"Let's see what this baby can do."

I have one of the new Ikawa Home machines released late last year. It called out to my hacker DNA...

I've retrofitted a custom thermocouple into the chamber to monitor the temperature of the spinning bean mass. It's adhered using high-temperature Kapton tape, so it hugs the wall of the chamber without interrupting the flow of the beans as they spin around. The tip of the thermocouple is shaped to position the tip squarely in the bean mass.



The thermocouple is monitored with a data logger I custom-built using a Raspberry Pi Zero W, a thermocouple converter, and a tiny display for temperature readout/graphing. The hardware and software are designed for very fast response and high sample rate: >100C/second and ~7 samples/second. The data logger is WiFi enabled so I can conveniently transfer the thermal data into a spreadsheet, roughly 4,500 data points for a 10 minute 'run'. Not shown is a 3D-printed/laser-cut enclosure, and a substantial amount of shielding to keep the Raspberry Pi's radios from interfering with the thermocouple converter.



Elsewhere on these forums, I mentioned how I thought the Ikawa could thermally 'turn on a dime'. Now that I have this spiffy temperature monitoring setup, I can test that assertion.

Using the (free!) Android version of the Ikawa app, I created a set of setpoints to pulse the temperature in the chamber. That will show two aspects of the Ikawa's thermal performance: how fast can it change temperature, and how it settles into the new setpoint. I picked 200C and 250C for the setpoints, which corresponds to about 160C and 210C in the chamber, which nicely line up with the browning phase and first-crack temperatures.

I did two runs: first without any beans in the chamber, and again with a small, 50g dose of natural-processed Panamanian beans. Yeah, I know they claim up to 100g, but I'm currently interested in the tightest connection between the temperature setpoints and the actual bean temperature - the lower the bean mass the faster it will respond and track.



Sure enough, the Ikawa can change temperatures pretty fast: 50C in about 10 seconds. The hitch, however, is ringing and overshoot when pushing that fast. It takes about 30 seconds to calm down if you tell it to instantly change temperature, which could lead to scorching at the higher setpoints. So even if the bean-mass temperature (the green curve) is changing more smoothly, the exterior of the beans is being subjected to more extreme temperature excursions. Probably not an issue for reaching the browning point, but I'd avoid it at the FC and SC temperature levels.

Also, interesting to note how the bean(green)/no-beans(blue) curves completely overlap in about 2 minutes, which I think implies the beans have reached equilibrium with the chamber temperature.

Next I tried slewing the temperature more gradually, 30 seconds for each transition:



That largely mitigated the ringing/overshoot for the 50C steps, but no improvement for the initial 100C step. So, I'm inclined to limit slew rates below 2C/second - which is still pretty fast IMHO. The bean-mass curve looks better behaved and still reaches equilibrium 2 minutes after start of the temperature ramp.

I think I understand why Ikawa omitted the bean temperature sensor for the Home versions, but it's still a shame we're forced to fly blind when developing roast profiles for our own beans. However, based on the test results there are a few rough rules of thumb you can follow:

1) Bean temp is about 40C lower than the setpoint temperatures.
2) Take at least 30 seconds to ramp between setpoints.
3) Allow at least 2 minutes for the beans to fully 'soak' at the target temperature.

I have a few more experiments in mind, and open to any suggestions from the 'bean gallery'

My Ikawa has schooled me again.

I rejiggered the temperature setpoints to produce a roast result that would be reasonably drinkable (I'm using my 'good' beans for this). I separated the roast into three distinct phases: drying, browning, and development. I went into this thinking a 100g batch size would be sluggish, so I allocated 3 minutes for each phase.



50g, 100g, whatever. Once I overlaid the curves it was clear I could skip the 75g batch size. The difference in temperature readings between 50g and 100g is negligible, and the ramp profiles for the steps into browning and development are virtually identical. Regardless of batch size, each roast reached equilibrium with the empty chamber readings in just 2 minutes. I did not expect this result.

I got a few theories:
1) The airflow is so high it overwhelms the thermal mass of the beans. Nonetheless, whatever amount of heat energy the Ikawa pumps through the chamber will still be spread across all the beans, resulting in a lighter roast for larger batches.
2) The distinct drying phase at the start of the roast is doing its job, by eliminating the effects water would have had on the later phases. I think this will become a regular part of my general-purpose roast profiles to get a more predictable result across a range of bean origins and age.
3) The tip of the thermocouple is sampling at a position where the Ikawa's airflow is particularly good at controlling the temperature at that specific location, regardless of batch size. It happens to be located pretty optimally for 50g batches - maybe not so much for 100g batches.

Despite the temperature results, the physical dynamics are a LOT different between 50g and 100g batch sizes. Bean motion is very fast and consistent across the 50g batch, whereas there's a lot of velocity variation in the 100g batch. The beans in the 100g batch circulate through parts of the chamber that likely run cooler than the outer rim where the thermocouple is located. Evidence of that can be seen in the resulting color difference:



It may not be super-obvious, and I don't have a color meter to quantify it, but the 50g batch size is a shade darker, and more uniform - it just looks better. At least for the beans I'm using, doubling batch size seems to result in about a one-step reduction in roast level, e.g. City to American.