Coffee Data Science
Espresso Basket Rumble: Robot vs Pullman vs Decent
A throw down of precision basket superiority!
I had a few espresso filter baskets that I haven’t yet analyzed, and I also updated my analysis code, so this was a good opportunity.
I am to look at hole size distribution for the Robot, Pullman, and Decent filter baskets.
Analysis
I use standard image processing techniques to do filter hole analysis. I have previously noticed that most analysis done on baskets counts the number of pixels above a threshold but my technique is more advanced using the actual amount of light to get some sub pixel accuracy.
Additionally, I updated the algorithm to correct all hole openings to circles. Some times they are elliptical. Then I applied a global normalization to accounts for any skew for the camera. This new algorithm accounted for a lot of the camera error and lens distortion.
From the Bottom
Here is the view from the bottom with a hand annotated blue circle for the hole detection algorithm.
From the top
I computer the hole diameter for each filter.
While a box plot is nice, a summary table shows some interesting bits. I also have the hole spacing. In terms of hole size, all the baskets have a similar STD or standard deviation which is awesome because they’re all supposed to be precision baskets.
The two Decent baskets had a slight different hole sizes especially from the top. I’m not sure how that effects flow because the number of holes is different as well. There is a slight difference in the number of holes which is due to the threshold and the original image. Imaging filter baskets is trickier than it seems.
In terms of the total area with holes, they all have a very similar area except for the Robot. That means the flow would be more restricted for the Robot filter, but the Pullman and Decent would preform similarly.
We can also look plot hole size average. The top holes are larger than the bottom for the Robot, but for the other three, the top holes are larger. That means even the way the coffee exits is different. The trouble is that the amount of area for the Robot is smaller than the rest, so it is hard to know what the effect would be.
We can look at all the bottom values and top values separately. The difference in DE filters is odd. I would think the hole sizes would be closer. The Pullman has smaller hole diameters but a larger number of holes.
Spatial Analysis
I plotted each of the images with a false color for the hole sizes. This is to understand if the distribution of hole sizes is random or do certain areas have clusters of smaller or larger holes that would contribute to channeling?
To understand the false color below, the color is adjusted from 0 to 1 based on the minimum and maximum hole diameter of that particular filter.
From the Bottom
The Robot has a cluster of holes in the center which could cause a faster flow channel to develop in the center. The Pullman has a similar issue to one side. For both the DE filters, they have some clusters, but the hole sizes seem more random.
From the Top
From bottom, there aren’t any major issues in terms of hole distribution. The DE 18g has some holes that are merged at the top. This was due to thresholding that particularly image. The filter baskets are metallic so they reflect a lot of light. Hence, the photographic challenge.
Total area seems to be the statistic that really differentiates filter baskets from one another. It isn’t to say the Robot basket is better or worse, but different. It will have a lower flow, so adjustments would have to be made between the two baskets to get a similar flow.
I enjoyed looking at these four filters. The key for this analysis was having really nice pictures. I’m glad the patterns of hole sizes across the filter weren’t too localized. The main conclusion of this test was a prediction on flow. It seems the Robot filter basket has a lower total area which is directly proportional to the flow. If the same amount of coffee goes into the basket at the same grind as the others, the water through the coffee will go slower as a result.
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.
Further readings of mine:
Collection of Espresso Articles
A Collection of Work and School Stories
Espresso Filter Baskets: Visualized
Espresso Filter Comparison: Pesado vs VST
Espresso Baskets (VST): Ridged vs Ridgeless
IMS Superfine vs VST: A Small Sample Espresso Filter Comparison
Espresso Simulation: First Steps in Computer Models
