some aspects of espresso extraction - 5

conclusion

none of these complications should be allowed to obscure the basic point : the taste of espresso varies by the level of extraction, and that level can be controlled.

and this result points to the irony of unintended consequences.

in italy, espresso is a mass consumption item, mostly made from coffees of the same low quality as is found in supermarkets everywhere. since the aromas of such coffees are not all that great, staling is of little consequence, while keeping doses precise and yields high is of great consequence, since one needs to extract every iota of caramel to make the shots palatable. so the ground coffee sits in dosers going stale, but is precisely dosed, 6.5 grams into single baskets, 13 into doubles.

in the non-mediterranean world, espresso is specialty coffee ( 11 ). cafe owners rightly noticed that the ground coffee was going stale in the dosers, and went to alternative dosing methods. what they didn't notice is that dosing by leveling the freshly ground coffee to the basket's rim, a dose far higher than the italian norm, gets solubles yields of 16% to 20%, rather than the 20% to 24% one gets with a properly adjusted doser. "specialty espresso" was almost always under-extracted

at these low extraction levels, high grade coffees become a bane rather than a boon, producing jarringly acidic or sharp shots. so, in the specialty coffee world, there is a feverish search on for ultra-sugary high grown coffees that are still sweet when roasted light and under-extracted. when such coffees are not available, one gets the ubiquitous medium-dark roasted blends that are a far cry from the quality of the specialty coffees sold for regular brewing.

and all this because of an unintended consequence of using fresh coffee. ı think it's high time for baristas to relearn their dosing.

as a final note : since posting early drafts of this paper, i have found out that this is changing already. ın scandinavia and australia, many top competing baristas are replacing their fingers with curved swipers that scoop out ground coffee below the rim level of the basket. by having a french-curve like set of these, they can efficiently vary the dose in a workplace or competition context. i'm sure these "3rd wave" dosing tools will become much more prevalent and developed as word on working the solubles yields gets out.

to watch it, here is link : https://www.youtube.com/watch?v=yKQRdexIpdw

scottie callaghan dosing tools explained ( as pdf document ).

notes

1.  for a more detailed explanation, consult ted lingle, "the basics of brewing coffee" , 1996, scaa.

2. the classification used here differs from the one in ted lingle, "the coffee cupper's handbook", 2001, scaa. he divides flavors into three main groups, enzymatic, sugars browning, and dry distillates. however, the herby and nutty sub groups contain flavors that derive mostly from maillard reactions between sugars and amino acids. these are the flavors that typify toasted grains, malts, smoked meats and barbecued foods.

i believe they deserve their own overall classification for three reasons : they are associated with sharp or bright-bitter tastes, rather than the sour or sweet tastes of the neighboring groups; they have similarly fast solution rates; and they are mostly produced in the part of the roast running from 300˚f ( 148.9˚c ) to the first crack, and hence can be roast-profiled as a group.

3.  aroma slightly improved at high extractions, but the effect is only marginally significant ( 0.95 ) : 

mouthfeel and crema are, as expected, well related to the concentration of the shot. here are the relevant partial regressions :

4. the residual correlation of sweetness to yield, after removing all other factors, was 56%, with a t-value of 3.12.

5.  the residual correlation of acid/bitter to yield, after removing all other factors, was 73% with a t-value of 4.82.

6. the t-value for the simple linear regression of filter.area/dose to yield is 7.89; if one uses the best predictor: filter.area/dose*log(shot.time*shot.weight) the t-value rises to 8.69. the graph is shown with the predictor inverted to the more sensible dose/filter.area, so the regression line is transformed to a hyperbolic curve. with t-values this huge, the magnitude of the change in the correlation coefficients from 75.2% to 78.3% has a high degree of certainty, and shows that while shot time and weight do expalin some of the yield changes, they definitely don't explain a great deal.

the small effect of shot time and weight/volume is probably an artefact of cutting shots when the flow blondes. this cuts fast flowing shots short, and lengthens the slow flowing ones. this practice counteracts the physically mandated higher extraction rates of fast flowing shots. however, since cutting a shot as it blondes is proper barista technique, the result applies to actual shot making.

7. dosing changes can also affect the shot because higher doses can come into contact with the shower screen, while lower doses do not. this depends also on the height of the basket. m. petracco warns against puck contact with the shower screen in chapter 7 of illy and vianni", eds, "espresso coffee: the science of quality", 2005, elsevier."

...the host inclines to overdose to serve the guest the best possible cup. this practice is risky ... because an excessive amount of ground coffee does not permit sufficient expansion during cake wetting.

on my elektra, the shower screen is like a third rail, so all the data in this paper are from shots with head space. ın other groups, shots seem to survive compression by the shower screen without harm.

8. i could not find the original alt.coffee posts, this early post is pretty representative.

9. the 12 second reading only contains one observation, since ı messed up the other one.

10. chapter 7 in illy, as cited in note 7 is "a discussion of espresso percolation". it focuses mainly on how the puck dynamically affects the flow, and helped me appreciate the important role of the grinds absorbing liquid. however, it does not go into the timing of solids extraction. "discussions of percolation for instant coffee", pp 127-128, clarke and vitzthum eds, "coffee: recent developments", 2001, blackwell science, are frightening; but they emphasize the role of how the initial percolation column gets wet (top down versus before starting the percolation).

11. historian jonathan morris in his inaugural lecture given at the university of hertfordshire, november, 2005, the cappucino conquests, tells the story from a british perspective.

data sets used in this study

taste versus solubles yield data

   f.loss puck.b puck.a p.loss sh.wt sh.time aroma ac.bi dr.sw body crema roast

1   0.962   11.8    8.7  0.234  22.7    28.0  8.50  4.00  6.50 7.00  8.50     1

2   0.962   14.8   11.5  0.192  23.2    35.5  7.50  3.75  6.75 7.25  8.50     1

3   0.962   17.9   14.2  0.175  21.9    37.5  8.00  3.25  5.25 7.00  8.75     1

4   0.962   20.7   16.3  0.181  30.8    22.0  8.00  3.00  3.75 7.25  7.50     1

5   0.976   12.0    8.8  0.249  18.4    28.2  8.00  4.50  7.00 7.50  8.50     1

6   0.976   14.3   11.3  0.190  21.0    35.0  8.50  3.75  5.75 7.00  8.50     1

7   0.976   16.9   13.2  0.200  26.7    28.4  8.00  3.75  4.75 7.50  8.00     1

8   0.976   19.6   15.9  0.169  30.5    23.2  8.00  4.00  5.50 7.00  7.50     1

9   0.975   12.0    9.0  0.231  26.2    29.4  8.50  4.50  7.00 7.00  8.50     1

10  0.975   14.1   10.9  0.207  26.4    31.2  8.00  4.25  6.50 7.50  8.50     1

11  0.975   16.6   12.9  0.203  31.5    41.2  7.50  3.75  5.75 7.50  8.00     1

12  0.975   19.5   15.6  0.179  30.5    36.4  7.50  3.00  5.00 8.00  7.50     1

13  0.975   12.0    8.9  0.239  25.3    30.0  8.50  3.75  6.50 7.00  8.50     1

14  0.975   14.1   11.1  0.193  23.3    30.0  9.00  3.50  7.00 7.00  8.50     1

15  0.975   16.6   12.8  0.209  47.7    30.0  8.50  3.75  5.00 6.50  7.00     1

17  0.981   12.0    8.9  0.244  22.2    58.0  7.50  6.50  7.00 7.00  8.50     3

18  0.981   14.4   11.3  0.200  25.0    53.0  7.00  6.00  6.50 7.50  8.50     3

19  0.981   17.0   13.6  0.185  24.0    42.0  6.00  5.75  5.75 6.50  8.00     3

20  0.981   19.5   15.4  0.195  38.4    30.0  6.00  5.75  7.00 6.00  7.00     3

21  0.981   12.0    8.6  0.252  25.8    36.0  8.00  7.00  6.50 7.50  8.25     3

22  0.981   14.5   12.0  0.194  19.7    51.0  6.50  4.50  5.50 8.00  8.50     3

23  0.981   17.0   13.7  0.204  33.0    46.0  8.50  5.25  5.25 7.00  8.00     3

24  0.981   19.5   15.5  0.173  22.9    32.0  6.75  5.00  5.75 7.50  7.50     3

solubles yıeld to shot varıable data

   dose yield sh.wt sh.time filt.d roast

1  10.5 0.174   9.1    45.0    2.9     2

2  10.3 0.168  14.1    59.0    2.9     2

3   8.4 0.186  22.9    25.0    2.9     2

4   8.0 0.171  12.0    28.0    2.9     2

5   8.3 0.189  22.0    36.0    2.9     2

6  17.6 0.194  17.8    39.0    4.3     2

7  18.5 0.189  18.0    32.0    4.3     2

8  19.5 0.205  20.1    55.0    4.3     2

9  14.8 0.214  28.4    43.0    4.3     2

10 20.4 0.190  17.5    52.0    4.3     2

11 17.0 0.196  17.8    36.0    4.3     2

12  9.9 0.155   8.1    24.0    2.9     2

13  9.8 0.157  11.5    35.0    2.9     2

14 21.4 0.180  18.9    26.0    4.9     2

15 18.2 0.181  16.2    43.0    4.3     2

16 13.7 0.203  31.3    26.0    4.3     2

17 18.3 0.197  23.3    32.0    4.3     2

18 23.5 0.175  21.3    65.0    4.9     2

19 17.2 0.187  15.7    60.0    4.3     2

20 16.7 0.206  26.4    31.0    4.3     2

21 10.2 0.180  16.7    26.0    2.9     2

22 17.8 0.160  15.7    38.0    4.3     2

23 10.9 0.132   8.7    36.0    2.9     2

24 10.0 0.156  15.1    26.0    2.9     2

25 18.2 0.178  20.7    47.0    4.3     2

26 11.8 0.234  22.7    28.0    4.9     1

27 14.8 0.192  23.2    35.5    4.9     1

28 17.9 0.175  21.9    37.5    4.9     1

29 20.7 0.181  30.8    22.0    4.9     1

30 12.0 0.249  18.4    28.2    4.9     1

31 14.3 0.190  21.0    35.0    4.9     1

32 16.9 0.200  26.7    28.4    4.9     1

33 19.6 0.169  30.5    23.2    4.9     1

34 12.0 0.231  26.2    29.4    4.9     1

35 14.1 0.207  26.4    31.2    4.9     1

36 16.6 0.203  31.5    41.2    4.9     1

37 19.5 0.179  30.5    36.4    4.9     1

38 12.0 0.239  25.3    30.0    4.9     1

39 14.1 0.193  23.3    30.0    4.9     1

40 19.5 0.127  18.0    30.0    4.9     1

41 12.0 0.244  22.2    58.0    4.9     3

42 14.4 0.200  25.0    53.0    4.9     3

43 17.0 0.185  24.0    42.0    4.9     3

44 19.5 0.195  38.4    30.0    4.9     3

45 12.0 0.252  25.8    36.0    4.9     3

46 14.5 0.194  19.7    51.0    4.9     3

47 17.0 0.204  33.0    46.0    4.9     3

48 19.5 0.173  22.9    32.0    4.9     3

intra-shot tds data

1st set

     time tds.top tds.mid tds.bot str.top str.mid str.bot

[1,]    0     663    1112    1424      65      90     120

[2,]    6     650     798    1250      60      60     110

[3,]   12     440     815    1235      40      50     110

[4,]   18     478     571    1041      45      45      75

[5,]   24     409     512    1004      40      45      75

[6,]   30     321     357     690      25      30      55

2nd set

     time tds.top tds.mid tds.bot str.top str.mid str.bot

[1,]  0.0     854    1248    1676      75     100     120

[2,]  6.0     542     971    1595      40      70     100

[3,] 13.5     578     781    1180      45      55      85

[4,] 21.0     373     643    1113      25      50      85

[5,] 25.5     406     674    1058      35      50      85

[6,] 30.0     292     487     709      20      35      60

writen by jim schulman