I often heard worries in the coffee community about a difference of quality in the coffee grind size distribution when grinding with a full hopper versus a single dose of coffee in an otherwise empty hopper.
The idea behind this is that coffee beans forced through the rotating grinder burrs have no choice but to go through whatever openings they encounter between the burrs. In contrast to this, a single bean in an empty hopper will bounce around and may end up passing through a larger hole between the burrs when the opportunity arises, because nothing is forcing it to pass through very fast. This bouncing around of a bean is what “popcorning” refers to.
As a result of this effect, beans that popcorn will end up getting ground somewhat coarser on average. Grinding a single dose of coffee in an otherwise empty hopper will therefore generate two kinds of coffee grounds: a first batch of slightly finer grounds resulting from beans forced through the burrs, plus a smaller batch of slightly coarser grounds resulting from the last beans that popcorned. The result will be a distribution of coffee particles slightly wider than what you would have obtained if you ground a small dose of coffee with a full hopper.
This more uneven distribution of grinds will cause an increase in the amount of coffee particles larger than average, sometimes called boulders. As I mentioned before on this blog (e.g. see this article), only the surfaces of coffee particles extract efficiently when you brew the coffee, and this larger amount of boulders will limit the amount of coffee compounds that you are able to extract quickly and evenly. Several people therefore suggested that it is best practice to grind with a full hopper, and even to grind one bean at a time if you are extremely patient and want to grind a single dose of coffee at a time.
Now that I built a tool to measure particle size distributions, I decided to test all of these claims. They all make sense, but none of these arguments are really telling us how important this effect is. To do this, I ground 3 different doses of 10 grams each on my Baratza Forté BG grinder. I used the same coffee for all these tests, which is important (especially the roast profile may affect how the coffee shatters). In this particular case, I used a relatively light roast of an Ethiopian Guji by Saint-Henri roasters in Montreal. I ground them on setting 6L with the factory-set zero position. In my case, this means burrs would only touch if I went 3-4 ticks finer than 1A. The first dose was ground with a hopper full of beans, the second one was ground by dropping just 10 grams of coffee beans in an empty hopper, and the third one was ground one bean at a time. This last batch bored me to hell.
I measured the particle distribution of each dose by taking 12 different samples sprayed on a 8.5″ by 11″ sheet of paper and combining them together. I took that many samples to make sure that I would have good statistics to be able to resolve minute differences in particle size distributions. I decided on the number 12 because I noticed that comparing the first 6 data sets combined together looked similar to the last 6 combined together when binning the particle size distributions in 16 distinct particle surfaces, so having double that amount of data seemed conservatively good enough.
As a first test, we can ask ourselves how important the popcorning effect is, i.e. how much coarser do the grounds come out compared to beans forced into the burrs ? To do this, we need to compare the full hopper versus the bean-by-bean doses.
What I call the “fraction of available mass” in this figure is the mass of coffee that is available for extraction if you assume that only outer shells of 100 micron are extracted in each coffee particle. This is just an approximation, but it is already more meaningful than just looking at the total mass of coffee particles. For more information, I suggest reading this previous blog post where I discuss a very interesting experiment carried by Barista Hustle to explain why this approximation makes sense. Basically, we want our particle size distributions to contain some information about how the coffee will extract, so we don’t care about weighing the cores of coffee particles that will never be extracted. I also talked about this more here and here. Another thing to notice in the figure above is that the horizontal axis indicating particle surfaces is in logarithmic scale. This means that every shift of e.g. 60 pixels to the right corresponds to a particle size twice as large. On top of each distribution, there is also a single data point with horizontal error bars, that respectively indicate the average particle surface and the spread of the distribution on each side.
As you can see, we are able to see a difference, albeit a small one: the beans ground one at a time are indeed about 0.08 mm² coarser than those ground with a full hopper. To get a better sense of how coarser they are, I compared the bean-by-bean dose to other full-hopper grind sizes on my Forté, and determined that the closest match was to setting 6Q:
Another interesting part of this is that grinding bean by bean generates a slightly tighter distribution, therefore mimicking a higher quality grinder. It might seem tempting to adopt this practice, but do it once and you’ll see why no one does it. It is also possible that this is just an effect of having started the grinder motor before the first bean hits it; this means the motor was rotating at the same exact speed during the full grind. The “full hopper” and “empty hopper” data sets were taken with coffee already dropped on the burrs before the grinder was started, therefore the start of the dose was grinder at a slower motor speed while it was speeding up. I am under the impression that this doesn’t entirely explain “bean by bean” doing so much better, but I will be isolating out this effect very soon to test that hypothesis 🙂
As we saw, popcorning beans are ground approximately 5 clicks coarser on the Baratza Forté BG. A strategy suggested by Scott Rao to grind a single dose was to start at your desired grind setting, and then change your grind size to something slightly finer when you see that your beans start popcorning. This figure above tells us that, if you wanted to do this, it would be appropriate to grind exactly 5 clicks finer when the beans start popcorning. That may require mastery of the on-the-spot Forté fine controls.
However, let’s first ask ourselves another interesting question; does popcorning affects enough beans to have any effect at all on the particle size distribution of a full 10 grams dose ? The smaller the dose, the bigger the effect will be, as the number of last beans bouncing around will always be the same. To answer this, let’s compare the particle size distributions of the full hopper versus empty hopper doses:
As you can see on this figure, the two distributions are virtually undistinguishable. This means that the popcorning affects such a small fraction of the 10 grams dose that I was not able to see any difference with this analysis. As you can imagine, the effect will be even smaller on typical doses which tend to be around 15 to 25 grams. As we saw previously, the last few popcorning beans were clearly affected and they were ground coarser, so there has to be a difference between the two particle size distributions even though it is a very small one. But to put this in perspective, the overall difference on a 10 grams dose has to be much smaller than one click on the Forté, and also much smaller than the difference in grind quality between all different brands of grinders I have ever tested. This includes Comandante’s C40, Orphan Espresso’s Lido 3, Mahlkonig’s EK43, Lyn Weber’s EG-1 and Baratza’s Forté BG. All of these grinders generate particle distributions that a similar analysis can easily distinguish.
The take out message that I got from this experiment is that popcorning has a non-negligible effect on grind size, but it affects a negligible amount of beans in any reasonable dose of coffee. You should therefore not be afraid to grind single doses at a time, because any degradation that results in your particle size distribution will be much smaller than any difference between brands of grinders. I’m hoping this post will alleviate the admittedly first-world and very geeky problem of single-dosing anxiety.
For those geek enough to ask or even to make it this far in the post, I did make my data public on Github.
I’d like to thank Douglas Weber for useful comments, and Victor Malherbe for proofreading.