Coffee Data Science
During an online discussion, there was a suggestion that intentional channeling might not always have the effect of reducing extraction efficiency. So I decided to repeat the experiment and cause intentional channeling. Then I could look at the flavor and extraction efficiency.
I started with a regular disc and then, after beating it, I poked a hole in it with a toothpick.
I had the injection and although there was some channeling it seemed to partially heal on its own.
Then I pulled another shot without the intentional channel. I also pulled an additional shot with a concave distribution, which you can see in the image below. It’s subtle, but the edges have more coffee than the center, which makes them denser.
Coffee Maker: A decent coffee maker
Coffee grinder: Zero niche
Coffee: Home Roasted Coffeemedium (First crack + 1 minute)
Shot Preparation: Staccato Tamped
Preinfusion: long, ~25 seconds
Infusion: pressure pulsation
Filter basket: 20g VST
Other equipment: Atago TDS meter, Acaia Pyxis scale
To evaluate the differences between techniques I used two sets of metrics: Final Score and Coffee Extraction.
Final result is the average of a scorecard of 7 metrics (pointed, affluent, syrupy, sweet, sour, bitter, and aftertaste). These ratings were of course subjective, but they were calibrated to my tastes and helped me refine my shots. There is some variation in the results. My goal was to be consistent across each metric, but sometimes the granularity was challenging.
Total Dissolved Solids (TDS) is measured using a refractometer and this number, combined with the output weight and input weight of the coffee, is used to determine the percentage of coffee in the cup, called Extraction Efficiency (EY).
For these three shots, I poured them into three cups so I could see how the extraction efficiency developed over time. The intentional channeling worked the worst.
As for taste, the intentional channeling produced a noticeably worse taste, while the concave one was slightly better.
In the case of TDS and EY, targeted information sharing did not work.
In terms of time, the deliberate channelization occurred faster, which is probably due to the faster flow through the canal.
I looked at the underside of the discs and there was no obvious evidence of a channel on the underside of the disc. I thought there was, whereas the concave distribution had more murky spots.
