Ken Fox speaking: Before going into the details of our blind tasting experiment, I'd like to give a little background on how this whole thing came about. Those not interested in this can skip down to the next heading, entitled "GOAL."
How did this "experiment" come about? What, other than the desire to waste a lot of time and money, would motivate someone to modify their relatively new plumbed in rotary Cimbali Junior in this way? These are questions I've been asking myself for the last few weeks.
About 3 years ago I decided to "upgrade" my then 7-year old vibratory pump pourover Cimbali Junior. The reason was largely a bad case of "upgradeitis" spurred by reading too many posts on alt.coffee, which at that time frequently extolled the virtues of rotary pump driven espresso machines and the supposedly superior espresso shots they produced. Not wanting to miss out on the fun, I plunked down a pile of money and became the proud owner of a Cimbali Junior D1 of current vintage. Not too long afterwards, Jim Schulman came to visit and we did our blind taste test comparison of these machines and concluded there was no systematic difference in the shots when these machines were compared head on A Tale of Two Juniors, Part Deux. Then, a couple of years ago, Jim and I modified the rotary machine by replacing the pstat temperature control with a PID, whose installation I have further modified in the interim by mounting it entirely on the machine's chassis.
After Jim and I did our comparison, the old vibe machine went down into the basement, where it lived most of the last 3 years, only getting turned on every few months, to be sure it still worked and to maintain the seals. Several months ago I decided to PID the vibe machine also, and after doing so decided I'd keep it upstairs in the kitchen just across from the rotary machine. Doesn't everyone have 2 single group commercial espresso machines in their kitchen? In continued use the vibe machine had some operational problems which I resolved by replacing the vibe pump, the group solenoid, the input solenoid, the braided high pressure hose, and the gasket in the OPV, the latter with a homemade rubber gasket made out of a piece of group gasket rubber. The latter repair was necessitated by intolerable screeching coming out of the OPV due to the worn gasket, which no Cimbali dealer I contacted stocked.
Previous posts I've made establish, with Scace Device thermography, that a high degree of temperature stability can be attained with both of my Cimbali Juniors.
I started to alternate my use of the two machines, using one for a week and then the other one for a week. I was dumbfounded to realize that whereas I regularly produced sink shots on the rotary machine, I seldom did so on the vibe pump machine. This was happening way too often on the rotary for this to be attributed to chance. I picked up the phone and called Michael Teahan in LA, to see if he had any ideas on why I was having worse luck with the rotary machine and also for suggestions on how I might fix it. Michael told me it was probably due to the difference in the ramp up of pressure in the two machines, in the fact that the slow ramp up in the vibe machine was more forgiving to the puck than was the rapid ramp up one gets with a rotary pump. He suggested that I put in a "delay on make" timer on the rotary pump, which would enable water to preinfuse the grounds for a few seconds at regulated water mains pressure, after which the rotary pump would come on and complete the extraction.
The original cube timer I obtained turned out to be incompatible with my machine, and only weeks later did I stumble onto a timer that would work, at the Mcmaster.com website. When the timer came, it had 5 wire terminals on it rather than the 2 on the original one, and I ultimately got wiring instructions that worked from Andy Schecter. Once hooked up, the timer worked perfectly and continues to do so. This particular timer can be adjusted to delay the pump from 0.5 to 10 seconds; right now it is set on 6 seconds which seems in use to be about right. The other essential part of this pre-infusion system is to have water main pressure coming into the machine regulated to a desirable pressure; this requires an adjustable inline pressure regulator, which many people with plumbed in espresso machines will have installed anyway. Previously we posted about our testing on different preinfusion water pressures and found that we needed approximately 3 bar for the incoming water to homogeneously soak the coffee grounds rather than just wetting the periphery of the puck.
Jim Schulman had a previously scheduled trip to visit me in Idaho and being as I had finally successfully installed the delay timer, it seemed only logical that we would do some sort of blind taste testing to determine if preinfusion, obtained this way, contributed anything to the cup.
Hi, Jim Schulman here. Ken already told the story of why he installed the preinfusion and how it helps with shot making. The goal of our formal testing was to search for connections between preinfusion and shot taste. We found a few.
Ideally, we would have used two identically set up rotary Juniors, one with the preinfusion delay we described previously, and one without. However, Ken's second Junior is a pour over vibe machine which does a 7.5 second ramp up pre-infuse in a style similar to E61s. So our test procedure was to taste vibe and rotary shots side by side, one day with the rotary's preinfusion turned on and the next day turned off. The vibe machine's role was to act as a source for "reference shots" against which we could measure the rotary's output. (Ken Here: Before the tests were run the brew pressure was equalized with the Over Pressure Valves, and brew temps were equalized at around 203F)
We used three coffees. Two were commercial espresso blends I'll code name "spicy" and "fruity" in accord with their dominant taste, The other was Australian Mountaintop estate, a mild, milk chocolatey shot with hints of peat. Each pair of comparison shots was made from the same blend using two Junior grinders with close to identical age and wear, each adjusted to their respective espresso machine.
Ken and I both tasted and rated shots blind. I pulled some shots for Ken, but it ended with Ken doing all the grinding and tamping, since I couldn't imitate his dosing technique and wasted a lot of coffee with me just operating the espresso machines for the shots of Ken's tasting rounds. (Operational Note from Ken: when switching between coffees and after the passage of time with changes in the beans and in the ambient humidity, the vibe machine and the modified/preinfusing rotary machine both required less finicky grind adjustments than did the unmodified rotary, if the goal was to attain equivalent extractions and less sink shots).
We rated the shots on aroma, taste, and body/mouthfeel/crema. In each category we rated the shots even or the left or right, shot either a little better or a lot better. After the results were unblinded, these ratings were transformed to a -2 to +2 scale. -2 meant the vibe machine was a lot better; +2 meant the rotary was a lot better.
The numerically motivated are all invited to analyze themselves.
1, "pr", "js", "sp", 0.00, 0.00, 0.00
2, "pr", "js", "sp", 1.00, 1.00, 0.00
3, "pr", "js", "sp", 2.00, 2.00, 1.00
4, "pr", "js", "sp", 0.00, 0.00, 0.00
5, "pr", "js", "sp", 1.00, 0.00, 0.00
6, "pr", "kf", "sp", 1.00, 0.00, 0.00
7, "pr", "kf", "sp", 0.00, 0.00, 1.00
8, "pr", "kf", "sp", 0.00, -2.00, -1.00
9, "pr", "kf", "sp", -1.00, 0.00, 0.00
10, "pr", "js", "fr", 1.00, 0.00, 0.00
11, "pr", "js", "fr", -1.00, 0.00, 0.00
12, "pr", "kf", "fr", -2.00, -1.00, 0.00
13, "pr", "kf", "fr", 0.00, -1.00, 0.00
14, "pr", "js", "au", 2.00, -2.00, 0.00
15, "pr", "js", "au", 0.00, 1.00, 0.00
16, "pr", "kf", "au", 1.00, -1.00, 0.00
17, "pr", "kf", "au", 0.00, -1.00, 0.00
18, "no", "js", "au", 2.00, 0.00, 0.00
19, "no", "js", "au", 1.00, -1.00, -1.00
20, "no", "kf", "au", 1.00, 1.00, 0.00
21, "no", "kf", "au", 0.00, 0.00, 0.00
22, "no", "js", "fr", 0.00, -2.00, -1.00
23, "no", "js", "fr", -1.00, -2.00, 0.00
24, "no", "kf", "fr", -1.00, -2.00, 0.00
25, "no", "kf", "fr", 0.00, -1.00, 0.00
26, "no", "js", "sp", 0.00, -2.00, 1.00
27, "no", "js", "sp", 1.00, 0.00, 0.00
28, "no", "js", "sp", 0.00, 0.00, 0.00
29, "no", "js", "sp", -1.00, 0.00, 0.00
30, "no", "kf", "sp", -1.00, 0.00, 0.00
31, "no", "kf", "sp", 0.00, 0.00, 0.00
32, "no", "kf", "sp", -1.00, -1.00, 0.00
33, "no", "kf", "sp", 1.00, 1.00, 0.00
It turned out that the body ratings were almost all zeros; with both machines turning out identically good shots in this respect. So the body ratings were dropped from the analysis.
In an absolute sense, both Ken and I greatly preferred the blend coded fruity to the Australian Mountaintop or the one coded spicy. This is reflected in the scores, because we were able to develop preferences to one or the other comparison shot more frequently and more strongly when using this blend.
A pictorial representation of this dataset, using aroma and taste as axis, and our initials in red and black to show the no preinfusion and preinfusion results is posted here:
It conveys a graphically honest representation of how strong the correlations are that turned up in the analysis.
One can analyze this dataset by dividing the data into groups and testing whether the average shot rating of each group is significantly different (the stats name is "paired t-tests"). Since there are multiple ways to group the data, one can more fully use the information contained in the data by building a single model that predicts the rating based on the combination of factors (preinfusion or not, Ken or me, and which coffee) in each test (the stats name is "analysis of variance").
Grouping the data turns up that I liked the preinfusion rotary produced shots better, and that Ken liked the vibe shots better, no matter how the rotary was set up. Both these differences are significant to the 5% level. The analysis of variance model also showed that our reactions to the fruity blend were much more biased towards the vibe machine than the other coffees.
The model for aroma and taste together, as run in "S" (a statistics analysis program), is printed out below:
Call: lm(formula = I(taste + aroma) ~ I(coffee == "fr") + (preno + taster)^2)
Min 1Q Median 3Q Max
-2.08 -1.009 -0.08028 0.7408 2.706
Value Std. Error t value Pr(>|t|)
(Intercept) -0.7144 0.2700 -2.6464 0.0132
I(coffee == "fr") -0.9106 0.2700 -3.3728 0.0022
preno 0.2408 0.2316 1.0397 0.3074
taster -0.3658 0.2316 -1.5794 0.1255
preno:taster -0.4908 0.2316 -2.1190 0.0431
Residual standard error: 1.329 on 28 degrees of freedom
Multiple R-Squared: 0.422
F-statistic: 5.11 on 4 and 28 degrees of freedom, the p-value is 0.003212
The models for taste alone is roughly the same, while for aroma alone, the coefficients have lower t-values with the same sign, and an F-statistic significances at around 5%. The model residuals show no violation of normality assumptions. I won't go into a detailed interpretation of the linear model which would merely repeat the analysis above in technical terms
DISCUSSION and CONCLUSIONS:
(Jim Schulman Here)
The size of the dataset precludes finding a lot of statistically significant relations. My preference for the rotary's preinfusion shows up; as does Ken's preference for the vibe's preinfusion. The strength of these effects depends on the type of blend. I doubt anyone is going to do more data collection on the topic; but my guess is that in informal tastings, people who do install the time delay preinfusions will find it does well for some coffees and may actually do more harm than good on others. In any case, I recommend people install a switch to bypass the delay so that they can do quick checks when switching blends.
It's suggestive that the fruity blend, containing a high proportion of east African DPs, responded most strongly and positively to preinfusion. If one follows the coffee and espresso tasting discussions on the web, one gets the impression that people tend to form strong opinions on whether these coffees are delightful or fermented. "Over-opulent," "flabby," and "imprecise" are some of the terms used by people insisting on tightly focused, ultra-clean blends. Oddly enough, my sense has been that the same people use the same condemnatory vocabulary in connection with preinfusion, e-61 heads, vibe pumps, all of which are supposed to take away from the pristine elegance of a real espresso shot. They may be on to something, in an odd "chacun a son gout" way. Unfortunately, I don't have enough datapoints to show whether the washed, single origin, Aussie coffee, where we both slightly preferred the unpreinfusing rotary, also supports this.
(Ken Fox Here)
What is the "take home message" of this blind tasting experiment, comparing very similar machines which differ primarily by virtue of the type of pump used to produce the shot and whether some sort of preinfusion was present in the shot production? It used to be almost dogma on alt.coffee and other coffee related websites that rotary pump driven machines produce better shots than vibe pump machines. This is now the second blind tasting study that Jim and I have done in the last 3 years that fails to confirm this supposition. There are reasons why one might prefer a rotary machine to a vibe machine, such as reduced noise or ability to plumb it in, but better shots attributed to the pump is unlikely to be a valid reason. If you have a vibe pump machine there is no reason, on the basis of pump type alone, to upgrade, if the goal is "better" shots. There are, after all, disadvantages to rotary pumps such as added complexity and expense if repairs are needed. If you have a rotary pump driven machine and it does not preinfuse the grounds, the modification I have done is fairly easy and cheap to do. The result of doing this modification should be greater tolerance to grind settings, easier switching between coffees, with less sink shots resulting. This is to say that with preinfusion added a non-preinfusing rotary machine begins to more and more resemble the way that a vibe pump driven machine operates. For ease of operation and less hassle, I'd suggest just installing the delay timer in your non-preinfusing rotary machine, without a need for an on-off switch.