Coffee Data Science

Cross-Sectional Espresso Puck Slicing

Exploring particle distributions post-shot

Published in

Towards Data Science

6 min readAug 27, 2021

The espresso puck is quite a mystery in terms what goes on inside between preparation and throwing out a spent puck. While it is theorized that particle distributions shift inside the puck, there has yet to be much dissection of the puck to examine particle shifts. I aim to take a look at a cross sectional cut of an espresso puck to help determine how the particles differ from top to bottom.

I started with a regular shot pulled on the Kim Express. One fundamental challenge was the donuting inherent to how the water enters the shower screen. The other is that I wanted to extract as much as possible. I typically pull a 1.3 to 1 output to input shot around 22% Extraction Yield, so I pushed through a few more shots worth through the puck to get to an output yield more similar to most people.

Puck Slicing

Espresso pucks are solid but brittle. I ended up working through a few trials for feasibility to simply slice the puck. I started with a very sharp knife, but that didn’t help the brittle nature. I tried a diagonal cut, but the cutting still let things fall apart in an uncontrolled way.

Then I had a revelation when emptying out a wet puck from a moka pot; dry coffee grounds are brittle, but wet coffee grounds stick together. I tried cutting a few wet pucks before I pulled the shot I was really interested in studying.

I slowly wet the puck from top to bottom to make sure it was all wet but not soaked.

Then I cut at an angle.

I aimed to cut as many layers of the puck as I could. I ended up with 7 layers.

I let these cuts dry over night, and they dried pretty quickly.

Test Day

I decided to collect data on the outside of the puck and the inside separately, so I cut the outside bits away from the inside bits. Additionally, I measured gTDS to get an idea for how many solubles were still left in the grounds.

From the gTDS perspective, Most of the values were low. The only rising trend was that the inside gTDS went up as you approached the bottom but not by much.

Particle Distributions

I looked at raw particle count rather than estimated for the volume it takes up. As a result, the finer particles are a higher percent just because there are more. This is the combined result of inner and outer puck:

There are some minor differences, so let’s take a closer look starting with the Outer measurement with just the top, middle, and bottom layers:

Again, these follow the same trend with a slight increase in finer particles for the L7 or bottom of the puck. However, L4 is not midway between the top and the bottom.

Looking at the Inner measurement, we see a similar pattern. There seems to be big overlaps in distributions and no clear trend aside from a slight bump at 90um for the L7.

We can look at the combined measurement, and the pattern is similar.

Focusing on the Other Layers

One way to deal with errors in measurement and understand trends is to look at the different layers and see if there are trends and shifts in the distribution. We can start with the top (L1) through L4, and put the L4 through L7 below.

Both of these don’t show a layer to layer shift in any trend sort of way.

We can slice this data with a few other plots below that provide a similar conclusion.

Overall, this experiment showed the particle distribution is relatively the same throughout the puck, and there doesn’t seem to be a migration of fine or coarse particles in either direction as a result of espresso. Fines migrating should have caused an increase in finer particles in the bottom and a decrease in finer particles for the top, but this type of pattern was not observed.

It is quite possible there is a measurement error, and I would prefer to have taken multiple measurements per sample, but I was time limited.

If you like, follow me on Twitter and YouTube where I post videos of espresso shots on different machines and espresso related stuff. You can also find me on LinkedIn. You can also follow me on Medium and Subscribe.