Monday, September 27, 2010

French Press Fines

This post is a loose follow-up to the cliffhanger at the end of my long and headache-inducing post on the chemistry of French press foam.

After concluding that French press foam might act as a sort of chemical scrubber for coffee fines (insolubles), I was eager to start comparing different methods for removing fines from the brew.

While I stand firmly with those who hold the palate as the final judge of coffee quality, the palate is inevitably subjective. I set out to measure suspended French press fines concentrations objectively -- and [SPOILER ALERT] probably failed to do so.

I prepared coffee in five different ways. First, I prepared coffee traditionally, brewing in a French press for 4 minutes and shaking in the coffee bloom at 3 minutes. After that, I tried the Wendelboe / Hoffman technique, the Coffee Catcher press technique, the Coffee Catcher pull (referenced here and here), and the double Coffee Catcher technique (videos here and here).

Referencing a video at The Other Black Stuff, I decided to pour the prepared coffee in each case through a clean paper filter and weigh the results using a gram scale.
The remaining residue looked pretty scary!
After weighing the filters, I found the following (per approx. 900 ml of liquid coffee):

Traditional - 3.1g
Wendelboe/Hoffman - 2.7g (-13%) (56 g solids removed)
Coffee Catcher Press - 3.0g (-3%)
Coffee Catcher Pull - 2.8g (-10%) (103 g solids removed)
Double Coffee Catcher - 2.3g (-26%) (134 g solids removed)

Did these numbers surprise me? Yes and no. Yes, in the sense that the results are not necessarily correlated to my palate's perceived cleanliness of flavor (which I assume is a direct correlation to suspended insolubles). My palate tells me all four variants are cleaner than the traditional press, but I would have rated the Coffee Catcher Press method as substantially cleaner than the number would indicate.

However, I wasn't surprised if the results reflect heavier, non-suspended particulates. This, I think is the "mud" that sinks quickly out of the brew, and I suspect it is not related as closely to perceived cleanliness. I noted that the three methods that involved the removal of large particulates easily beat the other two methods. Also, the only double-removal method tested removed far more weight than other methods. It's worth noting that the Coffee Catcher Pull measurement is skewed slightly by experimenter error: the disc slipped slightly part way up the carafe and leaked an unknown amount of solids back into the brew.

Is any of this significant? I'm still working on that one. I set up the test hoping to test differences within tolerances of a tenth of a gram. I was careful to test as large a volume of coffee as possible in order to exaggerate per ml differences. My statistics knowledge is rusty, but preliminary results indicate that these results are not necessarily significant (i.e. differences in weights could be chance alone). My instincts disagree with this assessment, and I'll post more as I discover more.

Finally, I'm circling back to square one. How do I measure suspended French press solids? This experiment shows that they appear to be too light and/or too small to be measured via paper filters. As an example, notice the tiny residuals that leaked through the paper filters on the drying plate. A smear like this leaked through the drying filters in each case. Perhaps these are the particulates I'm hunting for.
A dry smear of particles too small to be trapped by the filter. Are these the elusive suspended solids?


  1. Statistics Update: I've discovered that the formula I originally used to test for statistical significance requires number values to be greater than 1 to work. Since my values are much lower than that, I will need to find a different formula for evaluating the significance of these results.

    The good news is they might yet be significant after all!

  2. Statistics Update: I changed the data ratios to reflect concentrations in larger volumes (so that numbers would rise above 1), and again, the test indicates no statistical significance for the data.

    I may be missing an alternate testing model that would be more appropriate to the data. I'll continue to investigate the problem.

    Of course, it may also be true that my kitchen lab equipment is simply not precise enough to generate results that stand up statistically. This would be a sobering word to coffee kitchen experimenters everywhere. How much are we merely pontificating?

  3. Statistics Update: I think I've solved this problem. Assuming the proportion of precipitated fines in any given press made by x method is distributed normally, I suggest that the standard deviation of said distribution is much less than the +/- .1g tolerance of the scale. The reason the standard deviation should be small is because the input (ground coffee) is controlled very tightly (+/- 1 g).

    Assuming the standard deviation is less than the tolerance of the scale, then the standard deviation is irrelevant. All results differing by more than .1g should be significant. Results with .1g would be a statistical tie.

    The results should be grouped like this:

    3-3.1g: Traditional, Coffee Catcher Press

    2.7-2.8g: Wendelboe/Hoffman, Coffee Catcher Pull

    2.3g: Double Coffee Catcher

    My conclusion is unchanged. Brew methods that remove particulates from the press yield significantly less precipitated solids than traditional brew methods. From this test, of course, I have no idea whether this difference is correlated to flavor differences.