Hi,

This is a continuation to the modification I did on an IKAWA Home roaster. I post it in a separate post as the data capture could be helpful on other roasters.

I have added a humidity sensor to the Ikawa exhaust, still lots of work to do but it captures well the FC. The idea is to take an air sample from the exhaust, cool it down via a long path and then use a sensor to measure humidity.

It could be fun to brainstorm what other sensors could be useful!

The sensor I used is HIH8121-021-001 an I2C sensor. I also like another sensor that has a faster response time but it was capacitative and harder to interface...

Here is the setup:



For the sensor, I used a plastic case just to be able to have the sensor get the most airflow, and shield it a bit:



Closeup of the sensor before "installing" it .



And 3 roasts same greens same program, the last is 50g the rest 75g. The FC coincides where the humidity spike is 280s in the roast.





the 50g one:



And two for a different profile / greens, I am quite sure the drying phase is too short. It's a profile I am still working on and looking at humidity it seems I must be pushing it...





Hope you find it interesting!

Best,
Esteve

Duuude, you're using Geisha for roasting experiments?

That's an interesting roast profile, especially in the context of your BT curve looking considerably more 'curvier'. It looks like two ramps in the first two minutes, then flat thereafter - is that right?

It looks like the second profile is the same shape, but 3-4C hotter. That has me wondering if you're primarily seeing the humidity readings peak whenever the inlet hits 236C, hence the shift in the peak for the latter run.

I scanned the datasheet for the humidity sensor: it shows the operating temperature range as 125C, but the 'compensated' range only goes to 50C. I interpret that to mean the readings could get wacky above 50C. Also, this humidity sensor looks to be meant for standard air chemisty, whereas the gasses coming off a roast are more complicated, and may include chemicals that mess with the sensor. Have you tried running the humidty sensor with the chamber empty?

As for other sensors, it might be possible to repurpose optical smoke detectors, but they may not be sensitive enough and don't come in breadboard-friendly form factors. Using an IR (3um wavelength) source/sensor would enhance sensitivity to water.

Another possibility might be to measure the dielectric strength of the exhaust gas, presuming it's lower when there's more 'junk' in it. As with the optical sensor, this would involve some custom design work, but at least high-voltage supply modules are readily available on Amazon.

Regarding the drying phase, I've experimented with deliberately making it short. My theory was to maximize water present during the Mailliard phase - analogous to the 'wet method' for caramelizing sugar, which is alleged to produce better flavors: IKAWA Home - profiles I'd also point out that IKAWA has several curated profiles that spike the temperature right at the start of the roast, which may have a similar effect. This thread on rehydration may also be interesting/relevant: Green coffee rehydration where you can pre-engineer the water content going into the roast. And it's just a matter of time before someone starts misting water on the air inlet of their IKAWA during the roast...

Also, cooler air holds less water so the humidity may measure higher to significantly higher than actual.

That particular sensor also measures temperature, so it can compensate and calculate relative humidity - at least up to 50C and 90% RH.

Yes, but if you take hot air and cool it with no other changes, the humidity goes up or so I understand.

Hi,

Thanks for the feedback!

That Geisha is the coffee I am drinking now, been short on time and needed more for my stock I like the profile, it is a modification from one of the Ikawa PRO's profile converted to home. I saw no risk at sticking a probe in the exhaust

The second profile is one that I am playing and not happy with the results, looking at the humidity and your 236C comment could imply that this is where most of humidity is released but the bean has not dried uniformly yet. If that makes sense... I was upping the inlet temperature as the BT probe was consistently lower than what I thought I wanted.

but the 'compensated' range only goes to 50ºC. I interpret that to mean the readings could get wacky above 50ºC.

Correct, the sensor compensates in that range.

I first tried with a shorter tube cooling to around 100ºC and got really bad results, a steady declining humidity line... That's why this second try the tube is so much longer and I have that "casing". I like GDM528's 1 inch fitting on the exhaust, it could be interesting to use something like that with an attached coil to extract and cool the sample. The main concern cooling the sample would be condensation, I think metal tubing and preheating, I always do an empty run to heat the roaster, would produce a temperature gradient on the coil avoiding condensation.

Ikawa Home thermal performance (the 1 inch fitting I am referring to)

Also, this humidity sensor looks to be meant for standard air chemisty, whereas the gasses coming off a roast are more complicated, and may include chemicals that mess with the sensor. Have you tried running the humidty sensor with the chamber empty?

Good point, I didn't think about the chemistry of the exhaust gases. On an empty roaster it produces a decreasing humidity on heating phase, stays flat and at the cooldown it decreases again. I tried once only. I'll try to put a graph, the one I have is cut as I was adjusting the location of the sensor.

To get exact moisture release data we would also need a humidity/temperature sensor at the intake, or we could use an empty run as the baseline humidity for the room. It needs to be the exact profile though, as changes in the fan speed change output RH.

Also, cooler air holds less water so the humidity may measure higher to significantly higher than actual.

I think there are 2 different things to account:

1) Humidity sensor compensation. Look at this capacitative (IST P14 Rapid-2) humidity sensor with no temperature compensation, you can see the formula :

https://www.mouser.es/datasheet/2/1426/ ... 084349.pdf

The temperature compensation is just to account that the capacitator's relation to humidity is different at different temperatures.

2) Then the fact that relative humidity changes with temperature. Assuming no condensation, no leaks, the absolute humidity at the start of my white tube (hot) and the absolute humidity at the end (cold) is the same. The amount of water vapor present in the air remains the same. However, as you cool down the air, the relative humidity will increase, as cooler air can hold less water vapor compared to warmer air.

Knowing RH and temperature at one end we could calculate RH at the other end of the tube. However I am not sure how useful it would be to try to calculate the RH inside the roasting chamber, as long as results are repeatable and changing greens and weight gives meaningful results, at the end is all relative. Maybe a rate of moisture release.

I got some of the ideas from this post :
https://www.williamstownroasters.com/bl ... ty-studies

As for other sensors, it might be possible to repurpose optical smoke detectors, but they may not be sensitive enough and don't come in breadboard-friendly form factors. Using an IR (3um wavelength) source/sensor would enhance sensitivity to water.

I have an air quality(SPS30). They basically measure particle size and quantity present in the air. I might give it a try. Using IR to measure the humidity could be a nice solution, I assume you could route a heat resistant fiber cable and avoid needing to cool down the air sample.

There are also spectrometer sensors, quite too expensive for me to just give it a try though...

Best,
Esteve

Great stuff! At a far less granular level, how would you use humidity readings to identify the start and end of FC, and perhaps its shape? Being able to tune FC would be crazy, but it could make a difference you can taste. I've seen experienced roasters comment on wanting a certain "intensity" of 1C, but that's very vague. And, different coffees vary in their audible intensity. The last raises another question. I've seen references to attempts to detect the sound of 1C. How would you tune a microphone to do so and where would you position it to avoid damage or accumulation of coffee oils and debris? You would want to be able to separate that signal from other roaster sounds. I know I just went off topic, but correlating humidity and sound could be useful.

Digital signal processing technology (chips, software) is remarkably commonplace nowadays, and relatively inexpensive (less than $10) to incorporate. The signal processing can 'learn' and ignore the background roaster sounds. The audio pickup doesn't have to be in the chamber, so it can stay out of harm's way. The engineering is only hard the first time, so I'd bet it won't be long before other roasters include what Roest offers.

The SPS30 particulate detector is interesting, as it can profile the debris coming off the roast. It should be actively reporting 'stuff' over the entire roast - hopefully the particle composition changes over the course of the roast so that landmarks can be picked out. I'd be concerned about vaporized oils fouling up the delicate internals of the sensor.

A few narrowband LEDs and detectors could measure the 'cloudiness' of the exhaust gas, similar how the roast color analyzers work. Placing detectors/sensors on and next to a water absorption peak could differentiate water from the other gasses. This chart of water absorption is from Wikipedia (credit: By Kebes at English Wikipedia, CC BY-SA 3.0, https://commons.wikimedia.org/w/index.p ... d=23793083)



Some university must have already done this...

Agree, the sound detection doesn't seems that difficult. The microphone can be on the case. Probably the hardest would be to make it very robust, so many different 'crack noises' to test... I would filter out the background roasting noise and then listen for high power signal with short duration. I don't think a matched filter would work...

I looked at some CO2 detectors that work in a similar way to what GDM528 is mentioning, some are quite simple to interface. (https://sstsensing.com/product/sprintir-co2-sensor/)

Have not been able to find much for humidity via light with good response time.

Wow. The modifiers have not just matched the Pro100, but now the Pro100x. Great job!