Coffee Data Science
The beginning of data on the Osma for me
I was at Chromatic to taste some coffees with Hiver, and I asked to collect some data. I have wanted to examine if the Osma had any issues with fines migration, so we grabbed some Brazilian coffee, and we pulled two shots. By examining the particle distributions, I didn’t find evidence to suggest fines migration in the Osma.
I looked at a shot on the Osma and one on the Decent Espresso. I put the videos side by side. Both have some donuting, but if fines migrate, this should still come out in a lateral cut.
The Osma is a cold (or hot) espresso brewer that uses harmonics to shake the coffee and extract a truly cold brew espresso shot. The espresso hits differently, and I suspect some solubles are extracted at different rates than in hot espresso. I’ve been interested in studying this machine by collecting some data.
My concern with some of other experiments, particularly forced fines migration, that the Osma may behave similarly to forced fines migration. I took a spent shot puck, and I cut it into three pieces.
Then I used imaging and computer vision to count the particles. I did this for a regular espresso shot with the same coffee, same grind (Versalab grinder) on the Decent Espresso machine. My initial measurements didn’t show a difference, and I wasn’t sure if the samples were dry enough, so I dried them for a day longer.
There wasn’t much shift, and I expected the bottom layer to have a slightly higher number of finer particles, but it didn’t.
Then comparing to the Osma, I found a similar pattern.
So to compare all of the distributions, I took a few measurements to look at the cumulative effects. Again, there was not an effect in the regular or Osma shots.
I zoomed into below 150 um in particle diameter, and the bottom layer has the same amount of finer particles as the top for Osma. The middle layer seems to have slightly more, so it is possible there was some migration from the top to the middle, but it is hard to determine.
My first samples before drying again didn’t show a difference, and for the first sample, I took additional samples to verify the measurement process. For each one, I took three samples, and for most, the variance was not large.