Erichimedes wrote:Count me as extremely interested in this. I have enough DIY skills and could source parts and assemble my own, as long as I have a list of resources and someone I can ask questions to, especially on the software side. I know someone in town I should be able to get 3D printed parts from. I haven't had the chance to read all of this thread yet, but I'm sure I could get up to speed on most of it.

I'm running out of room inside the main box of my roaster, but I would be interested in a stand alone version to start out with. I can always stuff it into the main box down the road, depending on how much I end up using the data.

I need to tweak some of the 3D component designs, then I'll start posting the CAD files. I'm presuming most DIY'ers have access to filament-style 3D printing, so I'll need to figure out if the gas flow covers for the sensors can be printed that way (I'm currently using high-resolution SLA resin printing).

Initial form-factor is for attaching to the base of an IKAWA, but I'll be posting .step format CAD files so people with CAD skillz can modify the design to suit their application, in advance of me coming up with a generic standalone design. FWIW the IKAWA version doesn't have to be attached to an IKAWA, if one can tolerate the odd shape

Erichimedes wrote:I'm not quite familiar with the data you're posting here, but man am I interested in it. Am I reading the graph right in seeing that there's a noticeable drop in moisture RoR at a specific BT there at the beginning? Maybe 130C or so? Do you see that peak at a similar BT between different greens or does it tend to change?

I'm not familiar with the data I'm posting, so that makes two of us

I suspect the initial moisture RoR (MRoR?) peak is from the typically 12%ish greens hydration turning to steam as the greens are heated above 100C. The MRoR seems to peak as the BT ramp rate starts to slow down, so that seems like a math thing to me. I tend to use the same fast ramp for the first 30 seconds of the roast, which might be why the MRoR curve peaks in the same place on the BT curve.

Addendum to post #99, adding a second greens to the chart.

Exact same batch size and thermal profile for two different greens:
Light blue lines: Klatch Nicaragua Buena Vista Obata - washed process
Dark blue lines: Klatch Sumatra Mutu Batak - wet hulled process

Note the moisture RoR curve was scaled and shifted to fit nicely on the graph. It's zero point corresponds to 12 on the absolute humidity scale.



So, yet another beautifully smooth moisture release curve! I'm now starting to think perhaps the stepped thermal profile may be the root cause. Data this orderly also puts me on the lookout for any systematic issues with the sensor performance. The tubing system is cooling the air at the moisture sensor into the 35-40C range, well within the operating limits, and stays nearly flat over the course of the roast.

The graph shows how different greens make different curves. Interestingly, the Sumatra starts out lower, but makes a bigger jump at first crack. Maybe there's a correlation here: the Nicaragua has been well balanced, smooth, and mild, whereas the Sumatra is distinctly more intense. I welcome speculations on what's going on here.

Also, despite the different greens with their different moisture curves, they reach first crack within 15 seconds of each other. I attribute much of that alignment to the very steep approach into higher temperatures. It kinda makes sense that a 'drier' green might crack sooner, but that's just one of many factors in play.

Some other stats to noodle on:

Moisture-based development time ratio (BT-based ratios are identical)
Nicaragua = 27%
Sumatra = 30% (don't have a roast color analyzer, but I can tell the Sumatra is a teensy bit darker)

Total moisture release (moisture AUC), relative to zero, which may not be realistic)
Nicaragua = 140 (3% 'wetter')
Sumatra = 136
If I calculate from a 'room-ish' 10% baseline, the difference is closer to 8%

Sorry haven't been able to play much with this project lately. I have recently had some time and was able to validate a similar setup to GDM528's. I am using different sensors and didn't want to change the software... I am still using the iOS app which connects to IKAWA and the humidity sensor as well as CO2 and TVOC. Still far from GDM528's setup though!

The pump works great, it is a much neater solution than my copper tubbing, I get a constant sensor temperature at around 30C. I just tried the humidity sensor, since it's basically all I am using, have yet to come up with a use for TVOC and CO2.

I just 3D printed the adapter from the silicon tube to the humidity sensor. It's PET-G 0.15mm layer height and 0.4mm nozzle. I guess this means GDM528 design should be ok with a filament printer.

GDM528 wrote: Light blue lines: Klatch Nicaragua Buena Vista Obata - washed process

Can you share your IKAWA Home profile ? I have the same coffee and I'll try to at least get the humidity curves using this setup with your profile.

Setup :





Here are 4 roasts of the same bean Costa Rica San Roque - SL28 using the pump.

Esteve wrote:Can you share your IKAWA Home profile ? I have the same coffee and I'll try to at least get the humidity curves using this setup with your profile.

Happy to see you back!

I did a crude test of the airflow delay through the tubing:
With the pump running off the 3.3V supply, it was pulling air through the tubing at about 2 seconds per meter. My integrated setup has about a half-meter of tubing. The shorter tubing = warmer air, so in my setup the gas temp is 35-40C, still well within operating limits.

Don't see an inline filter for exhaust debris in your photo, but FWIW the filter in my setup is hardly showing any accumulation.

Here's the recipe from post #98:
https://share.ikawa.support/profile_hom ... IAGgAiACgA

The above profile is an unusual stepped profile, so here's a more "Pro-like" profile for the Obata:
https://share.ikawa.support/profile_hom ... oAIgAoAA==

These were the curves from that RoR profile:



Both profiles are towards the darker end of the spectrum (City+). I can try going lighter if you'd like.

And just for fun, I tried making my Excel chart look like the display on the Pro100x app:



The actual Pro100x curve is lifted from mgrayson's post here: IKAWA Pro100x. (hope you're ok with it Matt, consider it the sincerest form of flattery We used different greens and batch sizes, so the comparasion isn't direct. Despite that, I find it encouraging that a homebrew setup is getting pretty close to factory - it means we can share and collaborate with Pro100x owners.

GDM528 wrote: Don't see an inline filter for exhaust debris in your photo, but FWIW the filter in my setup is hardly showing any accumulation.

No, I haven't got to it yet. Also missing the capacitor in the DC motor line, I did the bare minimum to test, hehe.

GDM528 wrote: The above profile is an unusual stepped profile, so here's a more "Pro-like" profile for the Obata:

Ok! I'll try this one, my roasts are somewhat lighter, total roast time around 6 minutes for 75g, targeting DIY meter around 110. I am curios about yours, it'll go for nice thicker espressos

I got this, sounds right ?

Name C4 170 270 D800

{'time' : 0 , 'temp': 500},
{'time' : 303 , 'temp': 1697},
{'time' : 1204 , 'temp': 1975},
{'time' : 2099 , 'temp': 2220},
{'time' : 3002 , 'temp': 2424},
{'time' : 3897 , 'temp': 2578},
{'time' : 4804 , 'temp': 2697},

{'time' : 0 , 'power': 204},
{'time' : 4804 , 'power': 205},

Ending time end_roast_time=4804

Graph as comparison from one of my roasts:

Ok, I did 3 x 65g roasts of the same Nicaragua Buena Vista Obata greens. The roast looks great, I got 16.7% WL (65g to 54.2g). Very uniform roast and can't see any defects, not tried it yet but so far looks great.

In terms of humidity I think I got very similar results to GDM528's, albeit I start from a lower absolute humidity points. Either my sensor is different or most likely it is less humid today here. I sent GDM528 all the data in XLSX format on a DM, if anyone else wants to play with it I am happy to share it.

All 3 roasts are very similar to this :


Here are the 3 roasts:

Finally got the time to compare Esteve's roast data with mine. Results are complicated, but potentially interesting to anyone attempting to share notes about their roasting recipes. This may be a legitimate question to ask another roaster: "Are you high?"

Esteve and I roasted the same greens, from the same supplier, purchased within weeks of each other. We roasted the greens using the same model roaster, and the exact same roasting profile. The dominant difference is our respective locations: Spain, and the United States (NW). So this is largely a test if location matters when roasting coffee.

These were the greens used: https://www.klatchcoffee.com/products/u ... ?bgIndex=3

This was the roasting profile we both used: https://share.ikawa.support/profile_hom ... oAIgAoAA==

This is a humidity/gas thread, so the chart includes the rate of moisture release (MRoR) data we collected with our respective roaster mods. We each ran three batches, and please note how consistent our respective results were, especially in the context of the scatter in raw data we gathered, which I will show later in this post.



Apparently, location matters. My data shows MRoR peaking at the 6:00 mark, right where the roasting curve calculator predicted it. Esteve's data peaks at about 5:25, considerably sooner. The difference between our sets of curves is far greater than the scatter within our respective runs. So I'm considering this a statistically significant difference, and 30 seconds longer development time is significant for an eight-minute roast.

We're using different electronics for measuring the exhaust humidity, but an earlier test (post #88) shows they all give pretty similar results, and the differences couldn't account for the 30sec gap.

Esteve has suggested that altitude might be a factor:
Esteve: 1,476ft above sea level
Gary: 200ft above sea level

Esteve also found this relevant video: http://www.youtube.com/watch?v=8xaXaqeaMNM
The conclusion of the video is that altitude matters: higher altitudes roast darker. You can jump to that section of the video at 34:45 and watch/listen for about eight minutes. Big Caveat: this conclusion is for the IKAWA - the conclusion is likely different for a drum roaster.

I speculate there could be a fan speed correction for altitude that IKAWA has decided not to implement... yet. I gather the Kaffelogic does implement an altitude correction (it's just math), so market competition may eventually drive IKAWA to get with the times. In the meantime, I suppose it's up to the users to figure out the correction factor to use when sharing recipes - topic for a different thread.

The sensor I'm using (BME688) can measure air pressure... I think I'm gonna start logging that data

Also noteworthy how this particular thermal profile curvature/ramp created a completely level stretch in the MRoR curve for three minutes during the browning phase. Compare this to post #99, where holding the temperature steady caused MRoR during browning to be slightly negative (and flat) and this latest run with a rising temperature caused the MRoR to be slightly positive (and flat). I invite speculations as to what's going here: how can significantly different thermal ramps cause a completely flat MRoR? And how might a positive or negative MRoR during the browning phase affect the taste of the coffee?

This next chart shows the raw data we each gathered from the runs, for chamber/BT and absolute humidity:



Our in-chamber temperature probes show statistically different curves. I expect that's because of different placement, different probe types, and different downstream electronics. This is consistent with my previous observations about the veracity of user-added thermal probes. This is why the roast curve calculator only references inlet setpoints - it's the only common, stable, immutable temperature for the IKAWA.

The absolute humidity readings are quite scattered in their offsets. The first run in a sequence is typically higher, perhaps due to some sort of carryover from run to run. One of my three runs was taken five weeks ago (the highest of the three), when the weather was considerably warmer, with different ambient humidity and pressure. Fortunately the MRoR calculation erases the baseline offsets.

Despite all the scatter in our thermal and humidity data, the MRoR data is remarkably stable and repeatable. I've already stopped listening for first crack, and now I'm thinking about ways to ignore my thermal readings too This feels like another magic moment from measuring exhaust humidity.

Great analysis! I agree that location and environmental conditions have to have played an important role at the time difference of the FC.

I have been double checking and can't really find any other source of deviation from measurement errors. 30 seconds seems a large difference. Even if the humidity sensor response time is different, it would be too much.

Esteve also found this relevant video: http://www.youtube.com/watch?v=8xaXaqeaMNM

Also from the video it seems that darker roasts get more impacted by this difference in location. One thing that I don't like about the video analysis is that an exhaust profile is being used, if altitude changes the behavior, they have an extra correction from regulating on exhaust temperature. It would have been a more fun experiment if they used inlet temperatures (at least more fun for home users).

Despite all the scatter in our thermal and humidity data, the MRoR data is remarkably stable and repeatable. I've already stopped listening for first crack, and now I'm thinking about ways to ignore my thermal readings too This feels like another magic moment from measuring exhaust humidity.

Agree! Humidity turned out to be a very actionable measure.

Now it's time that you two guys arrange a swap of your roasting equipment, to rule out any differences of the machines

Well, actually...

Spain and the U.S. have different mains voltages, so add that to the many reasons why it might be considered crazy to swap hardware. And speaking of crazy, Esteve's data included the response of the heater to the programmed profile, and the tracking was crazy-tight - the awesome power of electronic feedback loops in wiping out external influences.