How filter baskets affect espresso taste and barista technique

Behind the scenes of the site's projects and equipment reviews.
User avatar
Team HB

#1: Post by another_jim »


The introduction of the VST baskets has started a discussion of how filter basket construction affects espresso taste. This study of different baskets' construction, shot flow, and taste is designed to answer some of the questions raised.

The major findings are two: The cold taste of espresso is scale independent, i.e., a small dose in a small basket and a large dose in a large basket taste the same after the espresso has cooled, provided the same coffee, grind, shot time and ratio of shot weight to coffee weight are used. However, the taste of hot espresso is scale dependent. Espresso made from large doses taste punchier, i.e., the crema creates a more structured mouthfeel of astringent and bitter sensations while shots from smaller doses taste softer and creamier, having a smoother and more buttery tasting crema. This remains true even when the scale of the shots is held constant, as described above.

Punchy, higher dose shots are a trademark of Pacific Northwest espresso. However, conventional baskets at higher doses require coarser grinds that create a poor cold taste with more acidic, lighter roasted blends. The VST 18 gram basket has found a niche in making these higher dosed, punchy shots with the finer grind required by brighter coffees.


John Weiss determined the hole distributions and supplied the data for the first part of the study. Andy Schechter developed the concept of brew ratio, the ratio of dose weight to shot weight. Without this concept, it is impossible to compare baskets either for their shot making characteristics or for taste. Chris Tacy supplied a tasting report of conventional versus Strada baskets (clones of the VST 18 gram), and James Hoffman has challenged our ideas about crema. Without their work, the format of the third part of the study would have been impossible. Finally, without team HB's collective support, this study would not have happened.


The study is in three parts. In the first, the holes size distributions and other physical dimensions of fourteen baskets were taken and compared. In the second, the same coffee (Metropolis Red Line) at the same grind setting was pulled at six different weighed doses in each basket for 25 seconds, and the shots were weighed. The resulting graph of shot weight to dose weight was modeled, first categorically, estimating a separate straight line for each basket or each type of basket, and then physically, using a simple model of flow and plugging in the dimensions garnered from the first part of the study. In the third, five of the baskets were used to make shots at doses calculated to produce 0.67 brew ratios in 25 seconds (the par shot for this study). The tastes of the shots were compared hot and cold.
Jim Schulman

User avatar
another_jim (original poster)
Team HB

#2: Post by another_jim (original poster) »


Eight different types of baskets were used in this study; for five of the types, a pair of baskets was used, for a total of fourteen individual baskets:
  • A pair of Faema singles, and a pair of Faema doubles, shown in black on the graphs
  • A La Marzocco single and a pair of La Marzocco doubles, shown in blue on the graphs
  • A pair of MicroFine doubles, shown in green on the graphs
  • A pair of OEM triples, shown in green on the graphs
  • One VST 15 gram, 18 gram, and 22 gram baskets, shown in red on the graphs.
The baskets' bases were precisely measured and photographed. John Weiss wrote a program that analyzed the images and counted the pixels in each hole. From that and the measurement information, he determined the complete hole size distributions for each basket. This procedure yielded five variables that were significant in the physical models. The total hole area, along with the dose, were key components of the flow model. The base diameter, which shows how conical or cylindrical overall the basket is (since the top is 58mm), the grid pattern (square or hexagonal), average hole diameter, and hole diameter variance were all significant contributors to the parameters of the flow equation.

The data are as follows:
         B.dia H.av    288 396.7 37.5 28.2    h    288 382.2 25.9 27.5    h    310 367.2 28.2 37.9    s    430 285.8 39.2 36.6    h    430 286.3 48.1 36.9    h    430 342.3 30.8 59.3    s    430 366.8 38.8 68.5    s    430 401.9 50.7 81.9    s    430 409.9 46.3 85.0    s
vst.15     494 310.6 14.3 54.3    h
vst.18     494 351.2  9.8 69.3    h
vst.22     494 377.1 15.4 80.0    h    489 384.0 41.5 87.7    h    489 387.0 42.5 88.7    h
Columns: B.dia is the basket diameter, H.av is the average holes size, is the hole variance, is the total area of all the holes, is whether the grid is hexagonal (0), or square(1).

Rows: fa is Faema, lm is La Marzocco, mf is the microfine basket, is an oem triple, vst xx is the VST basket in its gram size; si and do refer to single and double

It is clear that the manufacturing tolerances of the VST baskets are about three times as tight as those of the other baskets, as shown by their standard deviations. The Micro-Fine baskets claim more precise holes, but show no exceptional increase in manufacturing tolerances. It is also true that the VST hole sizes and total area are staggered roughly in proportion to their nominal size. Therefore, VST's claims about the purely physical characteristics of their baskets are verified.

Here are the estimated frequency distributions of the three VST baskets as compared to those of the other doubles in the study.


The Faema doubles have exceptionally small holes and a small total hole area. However, the La Marzocco and Micro Fine baskets use a square hole grid which requires a larger hole area for equal flow. Therefore, all three doubles performed roughly the same.
Jim Schulman

User avatar
another_jim (original poster)
Team HB

#3: Post by another_jim (original poster) »


It is impossible to gather data for these baskets on all machines, all coffees, and all grinders. This study looked at each basket using one grinder, one coffee and one machine. The guiding assumption is that the data gathered from this will indicate how the baskets perform in general.

The procedure was to make six shots with each basket, for a total of 84 shots, using the same grind, coffee, and machine. Each shot was timed to run for 25 seconds and the resulting shot was weighed. The average dose weight for each basket was designed to produce a shot that weighed about 1.5 times the dose (brew ratio of 0.67). The range of doses was roughly plus or minus 0.5 to 1 gram from the middle dose, depending on the sensitivity of the basket to dose changes. The idea was to get a range of shots that on the ristretto end weighed as much as the dose (brew ratio of 1.0) and at the lungo end, weighed about twice as much as the dose (brew ratio of 0.5). In some cases, with baskets that responded more strongly to dose changes, this range was exceeded.

This procedure generates a good deal of variation and noise. The flow is greatest at the end of the shot, so small changes in dwell time create fairly large changes in shot weight. All the shots in this data set looked OK, and would have passed muster on an automatic machine or if the barista ended shots when they paled. However, these much more reasonable procedures, used whenever shots are made for consumption, would not have yielded the directly comparable data required for this study.

There are alternative ways for gathering the same data. Instead of varying dose, one could vary grind settings. The reason this study went with dose is that it is easier to accurately record a weight than a grind setting. Another alternative is to make each shot to the same weight and record the time it took. This requires more elaborate synchronization, and either a fast responding, accurate scale, or a highly accurate flow meter. It is therefore a more expensive and error prone way of collecting these data.

The raw data is illustrated in the graph below. The squiggly lines show each of the fourteen baskets, the colors and numbers show the nine basket types. The legend is given in the graph.


The more the lines squiggle, the less regular the response of shot weight to dose weight. It should be noted that all these shots were visually OK, and that using fixed shot times strongly accentuates small differences in flow.

The data suggests that a separate straight line shout be drawn either for each individual basket or for each kind of basket. A line for each kind of basket explains about 78% of the variation, while one for each individual basket explains about 82%. The difference is too small to be statistically significant or to matter in everyday shot making.

The model with a straight line for each kind of basket is illustrated below.


The shallower slope (wider dosing range) of the LM baskets, and the steeper slope (narrower dosing range) of the VST baskets are massively significant. The models' accuracy drops precipitously if these effects are not included.

VST claims that its baskets are more consistent in commercial operation than regular baskets. The study does not support this claim. Random shot to shot variation is large enough that the differences in conventional baskets are only a minor exacerbating factor. Moreover, the narrower dosing range of the VST baskets means that it amplifies small random errors in grind and dose. In this study, this added error proneness offset any gain from tighter manufacturing tolerances.

The VST-18 and equivalent Strada basket have become popular among third wave professionals and serious amateurs. This section and the taste section suggest that more important for this acceptance than any consistency gain is that the 18 gram VST basket fills a niche in dose-to-grind relations, roughly a 2.5 shot, that was previously unoccupied. People wanting to make 18 gram shots would have to under-dose a triple or over-dose a double, with correspondingly extra fine or extra coarse grind settings. The relevance of avoiding this to making good shots will be discussed in the taste section.
Jim Schulman

User avatar
another_jim (original poster)
Team HB

#4: Post by another_jim (original poster) »


A simplified physical model was used:
Shot Weight = Average Flow * (Shot time - Dwell Time)
The two main terms of the model are: hole.area/(dose.weight+constant.a) for the flow rate and (dose.weight + constant.b)/hole.area for the dwell time. Since the shot time is a fixed 25 seconds, expanding the equation yields hole.area/(dose.weight+constant.a) and (dose.weight + constant.b)/(dose.weight+constant.a) as the two model terms. In practice, models using different constants were equally accurate, provided their values are different, so the final model omits constant.b and used the value of 1 for constant.a.

The parameters (i.e. additional constants) of the model are based on the other physical characteristics of the baskets. A model like this cannot work as well as the categorical model since the categorical model is based on assigning the baskets whatever characteristics fit the data best, rather than actual characteristics. In practice, this physical model works quite well and explains about 76% of the shot variations as opposed to the categorical model's 78%.


A physical model can be used to test what actual basket characteristics are significant in explaining the performance of a basket. These are:
  • Total hole area: this along with dose is the basic physical variable controlling the flow.
  • Base diameter of the basket: a small base implies a more conical basket and this has the effect of slowing down the flow, even in addition to the smaller hole area a small bottom implies. It also extends the dwell time.
  • Square Grid versus Hexagonal Grid: Holes arranged in a square grid extend the dwell time.
  • Hole Size Variance (?): a smaller variance in hole sizes appeared to both slow down the flow and decrease the dwell time. However, only the VST baskets had dramatically smaller variances, so this factor may be working like a categorical variable rather than a physical one.
  • Average Hole Diameter: larger holes increase flow and reduce dwell time.
Jim Schulman

User avatar
another_jim (original poster)
Team HB

#5: Post by another_jim (original poster) »


If one brews 100 grams of water with 7 grams of coffee and 200 grams of water with 14 grams of the same coffee at the same grind for the same time; one gets the same taste in both brews. This phenomena is called scale-invariance. If the assumption of scale invariance holds for espresso, then one can make identical tasting shots from different baskets by selecting doses that result in the same brew ratio in the same time from the same grind of the same coffee.

On the other hand, if some baskets are fundamentally different from others, then scale invariance will not hold.

These are the alternatives this taste test was designed to test.

To do this, five baskets of various dose sizes were selected, the La Marzocco single, the Faema double, the VST 18 and 22, and the OEM triple. Each basket was used in shots dosed to to produce 0.67 brew ratios in 25 seconds, thereby producing scale invariant shots. If scale invariance holds for taste, these shots should taste identical.


Coffee cuppers and those who compete in cupping competitions know it is easiest to distinguish different coffees after the cups have cooled. But surprisingly, the five scale invariant shots were indistinguishable cold and easily distinguished when hot.

The higher the dose, the punchier tasting the shot. This punch is a combination of astringent, cutting and bitter flavors that wine lovers get from tannic wines, and which they call "structure." The higher the dose, the more structured the shot; the lower the dose, the creamier and softer tasting the shot. Since this effect disappears as the crema dissipates; this structure difference is clearly caused by differences in the crema of high and low dose shots.

Italians like soft and creamy espresso; but Seattle and West Coast espresso typically has a lot more punch. Five years ago, West Coast blends were medium roasts and had low acidity; and one could get away with overdosing a double basket and grinding coarse. With the advent of more acidic and brighter SOs and blends, the choice with a conventional double is to grind finer, dose lower, and give up the punch; or to make a very rough shot with coarser grinds and higher doses. The VST 18 gram baskets uses the same grind as a conventional double at 15 grams. This means it delivers the high dose structure without the unacceptably rough and sour taste of under-extracted, lightly roasted coffee. This may explain why some people strongly prefer the VST18 or Strada baskets.

Finally, the VST 22 and OEM triple are very similar in dose requirement, and I compared them to check whether claims that better finished baskets taste better, or at least different, was supportable. This pair of baskets delivered the punchiest shots. So much so that I found them unpleasant and hard to distinguish. Therefore, I was unable to judge whether the hot taste of these baskets differed.
Jim Schulman

User avatar
another_jim (original poster)
Team HB

#6: Post by another_jim (original poster) »


The surprise in this research is the finding that espresso is not scale invariant; but that crema, mouth feel and the hot taste of the shot is punchier and more structured at high doses, regardless of grind settings or other extraction relevant variables.

The espresso machine itself is obviously not scalable; and the exact organoleptic characteristics of the crema and mouthfeel of espresso shots appears to also depend on its non-scalable group dimensions, not just on the scalable coffee, grind and shot factors. The results here suggest that the role of these machine characteristics should not be sought in the extraction and chemical composition of the espresso coffee; but in the physical structure of the crema.

A final word: The non-scale invariance of espresso taste creates a system of taste niches for different kinds of baskets. This reality would not have been suspected had the VST baskets not been created. While the discovery has nothing to do with the excellent manufacturing quality of the baskets, it was due to the willingness of VST to create baskets that depart from Italian grind, dose and flow standards to better deliver 3rd wave style shots. They are to be complimented for doing this.
Jim Schulman


#7: Post by earlgrey_44 »

Well, some insight finally after all the basket sturm und drang. Thanks to the team for this great effort.

I think I understand my own experience with dose variation now better than before, and I'm motivated to explore the VST "18 gram flavor space"! :)
Trust your taste. Don't trust your perception.

User avatar

#8: Post by boar_d_laze »

Nicely done. Your conclusions do a fair job of explanation.

The VST 18 gram baskets uses the same grind as a conventional double at 15 grams. This means it delivers the high dose structure without the unacceptably rough and sour taste of under-extracted, lightly roasted coffee. This may explain why some people strongly prefer the VST18 or Strada baskets.
That really resonates for me -- especially at my current dose which is very close to 17g (timed grinder, subsequently weighed). 17g is also the rough dose of my other double baskets, overfilled, excess pushed off and out, and leveled by a finger swipe.

At this dose, at least in my Casa, the additional head space provided by 18g Stradas, as compared to the other LM, LC and Pasquini double baskets is most probably also significant. I haven't compared the baskets with lower (aka normal Euro) doses. Perhaps your next go-round?

FWIW, I can't make any of my triples (Synesso, VST, Strada) work nearly as well as the 18g Stradas.

Drop a nickel in the pot Joe. Takin' it slow. Waiter, waiter, percolator

User avatar

#9: Post by JonR10 »

Nicely done, and much appreciated.
I'd like to re-read before commenting on content, as I've only had time for a quick read so far.

But I'd like to express my thanks and admiration to the HB team for making this study.

Y'all are a rare group of nutjobs, and I am extremely thankful to know you and even more thankful that the home espresso enthusiast community is lucky enough to have folks like you studying stuff like this 8)
Jon Rosenthal
Houston, Texas

User avatar
another_jim (original poster)
Team HB

#10: Post by another_jim (original poster) »

boar_d_laze wrote:I haven't compared the baskets with lower (aka normal Euro) doses. Perhaps your next go-round?
In the fourth posts's graph (taste test post), the point where the 0.67 brew ratio line intersects the dose/shot curve for each basket shows equivalent (scale invariant) doses for the grind setting of this study (roughly the classic Italian 7 gram single /14 gram double grind): it is 14 to 15 grams in conventional double baskets, 17 in the VST/Strada.

Let's hope there is no need for another go around.
Jim Schulman