Cracking into the Extraction: when and where the espresso puck brews.

[another_jim]

Andy Schecter and I have been studying the details of espresso extraction. Andy has worked on overall brewing ratios; I've been trying to find out how the puck extracts over time. We're also working on how much of the puck's weight extracts at different doses and basket geometries -- this is a work in progress. We're hoping when all this is combined, we'll have a model of the espresso extraction process that will aid in designing better baskets, groups, and grinders.

My work has been the rather thankless job of brewing and tasting partially spent pucks and measuring their TDS, which is a rough measure of how much brewable coffee they still hold. After a lot of trial and error, I hit on the following procedure:
1. I ground a fresh coffee sample
2. With the same coffee, I made six shots, stopped at the first drop (time 0), and 6, 12, 18, 24 and 30 seconds beyond that point.
3. I took the spent pucks, and divided it into a top layer, middle layer, and bottom layer.
4. The fresh sample and the puck sections were oven dried for 2 hours at 225F, then brewed at a coffee to water ratio of 1:20. In addition, the fresh sample was brewed at 1:40 (50% strength) and 1:80 (25% strength). I measured the TDS of all the samples, and tasted them, rating their strength by comparing them to the fresh samples.

There was a very close correlation between my taste strength ratings and the TDS measures, so I'm only reporting the TDS data at this point. The graph is illustrated below:

(Click on the image to view it at full size)

The horizontal axis shows the shot time. The vertical axis the TDS readings. The three pairs of dashed horizontals lines shows the range of TDS readings at 25%, 50%, and 100%. The pair of red graphs show the high and low TDS reading for the puck bottom, the green pair for the puck center, and the blue pair for the puck top.

Notice that the bottom of the puck brews stronger than the full strength fresh coffee, even at 15 seconds into the shot.
-- The puck basically brews from the top down. This is to be expected, since the coffee picks up in strength as it goes down the puck and has less ability to absorb solutes at the bottom until the top layers exhaust.
-- What is surprising is that the bottom gets stronger. What is happening is that the ground coffee at the bottom is absorbing the overbrewed coffee from the top, and retaining this until the second half of the shot.
-- Blonding only occurs when this absorbed, super-concentrated coffee is released in the last 5 to 7 seconds of the shot.
-- In essence, the puck bottom is acts like a coffee battery.

The result implies a revision of the conventional wisdom, at least to me. The usual idea is that poor grinders produce lots of fines. These clog the bottom, brew too fast, and are responsible for the exaggerated flavors in the first third of the shot. However, it looks like the fines at the bottom of the puck absorb the overbrewed coffee from the top and release it slowly throughout the shot. In addition, we are just beginning to see laser analyses of grinder particles, The scant data so far does not support the idea that good grinders produce less fines. My guess is that the fines required by espresso extraction are being confused for the dust produced by whirly blades. Based on this, I'm guessing they are quite different in their size.

I find these results somewhat surprising. I've offered my take; but I would appreciate everyone elses' ideas -- this is new, and my interpretation is likely to be off base.

[hbuchtel]

A quick question- does the machine you are using have a 3-way valve? If yes, how do you think it would affect the measured TDS of the puck? (especially the top layer)

My work has been the rather thankless job of brewing and tasting partially spent pucks and measuring their TDS

Thanks!

Henry

[another_jim]

A quick question- does the machine you are using have a 3-way valve? If yes, how do you think it would affect the TDS of the puck? (especially the top layer)

Yes, it does; like all pump machines except budget home models.

You have a point; it could affect the result. The liquid coffee left behind in the puck gets dried out and leaves the solubles behind, like instant coffee. If the top is dryer than the bottom, fewer of these instant coffee solubles get deposited. However, I sit the basket down upside down before knocking out the puck, and let it sit a few minutes. I was thinking this equalized the remaining moisture; at least it seemed that way when I was handling the pucks.

[Jepy]

I wonder how different your findings would be with a falling pressure at the end, lever-like, since flavor and mouthfeel can be very different from a straight pressure machine.

[jesawdy]

The result implies a revision of the conventional wisdom, at least to me. The usual idea is that poor grinders produce lots of fines. These clog the bottom, brew too fast, and are responsible for the exaggerated flavors in the first third of the shot. However, it looks like the fines at the bottom of the puck absorb the overbrewed coffee from the top and release it slowly throughout the shot.

Jim-

Just an idea. It may not be that the fines at the bottom of the puck absorb the overbrewed coffee at all. It may be that the other (non fine or super fine) particles control the migration of fines from the top of the puck to the bottom of the puck. These other sized particles may slow the movement of fines from the top of the puck to the bottom and even entrain them for a short while. If these fines or superfines migrate from the top of the puck to the bottom over the course of the shot, your TDS will be higher in the lower puck sections when you go to do the analysis. Now, can the coffee in the lower half of the puck actually reabsorb dissolved solids, or just impede or entrain them? I dunno.

Too many fines may still be a bad thing, but the ratios of different particle sizes may have more of a bearing on how well the coffee (and grinder) perform.

Out of curiosity, what machine and grinder(s) are you using for the above experiment so far

[HB]

However, it looks like the fines at the bottom of the puck absorb the overbrewed coffee from the top and release it slowly throughout the shot.

I'm no food scientist, but perhaps the bottom section absorption was higher because the coffee was initially exposed to lower temperature water. The speed of full pressurization / length of preinfusion affects not only how quickly water progresses through the puck during the critical early seconds, but the speed of the temperature rise:



This in turn is affected by the temperature profile (a "humped" profile would close the top-to-bottom temperature gap more quickly). Many believe that a ironing board flat temperature profile is superior, but measured mid-puck, there's little or no difference between the highest hump HX and the flattest flat double boiler temperature profile. The trend in thinking is that super tight stability of temperature and pressure is a good thing, despite that the difference is immeasurable beyond the noise of the first half of the puck; your data suggests the extraction experience of the top affects the extraction experience of the bottom more dramatically than I expected. That's good news, because it's a lot easier to precisely manipulate the temperature and pressure of the top than the bottom!

[jesawdy]

I'm no food scientist, but perhaps the bottom section absorption was higher because the coffee was initially exposed to lower temperature water.

Dan, that's a very good point, the bottom of the puck sees a lower temperature AND the water that passes through the bottom of the puck is less likely to extract more solubles as well. Once the water reaches the bottom of the puck, it is less likely to extract anything because it is already somewhat saturated and the concentration gradient is less. The steeper the gradient, the greater the extraction. The concentration gradient is always steeper at the top of the puck. This results in less extraction of solids, oils, whatever.

With that being said, and if I understand what Jim has shown, neither temperature gradient nor concentration gradient explain the "supersaturation" that Jim is seeing at time zero at the bottom of the puck.

One things is for sure.... there is A LOT going on..... thermodynamics, fluid dynamics, extraction chemistry, packed fluid bed chemical processing, etc, etc. Explaining it all away is the work of a lifetime.

[another_jim]

Jes: I was using Ethiopian DP Lekempti coffee roasted just past the first pops of the second, the Versalab grinder and Elektra Semiautomatica for the experiments. The '0' time on the graph is when the first drop came out of the naked PF.

Dan: The water dispersion on the Semi is excellent, so the '0' puck was fully saturated, despite the short dwell time. This means the dropping temperature could be a factor in the supersaturation. I'm no chemist, but presumably, if the coffee is fully saturated at the top, and cools off, it will initially precipitate out solubles at the bottom. As the solution gets less saturated, and warmer at the bottom, the precipitation will cease, and eventually it'll start absorbing again.

Also, with the fines at the bottom, the pressure in most of the puck will be near 9 bar, and it'll drop fast as it passes through the fines at the bottom. I have no idea how this affects the way the solubles absorb. Is the high pressure only a device for forcing water through more finely ground coffee, or does it also effect the extraction in other ways?

Basically, we are quite close to creating a detailed simulation of the extraction. Dan has the temperature map over the course of the shot at various puck depths. I have (if the data holds) a map of the extraction patterns. Finally, once we organize the laser counter, we can get the particle distribution of the puck over the course of the shot the same way I did the extraction map. This particle map will allow allow the pressure gradients to be computed; and add detail to the extraction dynamics.

In other words, the three maps together will give a fairly complete physical model of the espresso extraction process. In theory, such a model would allow the grinder's particle distribution, the machine's temperature and pressure profile, and the puck geometry to be used as boundary conditions, so one could predict how they affect the extraction at the nitty-gritty level.

[HB]

FYI, this topic is cross-posted on alt.coffee. The responses so far have varied from "Who cares?" (Ken Fox) and "What's TDS?" (JuleG) to "Anybody got an SEM handy?" (David Lewis).

[another_jim]

Yeah, not much action.

I've had a bit of serendipity today, and want to suggest an easy experiment for people inclined to try it. Do several shots, trading off grind and dose, that is, smaller doses at finer grinds versus larger doses at coarser grinds, so you get roughly the same shot size in the same time. Report your findings. I'll start a poll thread to gather the results.

[gscace]

-- What is surprising is that the bottom gets stronger. What is happening is that the ground coffee at the bottom is absorbing the overbrewed coffee from the top, and retaining this until the second half of the shot.

Jim:

The brewing temperature within the puck is dependent on radial and axial position within the cake. The bottom definitely heats up later than the top. Is it possible that the bottom also extracts at a slower rate in the beginning because it is colder than the top? Also the pressure near the bottom is near to atmospheric pressure, unlike the pressure at the top

I'm wondering whether these factors drive extraction rate more than concentration of solutes in the water as it passes through the bottom of the cake.

Interesting work. I wonder how this relates to basket geometry. My laSpaz, with 53mm baskets produce a different taste compared to the straight sided Faema double baskets that I use in my Linea. Brew parameters are otherwise equal in temperature and pressure at the top of the cake, although the pressure rampup may be slightly different (0.6mm gicleurs in the Linea and unknown in the Spaz).

-Greg

[another_jim]

Interesting work. I wonder how this relates to basket geometry.

This and group head design is one of the things I'm working towards (the other being the grind). It seems to me if you have two espresso machines, shots from the same machine at slightly different pressures and temperatures generally taste more alike than shots from the two machines with pressures and temperatures adjusted to be the same. Machines and baskets seem to have a "style" of some sort.

In order to get at this, one needs to be able to see what's happening inside the puck. Interrupting shots and analyzing the partially spent pucks seems the obvious way of doing this. I'm also hoping to be able to organize a laser particle count/sizing on the same sequence of pucks.

The ultimate goal is to build time domain equations of the extraction for solubles with different solution rates. This will give an "extraction map" of the puck for each different tastes in an espresso (providing Lingle's Taster's wheel is right about the taste & molecular weight sequence). Pressure and flow maps can be inferred from the changing fines and coarse distribution, and we already have data on the temperature profile at various puck depths.

[cai42]

Greetings,

I know temperature affects solubility. I don't know "beans" about bean constituents and their solubility but couldn't this be one of many factors going inside the puck? Higher temps at the top of the puck and dropping off as the extraction ends.

Cliff Isackson

[King Seven]

I wonder if the movement of the fines and other smaller particles affects things?

[PaniniGuy]

Taking all this into consideration, if you were to design a basket from scratch, would it be ridgeless or ridged similar to current models or something different?

[another_jim]

if you were to design a basket from scratch, would it be ridgeless or ridged similar to current models or something different?

I certainly would go either ridgeless or with outie, rather than innie, ridges. One upshot of these tests is that one may want to use the same basket for radically different doses: large doses, coarsely ground or small doses, finely ground. A ridge almost always screws up the levelling and tamping for small doses.

I wonder if the movement of the fines and other smaller particles affects things?

Fines is what got me started on this.

There's an old alt.coffee exercise of using three cups for a shot, putting the first few seconds in the first, the next few in the second, and the final part in the third. Unsurprisingly, the third section tastes weak, and slightly bitter. It's the first two sections that are puzzling: the first section tastes very intense, both bitter and sour; while the second section tastes sweet and creamy.

I always thought that the first section's taste was marked by overextracted fines. But the Illy chapter says the fines migrate down toward the bottom of the puck, and are necessary to control the rate of flow. Now, if you have a column of ground coffee, and send water through, the coffee will brew from the top down, since the water will extract all the top coffee, get saturated, and be unable to pick up solubles further down.

Bottom line: the fines brew faster, but are further down the puck; so when do they brew, into the first or second third of the shot? Turns out, it's in the second, sweet third of the shot, and that they may have been getting a bad rap.

[hbuchtel]

I think that the 3-way valve is throwing off your measurements, can you repeat your test with the valve disabled and let the pucks 'drip-dry'? Or do the same test on your Peppina?

Henry

[another_jim]

I think that the 3-way valve is throwing off your measurements, can you repeat your test with the valve disabled and let the pucks 'drip-dry'? Or do the same test on your Peppina?

Henry

With all due respect; if you think the results are invalid, you should repeat the trials. After my initial trials, I thought about this; then some more after your first post. I see no reason to repeat the trials without a three way.

Puck concentration increases in the direction of flow. The three way pulls the liquids upwards. I immediately removed the puck, and left it upside down to drain a few minutes, keeping the liquids moving further towards the top. This will tend to move solubles towards the top of the puck and even out the readings. If you let the puck drain and dry in the PF, the liquid will move downwards, rather than upwards as in a 3 way, and the puck bottom will get even more solubles. That at least is my take. If you think my reasoning is wrong; repeat the experiment with your own protocol.

[hbuchtel]

With all due respect; if you think the results are invalid, you should repeat the trials. After my initial trials, I thought about this; then some more after your first post. I see no reason to repeat the trials without a three way.

Puck concentration increases in the direction of flow. The three way pulls the liquids upwards. I immediately removed the puck, and left it upside down to drain a few minutes, keeping the liquids moving further towards the top. This will tend to move solubles towards the top of the puck and even out the readings. If you let the puck drain and dry in the PF, the liquid will move downwards, rather than upwards as in a 3 way, and the puck bottom will get even more solubles. That at least is my take. If you think my reasoning is wrong; repeat the experiment with your own protocol.

I hear you, I'm being an a**hole about it 'cause a lot of people are going to look at your graph and say "ah, this is how it is" ...

Besides causing the liquid in the puck to flow upwards the 3-way would also 'vacuum' off a bit (how much?) of coffee liquid. It could be exaggerating the difference between the three puck levels or it could be evening them out ... Allowing the puck to dry by itself also has problems, but at least it would provide another result.

As for me repeating the test, I don't have the skills (or a TDS meter) that would make my results useful ... so I'll just settle back in my armchair but would you mind posting how you got the TDS results?

Going on the results so far, it would seem like the cone shape of a single basket should extract more evenly then a double basket.

Thanks again for doing the legwork!

Henry

[another_jim]

Sorry to sound snippy. I post data from 2 to 3 experiments a year; and always get lots of suggestions on how I can improve them, and never, ever a volunteer to do the improved version.

3 way valves don't suck anything -- the burp discretely as they vent the built up pressure. In groups that hold a lot of water between the puck and the three way, they will spit a lot of water. The Semi has almost no storage inside the group, and the pressure is equalized with a tiny run-off. On low dose shots, water remains over the puck.