“Study hard what interests you the most in the most undisciplined, irreverent and original manner possible.”
— Richard Feynman, November 30, 1965
At the time I had absolutely no idea I was looking at a representation of sheer-jamming in two-dimensional granular materials with power-law grain-size distribution. At all. To me this was obviously a portafilter, those big circles were boulders, the small ones fines, and the picture was showing me the school of tamping towards maximum density was rather flawed. “Hopefully that is interesting?” Christopher Hendon had asked me, and after staring at it for a good day or two, I thought I saw what was so interesting. But to be honest it was really Hendon who told me what was interesting, so I’ll just quote him: “what matters is the difference between particle A and particle B, and then what matters is not how hard you push, rather the forces at their interfaces. And as you can see, this is inhomogenous.”
Computational representations of ice seem pretty far removed from coffee, but if you replace jamming with tamping and ice with coffee, you realise quickly the physical interactions of jamming and tamping, ice and coffee, are the same. The x and y axis on the picture are spatial co-ordinates, so it was easy for me to see that when force was applied it was distributed towards the centre. The more force, the less evenly the material was distributed.
But I was still well away from understanding the idea of homogeneity or what forces applied at interfaces meant just yet. I needed data to help tell me the story, so it was time to do some science, retire some filters, and refract.
Ask a Good Question
So Hendon may be a computational chemist with a pretty good by-line in old-school platinum chemistry and electrocatalysis, but he’s also provided some damn good presentations on coffee. In his presentation “Performing a Good Experiment” Hendon says that to be a good experiment, you need to ask a good question. I thought I had a damn good question to ask: can you make better coffee with higher extraction yields and less variance at low pump pressure, compared to 9 bars? So I lowered my pump pressure to 6 bars, and started taking samples. Hendon had already told me that I’d get a higher extraction and less variance if I did this, but I still think he was a little surprised I took him at his word and went for it. He was right though – my extraction percentage went up, and my variance went down.
How much? Over 20 samples (n) at 9 bars I had an average (M) extraction yield (EY) of 19.08%, with a standard deviation (SD) from the mean of 0.72%. At 6 bars this was M = 19.58, SD = 0.33, n = 20. This was a statistically significant difference (p = .018) with my variance dropping by over half. And taste? Never been this good. This was at a 1:1.55 ratio, as this was how we’d always pulled our shots, and I had never reached great tasting extraction yields at a more progressive 1:2 ratio. I had a strong feeling though I could raise both EY and flavour by stretching those shots out at low pressure, but I was also mindful of another part of Hendon’s “Performing a Good Experiment” presentation: determine your dependent and independent variables.
Know Your Variables
So what were my variables? Dose, time, and yield seemed obvious, with dose being set at 20g for a 20g VST basket. Time and yield were triangulated to achieve at a 1:2 ratio of 40g in 30 seconds – straight out of 9 bar theory and one I had never really questioned. All my initial testing was set up to achieve these metrics.
There was also brew temperature, or the temperature set at the boiler for our La Marzocco GB5, and that was set to 93˚C – again, never questioned, that’s just where you set the temperature at 9 bars. Sometimes higher at 94˚C – I’d even worked with roasters who demanded 96˚C set on the machine – but ever since starting with the Mythos it made sense to me to lower brew temperature slightly to compensate for the heated grinds.
Roast age was important to me as far as degassing goes – so I made sure I always used beans 10 days post-roast leaving adequate time to de-gas, and before flavour fell off a cliff after day 14. I had no Scace device at the time, so I could only ensure at 9 bars the dial read “9.5” with three groups engaged, and down low read “6” with three groups engaged. Before each morning of testing I’d set the volumetrics for my ratio using scales, each group individually set. For initial testing I used a Pullman 58.5mm tamp, and my entirely scientific manner for determining tamp pressure was to place my thumb and forefinger halfway down the outside of the side of the tamp, and stop when the tips of my finger hit the top edge of the basket. My method of grinds distribution, including collapsing, I laid out in this post here. Testing with the refractometer followed Matt Perger’s technique here, with a few more added alcohol wipes in between and plenty of sample stirring – oh, and a small fortune in VST filters (thanks Roger!).
I’d ride out the first hour as the grinder warmed up and service got busy, then I’d randomly start pulling shots off the bar. I decided I’d have a +/- 0.1g tolerance with 20.1g my target dose, as I wanted to mimic real-world conditions – also why I was randomly pulling shots during service. I had no desire to create “laboratory” conditions as I wanted my results to be indicative of how I ran my shop, not how I ran a lab. You could consider that a weakness or a strength, but I was never intending to have these experiments submitted to a journal for peer review – I was doing this to help me make better coffee. In my mind internal validity – or the ability to compare my results between my experiments – was an important factor to maintain.
With that in mind each evening before testing I would tear apart the Mythos, wash the burrs with soap and Isopropyl, then align them as best I could. I also wrote how to do that here, and if you’re a Barista Hustle Facebook member (you should be!) have a look at Mat North’s amazing pictographic post for Mythos alignment. And go get to whale hunting …
I usually cleaned my burrs every fortnight, after around 100 – 120kg coffee. Mat, on the other hand, cleaned his every 8kgs. Consider my results then to be at one end of a spectrum. For comparison sake however, during one lot of testing before cleaning the burrs I had an EY of M = 20.24, SD = 0.29, n = 10, and after cleaning M = 20.75, SD = 0.16, n = 10. I was running tests every week at this point, so this was five working days difference – half a percentage in extraction yield added from cleaning the burrs with soap and aligning. I can’t tell if it was just cleaning that added that half percent, or just alignment, or both together – unfortunately there was no way I was going to be at work for the eleventh hour pulling apart the Mythos and not make it worth my time. If the burrs came off they were cleaned and aligned. I have other tests where I didn’t use filters, and they showed the same trend – at least half a percent difference, sometimes a full percent, in EY pre- and post-clean and alignment. What those results didn’t show however was the difference in clarity. My coffee simply tasted better after cleaning and aligning.
Did you notice the variance as well? 0.16? I got that down to 0.02 for one test the week after this lot, and after that it stayed well under half a percent – always. And I have the Puqpress to thank for that.
Death of the Barista – First Blood
First blood? Perhaps that was when we decided to weigh our shots, or place temperature stickers on our milk jugs, or use volumetrics to measure our shots, then used scales to do the same. Or was it when we placed timers on our grinders, and soon enough scales will be there too. We’re slowly but surely abstracting away user error in espresso preparation. The Puqpress was always coming, perhaps the Eazytamp lead the way. I remember first coming here to Australia eight years ago and starting work in a small cafe with a lever-like contraption for tamping coffee. And somehow we still have around the Swift grinder with its 2g variance dose to dose. Automation is inevitable, it’s a natural process of evolution, and the Puqpress is part of that evolutionary step.
And I’m happy for it. Because coming home, cooking dinner for my wife, and my hands aching while I cut potatoes is a shit thing to have to deal with. Wringing dry a cloth and pangs of pain shooting up my wrist is a shit thing to have to deal with. Tamping that little bit hard by accident and the pain seizing up my arm for a moment is a shit thing to have to deal with. Going through 10kgs of coffee in four hours while trying to maintain a consistent tamp pressure and high level of energy and concentration, is a shit thing to have to deal with. RSI straight-up sucks and I no longer have to deal with it. Because I have a Puqpress. And it’s glorious. Can’t afford AU$1600? Get an Eazytamp 5 Star Pro. Your wrists will thank you.
The Puqpress comes with a stock standard 58.3mm tamp, but I asked for a 58.5mm tamp – I wanted to make sure I could maintain my internal validity with the Pullman tamper. It has an adjustable tamp pressure, going up in 1kg degrees, from 10kg to 30kg. I set mine to the lowest setting of 10kg as I wanted to grind as fine as this would allow me, which is also why I stopped collapsing and just kept to palm-tapping. Day 10 post-roast, burrs washed and aligned, I started taking samples that morning: M = 20.18, SD = 0.02, n = 20. Yield? In grams, M = 38.9, SD = 0.75.
Was that the Puqpress, low pressure, no collapsing, 58.5mm tamp, or all combined? Fuck knows – that’s the realm of multivariate statistics and a) I probably don’t have big enough a sample set, b) I’d need better and tighter laboratory conditions, and c) fuck multivariate statistics. It’s the last subject out of three statistic papers I’ve had to take for my Behavioural Studies degree, and I guarantee you the only way I passed the previous two was because you’re allowed a double sided A4 piece of paper into the exam, hand written or typed, with whatever you want written on it. You better believe I had that in 6pt type, margin to margin, and highlighted the shit out of it – and so far I’ve got Credit and a Distinction out of this strategy. Defining, testing, and analysing the variables involved in coffee is a problem yet to be figured out, but it will be by multivariate statistics – and that’ll be the cornerstone of super automatics, but not this post. With the Puqpress in place and believing all my variables were now neatly lined up, it was time to answer the original question and pit 9 bars against 6.
Nine Bar vs Six Bar
Day 10 post-roast, burrs washed, burrs aligned, and for the first time in three months I was raising the pump pressure back up to 9.5 on the dial. With my 4oz cups neatly stacked atop the machine, I started pumping out 20 samples during service after an hour of warming up.
Some problems immediately presented themselves. My metrics were simple enough – 20g dose, 40g yield set with volumetrics, in 29 – 34 seconds. I’d already decided to reject any shots outside these metrics, along with any visibly channeled shots, as this is what I’d do during a normal service. Within 20 shots I already had three rejections for long or short shot times and four rejections for visible channelling. This was kind of unique – I could not recall a single instance of channelling in those three months at 6 bars, and now I had four within 10 minutes. Strike one for variability against 9 bars, strike two for failing to consistently hit the shot time metrics. I also hadn’t realised how coarse I’d gone at 6 bars as I was far finer now at 9 bar. Otherwise the shots poured fairly well, and while they didn’t taste great compared to what I’d been drinking down low – they didn’t taste bad either. I wasn’t expecting that. But hey – more surprises were on the way.
At 6 bars, and besides an annoying five minutes dialling in to a coarser grind to hit my metrics again, there were precisely zero channeled shots, and precisely zero falling outside the shot time metrics. Two hours later I had 40 shots pulled, lined up in front of me and neatly labeled, and it was time to refract.
9 bar: M = 20.32, SD = 0.26, n = 20
6 bar: M = 20.41, SD = 0.17, n = 20
The yield SD for 6 bars was 0.83g compared to 1.17g for 9 bars, so there was that at least. Time was negligible really, 0.09 seconds separating the two means with a slightly higher standard deviation for 6 bars.
I raged at this result. For days. To see a statistically significant difference between the mean for 6 bars compared to 9 bars a power analysis told me I’d need a sample size of at least 80, four times what I had. And even then I thought it wouldn’t fucking matter – 0.1% difference in extraction yield would mean nothing to anyone casually looking over the results, especially when we’re lead to believe a higher extraction percentage equates to a better result, and forget about the flavour.
But the shots at 6 bars tasted better! But how do I convince you, the reader, that my subjective taste experience was better than the scientifically objective numbers the refractometer was showing me? I told anyone who would listen how crazy it was we can only measure “how much” we extract from the coffee, but we can’t measure “what” was extracted. Because whatever flavour compounds low pressure was extracting, it was subjectively better than 9 bars in comparison. But I have no objective evidence to prove that statement, because we rely on what really should be called a qualitative measure than a quantitative one. It was fucked.
And I would have continued to rage after these results, and more than likely I would have run that experiment of 80 shots at 9 bars, 80 shots at 6, and proven there was a statistical difference between the two. But I didn’t need to because Andy Schecter had already run his own experiment. His experiment helped crack open the door to figuring out this low pressure lark, as he’d already provided a solution for me.
The Effect of Bean Origin and Temperature on Grinding Roasted Coffee
All coffee ground shattered statistically the same, independent of processing method, degree of roast, and origin of bean. Cold coffee makes a more narrow fines spread. Dot point summary of Christopher Hendon, Matt Perger, et al’s peer reviewed study released in Nature two months ago. Also known as “Bad News For Mythos Owners”. Heat, according to this study, was not a good thing for shatter transitions.
With beans starting from around eight to nine days post-roast and organising them into batches, cleaning and aligning the burrs regularly, setting volumetrics daily, VST baskets, 10kg tamp pressure from the Puqpress, and a low water debit and low pressure – I knew heat was bad news because I still had variance. The grind temperature was one of the last variables left for me to figure out, but at least it was within a known range. Low pressure, along with my other controls, had a huge effect on mitigating any variance caused by the Mythos and it’s heat element. That was the main reason why I dropped to low pressure in the first place – to mitigate any variance caused by unpredictable shatter transitions of the beans at high temperatures. Variance caused by heat was a non-event for me.
But Andy, while busy changing the coffee lexicon from flow rate to water debit, or producing point of view videos of his EK43 prep, had also decided to do a test with beans at low temperature and room temperature. He wanted to see if he could replicate the results of Hendon et al’s study.
In his mind he didn’t have any luck. In a three-by-three test, performed twice, comparing room temperature beans at 20˚C to frozen beans at minus 16˚C, Andy found a negligible difference in EY. To me though I felt this absolutely proved the theory. There was a 34˚C difference between the two conditions, and yet the frozen beans still achieved a comparable result to the room temperature beans. Thinking these results through, I couldn’t help but think of my own experiment.
Because to me the frozen beans achieving a comparable EY to the room temperature beans was along the same lines as 6 bars achieving a slightly higher extraction yield than 9 bars – with a coarser grind. Temperature a known driving force for extraction – but so too is surface area. More surface area achieved through a finer grind should correlate to a higher EY. Conversely, a coarser grind should correlate to a lower EY. Yet in my test, the average EY at 6 bars was 0.1% higher than 9 bars. Maybe there was something to this low pressure lark after all, all I needed to do was pull this string a little further.
Remember Kids, Extraction Percentage Ain’t Nothing But A Number
I remember the first week Matt Perger opened up the Barista Hustle Slack to the general public. There was utter and absolute radio silence that first week. And if any of the other 1000 or so people who signed up were anything like me, they were silent because they were scrolling further back through each page. On every channel they would’ve found knowledge being dropped from Matt himself, and from World Barista Champions, World Brewer’s Cup champions, luminaries of our industry, and veteran baristas and roasters. Fuck me, learning espresso as an 18-year-old all I had was a ripped, tatty, stained-to-shit copy of David Schomer’s “Espresso Handbook” along with a bunch of voodoo hand-me-down dogma bullshit to contend with. I had to sift through shit managers beside shit roasters with shit baristas trying to find the tiny, fleeting, shiny gold nuggets of wisdom that shed some small shimmering glimpse of light inside the inner frustrating mystique of espresso. And these nuggets were now being tossed out left and right like candy. And I chowed down on that shit like a kid on a cupcake. It was glorious.
The Barista Hustle Slack was, and still is to a lesser degree, a brilliant medium for communication between peers, and a great way to foster a community around creating great coffee. Seeing Christopher Hendon readily reply to questions and queries on the #water channel motivated me to message him privately, ultimately leading me right to this point here. But the BH Slack quickly became a small satellite orbiting the behemoth that became the Barista Hustle Facebook page, and wandering into my attention on a post about low pressure came Mat North, from Full Court Press, in Bristol England.
We started conversing, I showed him the results from my experiment, and he reminded me of his post here, and how the extraction percentage is just a number – it also needs to be correlated to taste. I’d been chasing high numbers, going for those marginal gains, and I hadn’t really been paying much attention to taste. To illustrate this point, Kegman Kipp, way over in California, had asked me to pull a 1:3 EK style to see what I got. I achieved a 23% mean over five shots – and every one tasted horrid. Sure, easy strawman there to create, but it reminded me (and that’s what chasing high EY numbers do to you – you forget) that flavour should be king. Chris Baca showed here how a refractometer should be a tool to guide your palate, Scott Rao showed here how that tool can be used, and Matt Perger said himself on the BH FB page – “More extraction + better flavour = better”. Note the “+”. I’d missed that momentarily.
Kipp was one of the first people I contacted on Slack, and we struck up a productive friendship over our shared usage of the Mythos. He lowered his pump pressure when I asked him, and his results mirrored mine. This was however only after turning off his pre-infusion. For a month he was getting glimpses of the flavour low pressure can provide – but he was living half in the 9 bar world, and half in the 6 bar world. Low pressure demands commitment – pre-infusion needs to be off.
We would occasionally have conversations about him swapping over to an EK43 but he kept coming back to one simple point: he bought the Mythos because, in his own words “3 OG WBC champs and an Irish bloke WBC 4th placer have pretty good taste buds”. In his mind solubility created by the heat was the ticket to why Mythos espresso tasted so good. That stuck in my mind.
Finally, into the conversation Gregory Scace walks – so famous they named not one, but two pressure and temperature measuring devices after him. In two replies to Andy’s experiment he finally allowed me to fully understand that picture Hendon had originally sent me. In a moment of clarity in the middle of the morning rush, floor packed with customers, death-stares everywhere, on top of the grind, adjusting coarse and fine, wiping the tray, raking beans forward, stacking cups and smashing shots – solubility, heat, tamp pressure, pump pressure – it all finally made sense to me.
Fuck Yeah Science
The issue with describing an extraction as even, Greg explained in his reply, is that there’s actually no such thing. The temperature at the top of the puck is equal to the brew water temperature, but “the coffee at the bottom of the coffee cake comes nowhere near it – ever”. Factor in a seven degree change in temperature at the bottom of the puck compared to the top, he continued. Heat is a known driving force of extraction, so at the top of the puck the coffee is being extracted, but cooler water at the bottom makes extraction there a lot harder. But a solution of coffee and water is being created at the top of the puck – what I guess we’d call the actual “espresso” – which has to make its way through the rest of the puck out into the cup at the bottom. The concentration of this diffusion is called, funnily enough, the concentration gradient – which is stronger at the top of the puck than at the bottom. This further complicates extraction at the bottom as the water getting through at that point is actually a mixture of dissolved coffee solids in colder water – so you’re basically shit out of luck extracting flavour from the bottom of the puck. This idea is not new, Maxwell spoke about this in his post here, asking us to think about what our real dose is. How much, he asked, of the dry dose is actually being extracted during the espresso process? So really, what’s your real dose, and what does this mean to your extraction percentage read from the CoffeeTools app?
But in that moment of clarity behind the machine, Kipp’s idea of solubility hit me. As the grinds were already hot, the difference in heat between the brew temp and the grinds temp was smaller so equilibrium between the two can be reached quicker than if it was lower or more variable. Faster temperature equilibrium means the water can get to work on extracting flavour from more of the coffee faster than if it were lower. As it was already at a higher temperature at the top of the puck then water at the bottom wouldn’t be as cool either. Mat North had said that with the Mythos at least the grinds were within a known temperature range. We didn’t have that before, but the Mythos constraining temperatures to under 50˚C, now offered us this information.
So the brew temperature of the water was able to reach the bottom of the puck faster, adding to the already sped up extraction at the top of the puck. But could we make it reach the bottom faster, and in an even fashion?
Is that picture making sense yet? Again, from Hendon: “what matters is the difference between particle A and particle B, and then what matters is not how hard you push, rather the forces at their interfaces. And as you can see, this is inhomogenous”. Inhomogeneous? That would be the forces at the interfaces of particle A and particle B. Those particles? That’s your particle size distribution, and grinding finer creates a larger amount of little bits, and a smaller amount of bigger bits. If you follow the pictures on that original link, you’ll see how finer particles essentially distribute themselves evenly into a confined space with some slight agitation. In this case the confined space would be a portafilter, and we’ll just go and call that slight agitation “palm-tapping”. Tamping hard or towards “maximum density” creates inhomogeneous or uneven forces at the interfaces of the particles. So when pressurised water arrives the velocity of the water will be uneven, as it’s passage through the puck is past areas of uneven packed force. Uneven passage creates variable extraction. What’s needed is an even distribution of energy throughout the puck – a controlled tamp pressure.
In my case I had the Puqpress, Kipp had the Eazytamp, and in both cases we were tamping down at a controlled 10kg of pressure. I tried at 20kg and ended up with a mean of 19.46, SD = 0.43, n = 10. At 30kg I didn’t even bother testing – nothing emerged from the portafilter for over a minute, and adjusting the grind at 6 bars gave me coffee grinds looking close to French press. 10kg seemed to be the optimum tamp pressure for 6 bars – I’d hazard a guess this tamp pressure would change dependent on pump pressure, and would be why we grind fine and tamp hard at 9 bars. The world’s a different place down low however. Down low we grind fine and tamp light. Kipp found great results on the West Coast; I found great results in Queensland.
But grinding fine, while we know allows for a greater surface area, also creates resistance to the pressurised water. And at 6 bars that resistance is close to nothing compared to 9 bars, so consequently the shot time will blow out. But Greg also pointed out another essential feature needed for a consistent espresso extraction – time. Surface area, heat – and time. So it didn’t matter if our shot times blew out past 30 seconds. We could compensate for that with the brew temperature, bringing it down slightly so the extended “dwell” time the puck would have with the heated water wouldn’t over-extract our shot. I was already at 93˚C, taking it down half a degree to 92.5˚C, while extending my shot time to 40 seconds, gave me my best result yet: M = 21.06, SD = 0.30, n = 10. If I thought it couldn’t taste any better than I already had it, this espresso was beyond amazing. This was no longer about marginal gains – it felt like a different coffee.
But Mat wasn’t finished with me, saying. “I reckon if you lowered your dose, your EY would be even higher”. It took me a while to understand what he meant, but when I did, it made sense – a lower dose meant to hit my shot time of 40 seconds, I’d have to go finer again. Not only that, but setting my tamp pressure to 10kg would mean I’d have to go even finer again to compensate for the lighter tamp by comparison. Finer upon finer and I’d have a much greater surface area, and extracting within the same time limit I just lowered my volumetrics slightly to keep my 1:2 ratio.
That volumetric compensation was pretty shit – I set the weight a little shorter than I should have – and I hit a 35g average compared to my target 38g. Which would probably help explain why I was down to a mean of 20.28% EY, SD = 0.30, n = 10.
But don’t forget kids – extraction percentage is just a number. Let me instead try and explain that morning to you.
So this is it. If you’ve followed me all this way I’d like to commend you for your perseverance getting though not only this small novella, but also the two other parts of the trilogy here and here, with a slight departure here. During this time I’d like to think I’ve built up at least a small modicum of trust with you, because that’s all I have to trade on when I describe what happened that morning.
The recipe was simple: 6 bar pump pressure, 10kg tamp pressure, 92.5˚C brew temperature, between 40 – 50˚C coffee grind temperature, 19g dose in a 20g VST basket, 38g target yield, 40 second shot time. Staff member number one arrives, I pour a doppio for her, she takes a sip, frowns, looks at me and asks “what’s this?”. I ignore her (I can be a shit boss sometimes …) and instead fire back “how’s it taste?” “Amazing” she replies. “What single origin is it?”. I smile back. “It’s our blend”. She frowns and keeps sipping. If you can imagine a metaphorical ctrl-alt-delete being performed on someone’s brain, that was what her face was showing me. I completely understood – the first shot I pulled for myself earlier I had one of the tradies working in our building come up to chat with me. I have no idea what he said. I was staring down at my cup, trying to comprehend what I was drinking. Sweet, dense, lip-smacking acidity, voluptuous, concentrated, complete ctrl-alt-delete, reset my brain, because what the fuck am I drinking?
Staff member number two arrives, I pour another doppio. He sips. A slight grin emerges coupled with a frown. Ctrl-alt-delete. “Ok” he says. “What am I drinking?” “Our blend” I reply. He laughed slightly, not moving from that spot, a full two minutes of silence from him as he occasionally closed his eyes looking down at his cup, sipping away. Ctrl-alt-delete.
The owner arrives, we’re busy now; I pull his shot in between customers and hand over a piccolo to him. He sips, looks back, his frown bordering on confused anger. “What the fuck is this?”. I smile back, “how’s it taste” I ask him. “Holy shit” he says as he puts his glass down on the bench, finger pointing at it accusatory. “That” he points again at the glass. “Is the best damn coffee I’ve had here. Hands down. I can’t remember a coffee like that before. Holy shit. What’d you do?”
What did I do? I just followed down the path Hendon sent me on, and physics provided the details. It’s quite possible I’m interpreting those physics wrong, or attributing characteristics to mechanisms that don’t work the way I think they do. So call this my interpretation of low pressure theory – in the end it works in the cup, and it makes sense to me. I’ll let you figure out if I’m wrong, but here’s a recipe for you to first try at least.
Low Pressure Rehash Recipe
When we used to “dial in” every morning, it would always involve adjusting the grind to hit our 30ml in 30 seconds, 9 bars of pressure, 93 – 94˚C, tamping as hard as we could or at “30 pounds of pressure”. Then we got scales in, ditched Mazzer’s for EK’s, changed our ratios up to a 1:2 or more, weighed grams in, weighed grams out, extended the circumference of our tamps, nailed temperature stability for our machines (mostly) – but all this and we were still dialling in with incomplete information. You never adjusted brew temp, you had no idea what temperature your grinds were, and you had no idea if you were tamping at a consistent pressure. If you did experiment at low pressure, you probably adjusted the grind coarser, tamped hard, kept the brew temp at 94˚C, extracted in 30 seconds or less – and you would have come up with an under-extracted terrible tasting espresso. Tamp pressure and grinds temperature, in tandem with low pressure, never made it into the conversation because we had no idea what the fuck they were.
But now we do. And with that understanding we can really start to dial in our espresso.
Let’s start with pump pressure. But first you can’t keep water debit out of the conversation as it’s essential to your pressure. Think of watering your garden pot using the garden hose with a sprinkler attached, then replace that hose with a firehose – that’s the difference between a slow and fast water debit, and a nicely watered plant and one blasted over the neighbours fence. It’s a measure of volume of water delivered within a certain amount of time, and a good metric to try and hit is between 200 – 280ml in 30 seconds. Before lowering your pressure use the right tool for the right job and correct your water debit with restrictors – if you can. Which I hope you can. It’ll make your espresso taste better at 9 bars and simply magical at low pressure. It’ll also give back the body and thick mouthfeel you may feel you lose compared to 9 bars, while making your extraction more consistent. It’s your water debit that feeds your pump pressure, enabling that consistency. With a low pump pressure and slow water debit in place there’s a few things now you can achieve which you can’t at high pressure:
- We get a slow movement of water through the puck – let’s call this velocity. This allows both diffusional and temperature equilibrium throughout the puck. This is mainly at the top of the puck still, but more at the bottom if you have a known raised coffee grinds temperature. Through that temperature you have a lower dissipation of heat as the water passes through the puck.
- You also mitigate any variance caused by imperfect technique – I’ll send you the video I took if you want where I dug a 5mm divot into the surface of a puck, pulled a shot, and while it did show the water initially rushing to that corresponding point beneath the basket – it never came gushing out or overtly channeled. Within milliseconds the rest of the espresso started forming on the bottom of the basket as well. The next shot I pulled sans divot, poured in exactly the same time. That’s a 19g dose – 18g or 15g, as Mat or Maxwell would tell you, presents problems for channelling based on how thin the puck is. But still – you’ll get far less channelling at low pressure. In my main experiment I had four in 10 minutes at 9 bars, zero in three months at 6 bars.
- Not only does low pressure welcome one and welcome all with imperfect technique, it’s also forgiving of variable particle size distribution – again, because of velocity. Slow movement means bigger bits extract at a proportionally higher rate than they would have at a higher pressure, as contact time has been increased.
What about tamp pressure?
- “Even distribution of force throughout the puck” is another way of saying “even force applied at the interfaces of each particle”. So the tamp pressure needs to be light enough to not introduce uneven force at these interfaces. 20kg was too much, 10kg produced amazing espresso. It could be that 15kg might do the same, or even the original 30 pounds, or perhaps even as light as 5kg. But it needs to be reproducible – and that’s what products like the Puqpress and Eazytamp provide – reproducible and consistent tamp pressure.
- This consistent tamp pressure means the low pressure water is not meeting uneven resistance throughout the puck – it has a clear passage to the bottom, meaning we now have a true velocity, set by the pump pressure and fed by the water debit. You never had this clear passage before, it was always muddied by imperfect tamping, collapsing, made worse by high pressure. So now we’re extracting from a larger percentage of our puck than we were before, as the water has a clear and consistent pathway through.
- A lighter tamp also means a finer grind. A finer grind means we can control the water velocity, thereby controlling the extraction time as a function of temperature and diffusional equilibrium, while also gaining more surface area.
And brew temperature?
- This is a function of how we dial in at low pressure. A fine grind slows down the water velocity, while a high grinds temperature produced by the Mythos speeds up temperature and diffusional equilibrium. So we run the risk of over-extracting when using the Mythos. Half a degree down – for me – compensated for this.
- What if you’re using frozen beans? I do at home, and its great – it gives me a known temperature as well as an even particle size distribution. So I go in the opposite direction, 94˚C.
- Again, we’re dialling in with complete information now, so time needs to be experimented on, with the understanding that we need to provide enough time for temperature and diffusional equilibrium while compensating for the finer grind slowing down velocity. It will hit an equilibrium however, and once it’s there, extraction really begins. So for me, that took 40 seconds before I started encountering slightly bitter, dry flavours characteristic of over-extraction. Frozen beans were completely different. I ground fine enough to choke a Slayer and had some pours upwards of 1 minute 20 seconds. And it tasted great.
- Can I draw shot times out at work to 1 minute 20 seconds? Fuck no! Going up to 40 seconds was rough enough getting though 500 cups in a morning. An extra 10 seconds was a lifetime. But I adjusted – not just in workflow, but in dialling in. Brew temperature was kept constant at 92.5˚C, grinds temperature was constant between 40 – 50˚C, pump pressure was constant at 6 bars, tamp pressure was constant at 10kg, dose was constant at 19g, yield was constant at ~38g – and all I changed was the fineness of the grind to hit 40 seconds.
- VST recommend a +/- variation of 1g – so I took them at their word and shot down to 19g. VST baskets are manufactured to promote faster flow through the basket by having more filter holes than stock standard baskets. So already you’re going fine to compensate for this faster flow and the restriction it causes. Now you have a lower dose with less depth for the water to penetrate, which also leads to a faster flow. So you grind fine again to compensate.
- How does this affect acidity? No idea – but it does. In an amazing way. In a previous life when I was selling wine I always held acidity to be the backbone of a great wine – not to be confused with screw-your-face-up sourness. Instead you should be looking for lip-smacking acidity. It holds up all other flavours and suspends them there with full clarity. It’s no different with espresso, and that one gram drop in dose made the acidity shine.
And that’s the recipe, that’s how you dial in with complete information. You’ll need to experiment, but it won’t take more than a few shots to start seeing what direction you need to go in. Trust your tastebuds, correlate with a refractometer, and try and understand all these variables as sitting on a spectrum. You just need to find where on each spectrum your coffee, roast degree, machine, and grinder, can get you.
When viewed through this lens this low pressure rehash is really a clearer understanding of the relationship between variables. Water debit is related to pump pressure. Pump pressure is related to tamp pressure. Brew temperature is related to the coffee grinds temperature. Time is related to fineness of the grind. These variables combine to influence your ratio, while also promoting or demoting successful temperature and diffusional equilibrium. Physics guides all of this.
If I’m going to quote everyone else, I may as well quote myself here, as I’ve said this before and I feel it needs to be said again: “none of this is really new, it’s part of basic espresso theory we’ve all been taught … What is new is our understanding of the physics involved, and for that we need a new paradigm of thinking.”
I’m not alone in this thinking. Mat has written a brilliant piece here, outlining his own experiences with low pressure, with his thoughts on the possible mechanisms involved. I share and agree with them all, and while I may not yet fully understand the “how” – I know what it tastes like in the cup. Perhaps consider these two pieces the first sentence in a broader conversation, continued on by anyone else who decides to untwist the nut and go low.
It took me three months to cast off the shackles of 9 bar theory, lower my pump pressure and my brew temp, under-dose, tamp light, grind fine, and pull long. Along the way I’ve had the immeasurable help, either outright, tacitly, or unknown, from Christopher Hendon, Mat North, Chris Kipp, Matt Perger, Andy Schecter, Gregory Scace, and many others. But it was always the physics that lead me here. This low pressure lark is not a rehash – it’s how espresso should’ve always been made. It’s just taken us a while to get there.
(Agree with me or not here on Twitter, and find answers to questions like “what did Michael do at work today?” on Instagram here – every like gives me one more VST filter to use …)