developed as a future software update for the R2
OR,
would this require a complete re-working of the Bullet’s architecture from the bottom up (and hence is basically not going to happen …ever)?
This is not meant to be a nagging kind of a question, but merely driven by my curiosity.
I imagine that beside induction power, and fan speed, even drum speed would require continuous adjustment in order to replicate a profile for different volumes of beans.
This could be developed in the future, there are no technical limitation (the induction board power can be adjusted in 1W intervals), it is just about pouring enough resources at it.
Wow - learned something new here. Thanks for the clarification! One of our engieers explained that to me that discrete steps helped to prevent interferences at certain frequencies. So as i am a computer scientist and not an electrical engineer, i cannot judge on that.
Nevertheless, very insightful indeed. Thanks Jacob for your explanation - always keen to learn
