Grind Size and Extraction: the crushed garlic model - Page 4

#31: Post by **nuketopia** » May 17th, 2019, 6:18 pm

another_jim wrote: You can simply say "I don't believe you" instead of posting over and over again.

To repeat: I calibrated the Fuji for cupping according to the SCAA protocol (70% to 75% passing through a size 20 screen), and easily picked out the cups compared to the Bunn because they were more extracted tasting, i.e. softer, with more prominent sweetness and body, and subdued acidity (OK for brewing very bright coffees, miserable for cupping). Then I ground it far coarser (how coarse? -- well none of it passed through the size 20 calibrating screen; I haven't seen a grind this coarse since canned coffee for percolators) and the cups became much closer in taste. I could still pick out the Fuji, barely, for a slightly cleaner taste (presumably the lack of fines due to the much coarser grind setting).

Ok I'm not saying I don't believe you. Let's reset this, I'm not trying to tick you off.

I didn't see how you were gauging. I may have missed it from the other thread.

As you well know, and I mistook what you were agreeing with in a prior is that a simple 1-stage screening as in the SCAA cupping protocol doesn't tell us very much about the grind profile. All it says is that 25% of the grounds are larger than the screen. The composition of the grind could be sharply bimodal with sharp peak of 75% of the mass right below the screen size and 25% right above. Or it could be a perfectly flat mix of absolutely every size of particle evenly distributed. Or it could be a few whole beans sitting on the screen with 75% talcum fine powder passing through.

All of which pass the single sieve test in the protocol. All of which will brew very different cups.

Can we agree on that?

That would appear to account for the differences observed in the cup between your Fuji 220 and other grinders.

It would be interesting to take refractometer readings.

It would be interesting to produce a detailed graph of the particle sizes from this grinder and the others. While they may both pass the same sieve setting, it is unlikely that they're producing similar distributions.

#32: Post by **another_jim** » May 17th, 2019, 7:37 pm

My method was not to judge differences in cups when using the same prep method, but to produce as indistinguishable as possible cups using different prep methods, and then draw conclusions form the differences in prep. The Fuji at a 1200 micorn average grind produces roughly the same flavor as an 800 micron average Bunn grind.

So how can distribution differences explain that?.

#33: Post by **mikel** » May 17th, 2019, 10:06 pm

Regarding distribution, and assuming it's okay to bounce ideas around (not trying to lecture), the science-y model in my head, at least for espresso, has similarities to high performance liquid chromatography. You have a column (ground up coffee), you use pressure to generate flow through the column, and different compounds elute from the column at different times.

With HPLC, you send a sample through the column and different components interact (adsorb) for different amounts of time, based on how the components interact with the column properties (material, porosity) and the temperature-pressure-flow rate. Certain components will leave the column early, others later.

With espresso, the components are in the column itself (rather than your sample adsorbing to it first). The beans, dose, grind, and tamp (and likely preinfusion) influence column properties. The pressure influences flow rate past particles and flow paths through the column. And the interaction between temperature and flow rate (driving how water interacts with the particles) influences the types of compounds and their quantity that leave the 'column' and end up in the cup.

Who knows if that model is correct, but assuming some aspects are close enough, to me, particle shape, size, and distribution from a grinder ultimately influence how water interacts with the particle surface. Fractures and other aspects would be less important to extraction unless they have enough influence on the flow rate through the column and flow around particles to impact water interaction time, temperature, advection, and diffusion gradients.

In short, it's more about column properties that result from particle size, shape, and distribution than how water interacts with
a specifically processed and shaped/fractured particle?

Just throwing it out there.

#34: Post by **nuketopia** » May 18th, 2019, 1:37 am

another_jim wrote:My method was not to judge differences in cups when using the same prep method, but to produce as indistinguishable as possible cups using different prep methods, and then draw conclusions form the differences in prep. The Fuji at a 1200 micorn average grind produces roughly the same flavor as an 800 micron average Bunn grind.

So how can distribution differences explain that?.

Well, we're dealing with non-normal distributions. Terms that apply to normal (aka Gaussian) distributions, like average (mean) and median can be really misleading when talking about non-normal distributions.

Here's a fun, short youtube that shows how wildly different looking graphs, like dinosaurs and stars can have the same mean, median and standard deviations.

#35: Post by **CwD** » May 18th, 2019, 3:17 am

another_jim wrote:My method was not to judge differences in cups when using the same prep method, but to produce as indistinguishable as possible cups using different prep methods, and then draw conclusions form the differences in prep. The Fuji at a 1200 micorn average grind produces roughly the same flavor as an 800 micron average Bunn grind.

So how can distribution differences explain that?.

Different compounds extracting for a similar flavor profile but actual extraction yield not being similar. Extraction needs to be measured, you can't get around that with taste. For picking what to brew with, sure, not for challenging a model.

It could be the case that this is in fact how ghost grinders work. I personally think the idea is fascinating and could aslo have implications stretching to a certain very well known grinder with unusual looking burrs. But the default needs to be "it's probably something that works under currently understood models" until everything is ruled out. And not just for coffee. Any new model in any discipline needs to explain everything the current model can, explain it better, and explain things the current model can't.

#36: Post by **Wiz Kalita** » May 22nd, 2019, 11:28 am

What I'm really missing here is pictures, and it doesn't take an electron microscope to get a good look at a coffee particle. Even at 10x magnification it should be easy to resolve most of it with a good optical microscope. Not sure about microcracks, but if they are present they should give a more jagged shape in general, since you'll get chipping where cracks meet.

#37: Post by **vanilla_squash** » May 22nd, 2019, 4:51 pm

I have an old microscope and an Apex grinder, I can probably find a way to take some pics around 30x if you want

#38: Post by **another_jim** » May 22nd, 2019, 8:10 pm

Thanks, that would be very useful. We have microscope pictures of burr grind particles, so we might be able to see if they are systematically different.

#39: Post by **vanilla_squash** » May 23rd, 2019, 12:40 am

I figured a same day, same coffee comparison between ghost burrs and conical burrs could be interesting. Coffee is Big Trouble from counter culture.

These four are from the apex at 3 clicks towards fine from center:

The following pictures are from my kinu m47 at setting 4+6:

#40: Post by **another_jim** » May 23rd, 2019, 12:57 am

Great pics!

I'm not seeing any cracks in the particles, so the crushed garlic model may be a bust.

OK, set this thread up in another tab, and see if you see any shape differences. The electron microscope has much better resolution even at similar magnification to a light microscope. So the surfaces have a lot more texture showing

But unless I'm imagining things, it looks like the ghost burrs create cloven, flat edges that I don't see in the SEM pics.