Variable Pressure Infusion Modification Results: a Paper

Beginner and pro baristas share tips and tricks for making espresso.
User avatar
Team HB

#1: Post by dominico »

Variable Pressure Infusion Modification Results: Joint project between Mark Barnett, Synesso, and Herkimer Coffee

Below is a document I am sharing, with permission from Synesso, in very raw format.
I came across this document a while ago in a coffee shop near my office (Valentine Coffee Roasters) which runs with a Synesso Hydra, they were using this document as a guide to tweak their pressure profiles. I took the opportunity to guerilla snap photos of all the pages for future reading and experimentation, hence the fantastic quality of the photos below.

I later got permission from Synesso to share the document in a public space since I felt it would be interesting and possibly of use to home baristas who like to tinker with these sorts of things. This is one of the reference documents I used while playing experimenting with Pressure Profiling Techniques for Spring Levers. I found it quite nifty that a document initially about their attempt to make a modern pump machine mimic a spring lever taught me how to be more versatile with my spring lever.

I also found the 6 criteria under which they base their shot quality and their description of how each of them affects the outcome of the espresso very interesting in its own right, so even if the pressure profiling aspect of the document isn't of interest to you you may find some interest in how they describe the different espresso shot criteria and how they are affected.

Interspersed with the pages are some of my observations on content within them as well as any notes that I feel may make certain sections easier to understand.

Synesso also offered to send me a better quality version of the document, but they haven't yet; no problem, they are quite busy so I've decided not to nag them about it.
The first thing you will probably notice is that they list all their pressures in the document as psi. I'll convert them to bar for easy reference:
Control Phases
Initial (Stage 1) Preinfusion: (45psi) ~3 bar
Secondary (Stage 2) Preinfusion: (90psi) ~6 bar
Full Pressure Infusion: (125psi) ~8.6 bar
Dropback Infusion: (90psi) ~6 bar

To note in the paragraph that starts with "Tiny bit of background info" they state that they focused on mimicking HX levers and ignored dipper levers due to low (~1 bar) initial preinfusion and "inconsistent nature of direct water flow from the boiler"
So essentially they had in mind something like a Brugnetti Aurora rather than my Faema dipper.
I was undeterred, mainly because I don't understand what they meant by "inconsistent nature", and also because I'm less convinced than they are that a 1 bar preinfusion has much of an effect on the coffee unless it's preinfusing for 15 or more seconds.

The "Sweet Spot" in bar pressure
Initial Infusion Pressure of 3 bar. Duration of 4 seconds.
Secondary Infusion of Pressure of 6 bar. Duration of 3 seconds.
Full Pressure Infusion of ~8.6 bar. Duration varies from 8 to 11 seconds.
Dropback Infusion Pressure of 6 bar. Duration varies from 3 to 5 seconds.

As preinfusion progresses acidity moves from piquant and nippy to sweet to metallic and acrid
Lower pressures cause this progression to be slower (at the cost of sugar development), higher pressures cause this progression to go faster.
Acid at 45psi (~3bar) preinfusion:
0-2 seconds: nippy or piquant and faint
2-4 seconds: from piquant to sweet
4-6 seconds: from sweet to metallic or acrid

An interesting observation with sugar is that the full pressure portion of the shot should be shortened as coffee ages or on more humid days to bring out more sugar.

An interesting piece of info I gleaned from the "Sugar" section is that a declining pressure profile can lend itself to sweeter shots by allowing more time for the sugars to develop while inhibiting phenol development. A 8-9 second full pressure phase is desirable according to the sugar section.

Here it is said that the "drop back" pressure should only last 4 to 5 seconds before the negative effects of phenols kick in. It is important to note that a "drop-back" is only an approximation of what a spring lever does. The spring steadily ramps down the pressure, a "drop back" drops immediately back to whatever pressure the stage 2 preinfusion was set at. Thus, a Synesso doing 9 seconds of full pressure then 4 seconds of drop back is not exactly the same as a lever which really only hits full pressure for a few seconds before starting its decline steadily back down. In my opinion this would cause the desired timeframe for these equivalent stages on the lever to be a bit longer.

Spice (and everything nice)
The spice section talks a fair bit about the "spice" perception being largely affected by burr geometry / size. I imagine that all flavor attributes are affected to some degree by this; I guess they felt the effect was most pronounced with spice perception.

Essentially it was the Texture and Viscosity sections which caused me to experiment with a version of the "2 stage" preinfusion with the lever, looking for a way to increase body with lighter coffees. I've had enough success with it that I still use it semi-frequently with very light coffees, especially the finicky ones.

The following artisanally rendered graphs show the "Quality" scores of each of the depicted attributes as a function of the amount of time the extraction is in that stage. They determined the "sweet spot" listed above basically by finding where all the peaks correlated on each of these graphs.
Il caffè è un piacere, se non è buono che piacere è?
★ Helpful

Supporter ♡

#2: Post by samuellaw178 »


Thanks for sharing! This is a gem - the most extensive documentation of pressure profiling effect on taste I've seen!

One thing seems odd though - by adding up the numbers, it seems that they're pulling rather fast shots from 18-23 sec? Unfortunately the brew ratio wasn't mentioned. :P I normally pull shots between 35s-45s so this is quite interesting. Similar to you (thanks for introducing that!), I've found the stepped pressure (secondary infusion as they call) makes a positive difference to my shots. It's just I felt it's an extra step for potential inconsistency so I don't normally do it.

There were also quite a few insights/observations there along the way in that document (flat burrs vs conical burrs, drips during preinfusion, difference between low & high preinfusion pressure etc etc - things that I have no clue about). Great stuff!

User avatar

#3: Post by tohenk2 »

Thanks for sharing! Very interesting.
samuellaw178 wrote:One thing seems odd though - by adding up the numbers, it seems that they're pulling rather fast shots from 18-23 sec?
Maybe a coincidence, but my shots also run around 23 seconds If I use a profile like this. Depending on the coffee, it can get to 30 seconds.

User avatar

#4: Post by tohenk2 »

tohenk2 wrote:Maybe a coincidence, but my shots also run around 23 seconds If I use a profile like this. Depending on the coffee, it can get to 30 seconds.
I sometimes fiddle around :D

Supporter ♡

#5: Post by samuellaw178 »

tohenk2 wrote: Maybe a coincidence, but my shots also run around 23 seconds If I use a profile like this. Depending on the coffee, it can get to 30 seconds.
Interesting indeed. Is the ramp up time to target preinfusion pressure included in the 23 sec?

User avatar

#6: Post by tohenk2 replying to samuellaw178 »

Yes, it is.

I can flip the E61 lever so that it preinfuses without pump, but I rarely use that. Otherwise the timer starts when the pump gets going. (However, I do use this to clean afterwards because I don't want to count cleaning as a shot on the shot-counter.)


#7: Post by Ellejaycafe »

Dominick!!! Post like this keep me coming back to HB. THANK YOU. I know how much work a post like this takes and your passion for sharing knowledge keeps me motivated. You might not know it, but you've held my hand through my first 2 months of owning a commercial spring lever with your insightful post.

That aside, I find this profile to be the most helpful with certain coffees and the most distinguished. When I first got my machine I found the standard profile to make light roasted acidic forward coffees just too sour for my taste. So I read through your pressure profiling for spring levers post (one of my favorite HB post) and spent the past month exploring. By doing this I not only learned how to get the best out of my coffee but it also taught me a great deal about my machine.

Now with about 80% of the coffees I've tried the "synesso" profile and it seems to work best. Just pulling a shot this morning of the Ethiopian Idido (one of the lightest roast my roaster does) roasted last Monday I did the following profile.

7sec preinfuse
Lifted the lever till it caught and it held it there for 15 sec. for the second stage.
Then slowly ramping up to full pressure before letting the lever take over (I find this part is becoming an important aspect while doing this. If I just let the lever go right after stage 2 the flow is fast and generally gushes).

The shot this morning took 50 seconds till I pulled it away; another thing I'm learning with the lever is to not worry about the time as much and concentration on pressure and flow. (17g in EPHQ double and 36 out).

Doing this profile just brings out the absolute best this coffee has to offer. It's super bright, lingering blueberry, and still has some nuttiness in the background when it cools. It's one of the most balanced shots I've ever had. Without the pressure profiling the shot tends to turn sour really easy and the separation of flavors just isn't there. If I pull this coffee on my Aurelia it's definitely drinkable but not quite as interesting or balanced as with my spring lever doing the synesso style pressure profiling.

I can see why synesso has spent a great deal on trying to replicate the lever style shot.

FWIW I also have a dipper and don't get what they say about the consistency. A couple weeks ago I attached the OE temp strips to my groups and ever since I can get to whatever temp shot I want.
LMWDP #544

User avatar
Supporter ♡

#8: Post by vberch »

Great information, Dominick!!!

User avatar
Team HB

#9: Post by homeburrero replying to vberch »

+1. Thanks, Dominick, for that excellent post, and thanks to Synesso for allowing you to share their doc.

One thing in their "Tiny bit of background info" has me confused. They state the following:
"The spring's potential takes a small amount of time to be realized as it is not as strong initially as a rotary pump."

I would think it would be the other way around - i.e. that the brew chamber pressure on a spring lever is near instantly at the maximum when you release the lever. Whereas on a rotary the pressure at the gauge (which is on the pump side of the gicleur) would quickly hit the pump's max, but inside the brew chamber it would rise a little more gradually because of the gicleur (plus also because of the pre-infusion chamber in the special case of an E61.) Am I missing something?

Of course, as Dominick points out in his other thread, the ~6 bar "secondary preinfusion" effect can be accomplished on a spring lever if the operator slowly releases the lever and retards its upward progress for a few seconds before fully letting go.
nínádiishʼnahgo gohwééh náshdlį́į́h

User avatar
Team HB

#10: Post by another_jim »

dominico wrote: ... Below is a document I am sharing, with permission from Synesso, in very raw format

It confirms the basic pressure profiling result that the classic spring lever profile is "about right" when it comes to stable and forgiving espresso prep.

But to get nit-picky on the details: It's interesting to see them distinguishing the effect of small profile changes in a sequence of four shot segments. Each segment has three possibilities, so there are 81 different profiles (3**4). That comes to 6480 (81*80) possible paired blind comparisons if all possible variations are considered. This is an impossible experimental design, obviously. My guess is that the data are derived from doing blind tests for variations on each segment while holding the others at medium values (this reduces the number of blind comparisons to 12). But this assumes the contributions to the taste of each shot segment is independent of the others.

The independence assumption leads to very specific predictions. For instance, if I took two randomly selected shot profiles, I could look up the graphs and predict which of the two shots had more sugar, which one more acid, which one more viscosity, etc. Now, if I taught a panel to rate a pair of blind shots for more/less on each of these factors; would their results accord with the prediction? If one makes these claims, this kind of testing should bee done.
Jim Schulman