Why TDS/extraction yield measurements are not a good way to evaluate taste profile/quality? - Page 5

Want to talk espresso but not sure which forum? If so, this is the right one.
User avatar
another_jim
Team HB

Postby another_jim » Mar 16, 2019, 1:21 pm

Jake_G wrote:If you were to do EY% for each particle size, you would find a gradient of high extraction yield for small particles to low extraction yield with large particles with the average EY% being close to the extraction yield of the average sized particles.


I think this is more a theoretical difference than a real one. It assumes each particle is isolated and getting the same water and flow exposure.

Here's my theory: Highly extracted particles reabsorb solubles, while poorly extracted ones continue extracting. With enough time and interaction between particles, extraction levels even out, so that only the overall grind size and pour times determine the overall extraction level. Moreover, we know perfectly well how to do a badly and unevenly extracted shot: Overstuffed basket, no head space, no preinfusion, and above all laminar, "espresso-porn" flow through the puck, so there is no sideways diffusion or interaction between particles. Finally, finishing the shot by some weight standard, rather than when the flow is clear, so that the bottom of the puck (which recharges with the solubles from the top) is less extracted than the top. Pretty much everything many of the people who pontificate about extaction think is good practice. This is why these posts contain so much nonsense. If you have laminar flow through the puck, and are not worried about fully extracting the bottom layers, the only way to salvage anything is to have ann Uebergrinder producing perfect particle sizes.

It's the espresso version of Laputa.
Jim Schulman

User avatar
NelisB

Postby NelisB » Mar 16, 2019, 2:29 pm

Jake_G wrote:Sorry, poor wording.

If you were to take a wider distribution grinder through the range of EY% from very low (short shots with coarse ground coffee) all the way to very high (long shots with very finely ground coffee), my assumption (given my rambling talk of particle distribution and flow rates and contact time, etc...) is that the flavor would be best (tasting neither under, nor over extracted) for this particular grinder and coffee at a lower extraction yield than it would if the same experiment were carried out on a grinder with a tighter distribution.

Does that make sense?

This is my take on the whole "EK43s taste under-extracted at anything under 21%" sort of thing. It's not that the wider distribution grinder at lower EY% will taste better in absolute terms than a tight distribution grinder, but the relative maximum for balanced flavor for a wider distribution grinder should occur at a lower EY%.

As I said before, I should have spent more time putting my first post together...

Cheers!

- Jake


Check!

nuketopia

Postby nuketopia » Mar 18, 2019, 12:46 pm

Well, the problem is that human beings are horrible test instruments. If you put the exact same coffee at the exact same temperature into four identical cups and set a label card next to each describing each as wildly different, people will report tasting completely different things.

We need tools like refractometers to make objective measures of machine and process performance. Taste and other purely human observations are so subject to bias and unreliability that we really have no idea of making any positive changes or not.

The refractometer indirectly measures TDS. I use the VST-III and follow the recommended process of mixing, filtering and cooling before measuring so the tool works repeatably. So I can make comparisons.

The refractometer can't "taste" anything. That would be akin to deciding which mother loves their child most by counting the socks in the laundry.

But I will say that I haven't ever had a good tasting shot with a poor extraction yield, ever. And I've measured a lot of good and bad tasting shots.

I have to say that when TDS approaches 10%, I find the beverage too dense and it tends to set off my bitter detector. The same beverage, diluted somewhat, often tastes better. So there are limits to density and at least this human's taste senses.

My experience with grinders is limited. I've certainly tasted shots from an EK43, as a local shop I like has one they use for single-origin shots. They're good and generally, I have enjoyed them, especially if the barista has really got it dialed in. Haven't had an opportunity to measure these or pull the results on my equipment.

I do with some frequency, measure my own results to see how things are doing and to really dial things in.

I have yet to personally see a case where one grinder produced radically different EY% with radically different flavors. I haven't tried every grinder, of course. But I have tried a number of them, measuring both flavor and objective measures. I'm very happy with what I have.

cunim

Postby cunim » Mar 19, 2019, 9:20 am

nuketopia wrote:Well, the problem is that human beings are horrible test instruments. If you put the exact same coffee at the exact same temperature into four identical cups and set a label card next to each describing each as wildly different, people will report tasting completely different things.

We need tools like refractometers to make objective measures of machine and process performance. Taste and other purely human observations are so subject to bias and unreliability that we really have no idea of making any positive changes or not.


Actually, there are ways to make objective measurements of subjective dependent variables like taste. Trouble is, these methods require the use of statistical methods and those rely on a reasonable number of data points. That said, it's not hard, just laborious. Ask Nestle or Christie's how they tell if one product variant or another is more likely to succeed. They determine which tastes better using tasting panels. Same thing, but to provide external validity (relevance beyond the test conditions) you need a representative sample of members of the relevant population. That means random sampling of panel members from that population (which in our case is lovers of good coffee). Combine that sampling with blind (preferably double blind) testing, and you start to get somewhere.

Contrast that with a cupping session or a wine tasting. You are getting a small number of data points from testers that are absolutely not representative of the population of drinkers. The testers are experts, the drinkers are not. That is why it is so common for what the gurus tell us to diverge from what we experience. They are basing their judgments on a different context and skill set from ours. To them it tastes like blueberries and leather (Corinthian). To me it tastes like a frog on a lilly pad. Sigh.

nuketopia

Postby nuketopia » replying to cunim » Mar 19, 2019, 11:54 pm

Oh I agree. The entirety of color science for instance, is based on methodical and statistical approaches to human observation. The process used to develop CIE 1934 color space and models of human tri-stimulus vision are great examples of such science. Color is very critical to a variety of industries. The really tough thing is that we can't "measure color" in the sense that color is a perception, while we can very accurately measure light, wavelengths, reflectivity and a variety of other physical qualities. Humans are great at seeing minor differences between two similar colors, but really terrible at picking the same color on its own, and easily fooled by certain color combinations.

A very experienced professional wine, spirits or coffee taster will readily admit that his/her tasting ability is subject to error. They usually don't make critical decisions based on a single session. I know some wine makers and the good ones understand the human limitations.

Getting back to the refractometer question: I honestly don't have any data to argue for or against taste-vs-EY%. I can say, I have used the refractometer to establish at least for me, that espresso tends to taste "good" in the range of 8.5-9.5% TDS of density and around 17-21% EY most of the time. I use it as a tool to help me keep my own methods in line and sometimes as a means to dial in something that is difficult, or to assure that I've done everything possible with something that isn't working for me. I have to say, it is a reliable tool when used correctly. I have noted interesting things, such as EY goes up a little after a thorough cleaning of the grinder and burrs and that the bean temperature prior to grinding (say freezing vs. heated) makes a difference.

I haven't seen any particular difference between grinders for instance, where gross differences in EY materially worked for one and not the other. I have though, seen grinders that were inferior, and had a difficult time achieving EY in the range one expects for espresso output and duration. I have tried competent examples of flat and conical burr grinders and not found them to perform differently in terms of EY when adjusted for similar time/output. I think they may be different somewhat in flavor, but that whole double blind process and statistical sample size would be necessary to say so conclusively.

I may go visit a local shop I know with an EK43 and manual paddle machine they use for single-origins and take the refractometer with me, order a few shots and see what is happening with it, and try the same beans at home in my setup. I would suspect they're not going to be wildly different.