In October, I visited Mahlkonig in Hamburg to conduct some tests at their factory. We wanted to understand more about the EK43 and why it makes coffee taste better. They have quite a lot of equipment to analyse coffee grinders; the most important of which is the laser diffraction particle size analyser. This piece of kit measures the size of every coffee grind in a given sample and plots the results on a graph. That graph can then tell you a lot about the grinder, and help you understand why different grinders make the same coffee taste and extract differently.
Most grinders will produce a sample that creates a bi-modal curve, meaning there are two distinct peaks. More often than not, there will be a ‘coarse’ peak consisting of particles at the desired grind setting (eg. Grind Setting 6 aims for a peak at 500 microns) and a secondary ‘fines’ peak consisting of very fine particles that break off of the beans during the grinding process. Fines are considered detrimental to the coffee brewing process as their relatively enormous surface area extracts so much more easily than the larger portion of coarse particles; they are also unavoidable as coffee beans are brittle and will always throw off ‘crumbs’ when being ground.
The most desirable result from this test is a very tall and thin bell curve, indicating a large number of very similar sized grinds. This will make an even and delicious extraction much easier; it will also allow you to increase the extraction yield of the coffee to achieve more sweetness, transparency and balance.
It has long been considered desirable for an espresso grinder to produce a bi-modal curve. The theory behind this was that an uneven particle distribution (ie. more fines) would effectively block the flow of water from the machine, allowing the barista to easily produce a 30ml / 30 second espresso. This theory is flawed. An uneven grind size does nothing except create an uneven extraction. Uneven extraction is one of the most common problems in coffee brewing around the world; it’s present in every coffee that has ever been brewed!
The test at Mahlkonig was devised to determine how the particle distribution from an EK43 compared to other grinders for both espresso and filter coffee brewing. Instead of calibrating the grinders to the same burr gap, which generally gives dubious and irrelevant results, we decided to use extraction yield as our benchmark. By using extraction yield as the determinate factor of grind setting, we were able to discern between the different grinders’ abilities to extract the most deliciousness out of a given coffee. This of course being the way most coffee professionals around the world use their grinders (even if they don’t know it).
You may disagree with this methodology, and claim that a better test would be to match the brewing ratio of the different grinders. This is flawed for a number of reasons, the most relevant of which is related to my ultimate goal – a higher AND more delicious extraction yield (the ‘and’ is crucial).
If we didn’t push the grinders in this experiment to create the highest and most delicious extraction yield, we wouldn’t learn which was the best, we’d only speculate as to the results of laser diffraction tests and end up where we were before. That said, the brewing ratios are all quite close, but it was not deliberate.
This experiment is all about espresso. I plan to write another focusing on filter coffee, but it will have to wait.
The espresso test was conducted with one coffee, all from the same roast batch. This coffee was a Colombian Caturra from Finca ‘El Silencio’ in Huila. It’s a Virmax coffee, which means the screen size, density and sorting are all excellent. This is super important for a test like this, because you want as little variation between the brews as possible, leaving the grinders as the only variable. This coffee is also one of my favourites this year, so the tasting component wasn’t exactly arduous either.
I ‘dialled in’ each grinder as well as I could. I was aiming for an extraction yield that was as high as I could hit, without experiencing any dryness, bitterness or the usual side-effects of over extraction. This level of extraction yield ‘right below maximum’ is where a grinder is performing at its best, and it’s where I took the grind samples from.
Here’s some of the parameters
- The grinders used were a Mazzer Robur, Mahlkonig K30, Anfim Super Caimano and Mahlkonig EK43 (coffee burrs);
- The espresso machine was a Kees van der Westen Speedster;
- The baskets were 22g ridgeless VSTs;
- All shots were made with a +/-0.01g tolerance;
- Temperature was 94˚C;
- Extraction Yield was measured twice with a VST LAB2 Refractometer and a fresh syringe filter was used for every sample.
Here’s a quick rundown of how this graph works
X axis is measured in microns, from 0 to 2000µm (or 0 – 2mm). This is the size of the coffee grinds that the laser detected. Every particle is measured and counted. In a perfect world, a grinder would produce particles that were all exactly the same size. This is nigh on impossible, so we always see a ‘spread’ of particle sizes. The best grinders will produce the smallest ‘spread’.
[The X axis of this graph is scaled exponentially (non-linear). This is because we’re measuring the % volume of the whole sample. Fines are really small, so there has to be a lot of them to take up the same volume; if the graph was linear you’d barely see them.]
Y axis is the %/vol. from 0 to 10%. This is the amount of each size as a percentage of all of the grinds. If you look at the red line, you’ll see it peaks at 9% right above 400µm. This doesn’t exactly mean that 9% of the sample is 400µm wide; but it’s good to think about it like that. Don’t worry about the exact percentage too much; just think that higher means more and lower means less.
Mean is the average of the sizes. Mode is the most frequently occurring size (the peak). Median is the size that sits right in the middle.
Particle size graphs are quite often misinterpreted, so I’ve compiled a number of things to note and compare between the grinders. You can look at these numbers however you like and I’d love to hear your thoughts below (as long as they’re coherent, intelligent and polite).
The grinder with the least fines is the Robur [Green]. This was quite surprising although in retrospect it’s fairly obvious. The Robur is producing a massive number of large particles; half of which are larger than 444µm. Fines are produced when you slice coffee beans into smaller pieces. As it turns out, the robur isn’t making the beans anywhere near as small as the other grinders, so of course it should have the least fines.
[Side note: this is a really interesting spanner in the ‘you need fines for proper espresso flow rate’ argument. It may be the case that it is in fact the large grinds – rather than the tiny ones – that are restricting the flow of espresso with traditional ‘espresso’ grinders.]
The grinder with the widest (read: worst) spread is the Robur [Green]. This is unsurprising. Robur shots consistently taste and measure under-extracted, and are notoriously hard to get shots above 19% extraction.
The finest or ‘smallest’ peak (mode) is from the EK43 [Blue] at 295µm. This is super important. The EK43 is making grinds much much smaller and much more evenly than any other grinder. This means you can extract more, without getting over-extracted flavours.
The EK43 is making the most fines (gasp!) but I have two things to say to this.
- The EK43 is slicing each coffee bean so many more times than the other grinders; it’s a super small grind size. With this in mind, there should be much more fines; but there isn’t. It’s only producing 3% more 0-90µm particles than the Robur. It’s doing an admirable job of hitting such a fine grind and producing so few fines.
- The rest of the grinds are so much closer to the fines than with the other grinders. As I’ll explain later, this is one of the defining factors in the EK43’s superiority.
The Robur’s mean of 460µm is very large, and its mode of 567µm is also frightening because it’s so far away and almost double the EK43’s. Most other grinders have their mode and mean values well within 80u of each other. The Robur and Anfim both exceed 1000µm – this means that they create a much higher proportion of larger grinds.
Look at the K30’s curve. [Brown] It’s actually really impressive, and totally surprised me. It’s so much closer to the EK43 than it is to the Robur [Green] or Anfim [Red]. The taste results matched this. We were able to get a much higher extraction and sweeter shot with the K30 than either of those competitors. Its maximum is 800µm – 800µm less than the Robur’s and 1000µm less than the Anfim’s; and only 100µm or so larger than the EK43’s. Don’t just look at the minimum and maximum though; the mean and mode values, and the difference between them is very important too.
The Anfim’s curve [Red] also outperforms the Robur [Green} because the majority of the grinds are so much finer. There is a small bump up around the 1500µm mark, but this is pretty tiny compared to the rest of the curve.
These are by no means definitive results. There are many more tests that need to be conducted to completely understand things. That said, it has certainly helped me understand why my taste preferences swing towards the EK43.
In Part Three of this series, I’ll attempt to correlate this data to taste and extraction.
If you have found this useful and want to enjoy delicious coffee with the rest of the community – register for our monthly Superlatives coffee subscription. Or if you just want to keep up with every thing Barista Hustle – sign up to the Newsletter.