Coffee Data Science
Zerno Coffee Grinder Across Burrs
Coffee Analysis by extraction, particle distribution, and shape
A few months ago, Zerno sent me a grinder because they liked what I did on analyzing coffee. They also sent me two extra sets of burrs. There are a lot of data collection options, so I decided to start with a small test across burrs and grind settings to gain some intuition. These results were delayed a bit by summer; I’m glad to take a look at these particles and their shapes.
I will compare espresso shots, particle distributions, and particle shapes. I have completely switched from the Niche Zero to the Zerno, but I have not done a comparison between the two grinders yet.
For each burr set, I pulled three shots at 100um, 200um, and 300um burr gaps. I typically use a 200um burr gap. I used 4 month old test coffee, and I focused on extraction yield, not taste. For each shot, I also pulled them as salami shots (three cups per shot at the 1:1, 1.5:1, and 2:1).
Burr Sets
I think the labels on my boxes were wrong, so here is a key. I have Stock, Silver, and Red.
Equipment/Technique
Espresso Machine: Decent Espresso Machine
Coffee: 4 month old coffee from Chromatic
Shot Preparation: Staccato Tamped
Profile: Flat 1 ml/s without pressure restriction
5 minutes of Thermal Pre-infusion
Filter Basket: 20 Wafo Spirit
Other Equipment: Acaia Pyxis Scale, DiFluid R2 TDS Meter
Metrics of Performance
I used one set of metrics for evaluating the differences between techniques based only on coffee extraction. I focus here strictly on extraction metrics while acknowledging the importance of taste.
Total Dissolved Solids (TDS) is measured using a refractometer, and this number combined with the output weight of the shot and the input weight of the coffee is used to determine the percentage of coffee extracted into the cup, called Extraction Yield (EY).
Extraction Yield
The three burr sets had similar performance with the maximum performance at 200um, but what was interesting was that the 100um shots cut across and achieved the same EY by the 2:1.
The Silver burrs:
The Red burrs:
I compared the burrs to each other, and the stock burrs are a little better at the 1:1.
For the 200um gap, the Stock and the Red burrs are nearly identical.
There is more separation between the silver and red vs the stock at the 300um setting.
Particle Distribution Curves
I measured particle distribution using a camera and a board. The trends are pretty similar.
For 200um, the Silver and Red are very close.
At 300um, there is some differentiation at the larger particle sizes.
Looking at each of the burrs, the settings are wild.
But we can double check because there is some cross over in the curves for 200um and 300um. For each of these images, I have a zoomed in and zoomed out image from the tele and wide cameras on an iPhone. I use the zoomed in because the particle resolution can get below 100um in particle diameter, but the zoomed out is a good sanity check.
This zoomed out looks as expected for changing grind settings. There seems to be a wider distribution which could be due to resolution issues and multiple particles looking like one particle.
On to silver burrs, which is also odd.
For the Red burrs, I also took a few other measurements. I was surpirsed at how close 50um and 100um were as well as 100 to 150um. I don’t have a good reason for this difference.
Particle Shape Comparisons
Let’s look at particle shapes using Linear Binary Patterns (LBP) and K-means clustering. This is a way to discuss the similarities between particles. I used 16 clusters in the k-means clustering because that is usually the maximum for coffee particles that makes sense.
This part is a bit complex because the burr gap is in 100um, 200um, and 300um while I separate out four groups by particle size bins based on the diameter from 100um to 400um.
The lighter the color, the more similar they are which is why there is a diagonal of no differences. Not only do the particle shape similarities vary across burr but also particle size. This first one has all the data in order.
The particle size is in the title with the burr gap on the x and y axes.
We can sort this data by the burr gap. 200um has some large distinctions, and at 400um, the Red 100um is isolated and alone.
Looking at just the Red burrs, the Red 100um is similar until 400um. For 300um, every setting is pretty close relative to the other ones. 50um has some interesting shape differences for 100 and 200um.
While it may seem like there are large differences in the similarity matrices, we can compare the overall vectors using the percent of particles that fall into each group. For all the particles, there isn’t much difference.
For individual bins, there are quite a few binning differences.
Take a breath! This was a lot to digest, and I’m still digesting it. I’m not sure what this says about anything. Most people would say different burr geometries would produce difference grinds, and they certainly do. However, the particle shapes have a lot of overlap for some particle sizes. It seems burr geometry primarily affects some of the particle distribution but not all.
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.
