Truth be told, this post was going to be split into three parts, one each for the namesake accessories in the title. But before I went about making plots, I reached out to Jonathan Gagné to check if I could trouble him to take a look at my data and find any shortcomings with it. What followed was a very humbling (for me) conversation during the course of which Jonathan not only offered to plot the data for me, but also presented it graphically in a way that is a lot more intuitive than anything I could have come up with.

Wider and wider

If you ever head to EAF’s fancy-baskets thread, be prepared to have your mind blown with the seemingly infinite number of basket options that have become available. While I’m personally not enthusiatic about the myriad of designs that claim to affect espresso extraction in specific manners, I was specifically interested in high hole coverage baskets with spherical holes of the same diameter spread evenly across the basket’s bottom surface. Some popular examples of this are Weber Unibasket, Sworks high/standard/regular flow basket, Flair’s high uniformity basket and PCL’s high flow basket to name a few. These designs also tend to have straight walls, although this is not a given (traditional baskets like vst18 have a slight taper to allow easier knocking out of pucks). The idea is that having holes closer to the edge of the basket should provide even hydraulic resistance across the puck’s diameter. This is in contrast to traditional baskets that tend to have a thin area along the perimeter devoid of holes, which in theory tends to obstruct water flow in that region.

However, even if the basket allows for perfectly even flow across the bottom, it’s not guaranteed that water dispersed atop the puck was done in an even manner. There’s a few ways to try and control movement of water flowing into the puck:

• Shower screen: A bit of black magic in itself, shower screen design can have a tangible effect on how water flows into a puck but falls outside the scope of this post
• Puck screen: Lance Hedrick and Stéphane Ribes have both shown that puck screens have an effect on extraction in traditional shot regime
• Grouphead dispersion: Water isn’t entering the shower screen out of nowhere, but rather being dispersed in varying ways depending on machine.

Specifically in the DE1, water first enters through the inlet in the grouphead manifold into to the top half of the dispersion block, which has been designed in a way such that water first pools in a reservoir before it reaches the exit holes, at which point water should be exiting each hole at equal flow rate. After this, the bottom half attempts to make water flow in a way such that it covers as wide an area as possible when exiting onto the shower screen.

It should be noted that while the older dispersion block design also followed a similar design approach, the holes in the bottom half didn’t quite extend as much to the edges. I also wouldn’t be surprised that the brass dispersion was in turn already a step above most grouphead dispersion designs available in the market when it comes to water flow coverage and evenness.

What does it say

For most part we’ll be looking at the data (raw version can be accessed here) in terms of the difference between difference in extraction between the edge of the puck vs. the rest of it (which I’ll refer to as “center” for convenience but is actually a pretty large chunk). Jonathan kindly went about plotting it in a bunch of ways so we can try and visualize it in as intuitive a manner as possible. The first set of plots are a combination of box plots of individual shots overlaid with box plots of combine averages and a KDE (kernel density estimation) of the different categories of shots.

A quick reference on how to interpret these plots amidst dense amounts of data:

• Yield loss refers to (edge EY – center EY)/center EY
• The further away you’re from zero, the higher the delta is between edge and center
• Negative value indicates edge extracted less than center
• Positive value indicates edge extracted more than center
• The vertical line plots are bar plots. The longer the lines the higher was the variance in edge-center difference (so one shot can have low difference and the next one can have high difference)
• The curved plots, also called violin plots, show the probability of difference between edge and center. Wider sections of the “violin” correspond to higher probability
• Error bars are with 95% confidence intervals

With this in mind, here are some observations one can gather from these plots:

• Traditional shots with traditional basket suffer most from reduced edge extractiom
• Turbo shots don’t seem to suffer from this effect as much
• Using a modern basket nearly eliminates this effect and also seems to favor edge extraction a bit for turbo shots
• Puck screen mitigates both reduced and increased edge extraction. It seems to make flow more even where it’s not and doesn’t have a negative effect if flow is already even.
• New dispersion pattern improves evenness for traditional shots, and favors edge extraction heavily for turbo shots when used without a puck screen

While not totally unexpected, the extent to which a puck screen helps is quite illuminating. Also what’s interesting is that the function of puck screens doesn’t seem limited to just pushing water to the edges, but rather evening it out overall. One should remember though that these comments apply only in the context of these dispersion blocks, and it remains to be seen if it helps with setups that don’t have even water dispersion prior to the shower sceeen.

That’s not intuitive

If the earlier plots seemed confusing, worry not, Jonathan’s got you. He went through the extra effort of generating box function plots so that the findings are more intuitive.

When explaining the process of generating these plots to me Jonathan mentioned the following:

• A Monte-Carlo simulation was done by including possible sources of error when measuring extraction
• Tds was assumed to have a 0.01% uniform error
• Weight was assumed to have a uniform error of 0.1g
• The box function plots were made by scaling edge and center EY in a way that it matched shot EY.

Some differences to note between the smooth plot (fig. 8 top) and the box version (fig. 8 bottom):

• The smooth model has center EY aligned at the same values
• These were generated by stitching quarter of an ellipse onto a rectangle
• The width of the rectangle is considered to be a function of basket width
• The vertical size of ellipse drives the EY at edges of the plot.
• Since there was some mass loss during sample transfer, the shot’s EY was taken as an anchor (because it’s a direct measurement) and the sum of scaled sectional EYs was ensured to match shot EY.
• Making such plots more accurate will need more cutouts instead of the one big cutout that I did for the “center”

I need a drink

If you’re like me, you will have spent a ton of time looking at those plots and marveled at how tiny changes in design can result in incremental changes that allow for more efficient extraction. This isn’t me saying go rig your system with Decent’s (read Ben Champion’s) new dispersion, but rather that we haven’t paid enough attention historically to water flow evenness, and that there’s no one single trick that will solve issues with extraction (although a thick puck screen seems to be a good start).

Acknowledgements

• Jonathan Gagné – for doing the plots and detailing out his thought process for analysis
• Lance Hedrick – for putting me on to this (you owe me a coffee for being barely correct about the new dispersion)
• Misha from decent discord – for kindly lending me the sworks high flow basket to test thoroughly and sending my fascination for espresso through the roof

Subscribe to keep you pocket filled with coffee science.