Coffee Data Science

How does WDT affect Coffee Particle Distributions?

A dive into the finer things of coffee

Robert McKeon Aloe


In recent testing, WDT was found to be not as effective relative to a blind shaker for espresso puck preparation. One question arose from the ensuing discussions online: how does WDT affect particle distribution? Does it cause finer particles to migrate to the bottom of the puck?

Short answer: the data supports WDT making distribution more even in particle size and shape. Fine particles in this test did not migrate to the bottom after lengthy WDT.

To investigate, I collected some data. I dosed directly into the basket and tamped.

Then I took a center core sample.

I pushed this sample out slowly to get 7 samples from different depths of the puck.

Then I did WDT for 5 minutes to be as extreme as possible. I figured if finer particles migrate, it should get worse over time, but maybe that was a bad assumption.

I then took another center core sample.

Then I measured particle distributions using a drawing board for backlight and an iPhone.

Resulting Distributions

I focused on particles less than 200 um. The smallest this technique can resolve is around 15 um.

I plotted levels 1, 3, and 5. Level 1 is the bottom of the puck, and Level 5 is towards the top. This follows the pattern I have seen before in grinding where the particle distribution is changing throughout the puck.

I then looked at the same levels for WDT. They had less of a gap, and their order changed.

I plotted the cumulative particle distributions at a few diameter sizes for all the levels to see how they changed. The change is clear, but this could be improved. This does shot more evenness per each particle size in terms of cumulative distribution.

Based on Jonathan Gagné suggestion, I looked at D50 for all of the settings. D50 is where 50% of the particles are smaller than that diameter, and 50% are larger.

So the overall grounds are coarser for the first two layers in direct dosing.

Shape Analysis

I looked at the particles using Linear Binary Patterns (LBP) and K-means clustering. Even the particle shapes were greatly changes during direct dose, particularly for smaller particles. While WDT has some strangeness in particle shape in 200um, it has more evenness of similarities than direct dosing.

However, there is some weirdness in layer 5 of WDT as well as layer 2. I don’t have a good answer, but the distributions for these two samples were also odd.

Some of the shape results are a bit more challenging to interpret, and more data would be helpful.

WDT evens out particle distributions from direct dosing. I didn’t have a blind shaker to better understand how it affects particle distribution, but it wouldn’t be a bad item to try.

I think something like Moonraker or Autocomb would be interesting test cases as well.

If you like, follow me on Twitter, YouTube, and Instagram 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.

Further readings of mine:

My Second Book: Advanced Espresso

My First Book: Engineering Better Espresso

My Links

Collection of Espresso Articles

A Collection of Work and School Stories



Robert McKeon Aloe

I’m in love with my Wife, my Kids, Espresso, Data Science, tomatoes, cooking, engineering, talking, family, Paris, and Italy, not necessarily in that order.