Bluecold wrote:Before continuing this discussion, which I think is interesting, let's try to agree on what the concept of flow profiling is.
As far as I understand, flow profiling is a way to reproduce a shot which isn't exactly the same, by prioritizing flowrate equivalence over pressure equivalence. Ie, you have shot A, and shot A is amazing. Shot B will, despite your best efforts, not be exactly the same. Shot A is defined by dose, grind, pressure profile, flowrate profile and a host of other factors. Flow profiling advocates would say that one should try to keep the flowrate profile the same to reproduce shot A as closely as possible, even if it means sacrificing pressure profile equivalence.

Pressure and flow are inversely related by puck resistance (simplest formula is P=RF). Changing pressure changes the flow, and changing flow changes pressure. If you prepare similar pucks, you should be able to get similar extractions. (I won't say "identical" because that's not possible.)

Here are extraction plots from half a dozen DE1 shots that I pulled over the period of a week, using a simple declining pressure profile after preinfusion. Pretty similar, I'd say.


And yes, I agree that this discussion has become interesting.

At least for me, your definition is far too restricted and conflates the process with the objective that one might have.

For me, "flow profiling" is managing the flow rate through the bed of coffee as it is extracted.

One common objective of using flow control is be to be able to "best" extract coffee from the bed

At its simplest, flow profiling is choosing an infusion method over a percolation method, selecting effectively zero flow, or some flow.

From there it can take many forms, as flow rate at a given point in time is related to the pressure across the bed and the bed's effective resistance at that moment. These may include all kinds of restrictors, such as gicleurs, E61 mushroom valves, needle valves on the supply side, valves on the exit side (such as a Clever dripper), filter paper in the basket, even the type of V60 filter paper, that change the overall hydraulic resistance to flow at a given supply pressure. One can also control the pressure or meter the flow. There are, no doubt, other methods as well.

These apply whether open- or closed-loop control is being used. If closed-loop control, how you control the flow is much less important than what you control.

If your objective is to be able to "best" extract coffee from the bed and you agree that

• The flow rate impacts extraction "quality"
• The pressure impacts extraction "quality"

then you need to acknowledge that, as there is a fixed relationship between pressure and flow at any instant in time, you can pick one to control and the other becomes a dependent output.

Pressure-driven control is familiar to many. It is how a typical pump-driven machine or a spring lever works, especially during the extraction phase. (There are certainly valid considerations around, for example, when the flow is restricted by a gigleur during PI.) A simple, pump-driven machine measures pressure at the OPV or bypass valve, and closes the loop at that point. A spring lever provides open-loop control, based on the spring rate and the position of the piston.

Flow-driven control has been practiced for decades, but not by that name. If you read the various manual-lever threads you'll hear over and over, "I ignore the gauge and just watch the stream, pressing harder or softer on the lever depending on what I see." This is closed-loop, flow-driven control. Agreed the accuracy of the flow measurement isn't to 1%, nor is the system response time to perturbations in the tenths of seconds, but it works.


Once you get into what is "better" for achieving a specific objective when that includes significant puck-to-puck variation, then there is a lot of testing to be done to determine what kinds of variations contribute most to preserving or destroying the result in the cup that you want.

There are some rough explorations that have suggested that, for many grinders, for espresso, somewhere around 1-2 ml/s during extraction tends to be commonly in the sweet spot for medium roasts. The "18 g, 1:2 ratio, in 25 seconds" suggestion can be thought of as a flow-rate suggestion. 36 g in the cup, less 3 g of extract is around 33 ml of water. 25 seconds less 7 seconds for ramp-up is around 18 seconds. 33/18 ~ 1.8 ml/s.

Similarly, if the pressure is extremely low (0-1 bar), you're making "drip", in the slightly higher range, something more akin to the original steam-driven caffe espresso. What many consider "espresso" seems to be associated with pressures in the 4-9 bar range. Somewhere above 9 bar, some find that the quality of the espresso diminishes in various ways.

Again, none of this should be a surprise. With a conventional, pump-driven machine or a spring lever, the barista chooses a grind and dose for a given coffee and setup that has a fixed, pressure profile during extraction to be able to achieve a flow rate that results in "best" extraction of that coffee for that pressure. With a manual lever or a flow-driven profile, the barista chooses a grind and dose for a given coffee and setup to be able to achieve a pressure that results in the "best" extraction of the coffee for that flow rate.

The goals for (and reasons for interest in) flow profiling have changed quite a bit in the 4 years that Decent has been shipping.

When we started, both Rao and Hoffmann told me their interest in flow profiling was to make the "worst shots less bad". For example:

• Ie, instead of a "gusher shot" that is channeling terribly and outputting 8 ml of "espresso", while the espresso machine tries to maintain pressure (ie, what this is what happens with a 9 bar shot if a big channel opens)
• you could instead limit the damage of the gusher and cap flow at (say) 4 ml/s. You'd get lower pressure, the channel would be more likely to heal itself, and the shot would be less awful.

Then, after about 2 years of the Decent being in the wild, this goal for flow profiling "went away". What everyone realized is that flow and pressure both matter enormously, and the best espressos are those where both flow and pressure and in the range you want.

But also, as Hoffmann mentions in his video below, these days it's really difficult to make the Decent make a shot that channels. Hoffmann wasn't able to, for his slow-mo video. Coping with gushing channeling just isn't a focus any more, at least not for Decent owners.
So these days, flow profiling on the Decent is most often seen as a "step" in a profile that also uses pressure profiling, such as during preinfusion.

Flow profiling is sometimes used during the high-pressure stage of making espresso, but because small changes in flow yielding large changes in pressure, it can be quite difficult to dial this in.

• For instance, the final step on Rao's "Blooming Espresso" uses a flow step, and it's famously difficult to dial that shot in, made harder because the Blooming technique requires a very fine grind.
• However, the Allonge, which is a 5:1 "espresso" in ~35 seconds, because it uses a coarse grind, is not difficult to dial in even though it's entirely flow profiled.

To summarize, I'd say that "flow profiling" is "another tool in the toolbox" and has been found useful in a variety of situations. But it is not a panacea, and I for one would NOT want a machine that only did flow profiling. If I didn't have the Decent, I'd own a springless lever with SEP https://www.naked-portafilter.com/smart ... -profiler/ attached to it. That would give me manual control over pressure, and indirectly over flow.

-john

With my DC Mina, I use flow profiling to create the pressure (that I think) I need during extraction. The actual flow after the puck is the result of flow and puck resistance.

The DC DFR is a digitally controlled valve that can deliver 5 different flows during extraction. DFR# delivers #ml/s of water without coffee resistance.

So if I want to start with a very long PI, I program a low flow, DFR-2 (Slayer style).
If I want a blooming phase after PI, I program DFR=0, to let the pressure decrease to zero.

If I choose a low DFR after blooming, pressure will re-build up slowly. If I choose a high DFR, like DFR-8 pressure will build up fast and the puck will be compressed and pressure will stay high.

If I am able to not hit compression, with for example DFR-4, pressure will rise and after a while will drop slowly, because of puck erosion.

So, like John wrote, flow profiling is another tool in the tool box. I don't see it as a goal.

Edit: sep 15, 21:25 CET

Great thread. The comments by John (RapidCoffee) and John (Decent) agree with my experiences doing flow and pressure profiling on my heavily modified GS/3 AV (Android/Arduino-driven gear pump, needle valve, bypass solenoid, flow and pressure sensors, BT interface to scale, etc.)

Let's not lose sight of the primary use case for flow and pressure profiling, which are to increase extraction of light and light-medium roasts that don't extract fully with "standard" espresso profiling (e.g., 18g producing 36g in 25-35 seconds.) Some claim that flow/pressure profiling can bring out sweetness in darker roasts, but I haven't been able to verify that, as I rarely drink anything darker than medium roast coffee.

Flow and pressure profiling achieve greater extraction by allowing a much finer grind and increasing contact time. The flow rate during a pump's preinfusion (PI) ramp to full pressure is quite fast, which leads to the puck quickly being compressed. This precludes using an ultra-fine grind because the combination of the fine grind and rapid compression will choke the machine. When a much slower flow rate is used for PI, water slowly infiltrates and soaks the puck, which expands, compresses less at full pressure due to resistance by the water it now contains, and provides greater permeability. Thus, you get the greater surface area of the finer grind and longer contact time during PI without the machine choking.

But like the back-end of shots pulled with standard profiling, a few seconds after compression at peak pressure, puck permeability begins to increase, and not long after that the rate of increase gets quite steep -- i.e., the shot runs faster and faster, and the faster it goes the less contact time you have. By ramping down the pressure you can keep the flow rate constant, increasing contact time.

Like the Decent, my setup can do Slayer-like PI, and unlike the Slayer it can ramp down pressure during the back end. The one thing that I can't do easily is a "blooming" shot, where PI is stopped and the puck is allowed to soak at zero flow and declining pressure for some number of seconds before ramping back up to full pressure. I would have to simultaneously cut the gear pump speed to zero and close down the needle valve. After soaking I'd have to open the bypass valve and return the gear pump to the speed that produces 9 BAR at peak, the ramp to which isn't instantaneous. [I can't run the gear pump while the needle valve is closed because the gear pump has no bypass valve and the needle valve is between the gear pump and the expansion valve. I've considered installing a second expansion valve between the gear pump and needle valve, but that could change the stock flow dynamics of the machine.]

Anyway, I'm wondering whether trying to implement blooming would be worth it. I can do very slow PI, so I imagine that blooming would be necessary only for really, really light roasts. Comments from those who do blooming shots?

Peppersass wrote:Flow and pressure profiling achieve greater extraction by allowing a much finer grind and increasing contact time. The flow rate during a pump's preinfusion (PI) ramp to full pressure is quite fast, which leads to the puck quickly being compressed. This precludes using an ultra-fine grind because the combination of the fine grind and rapid compression will choke the machine. When a much slower flow rate is used for PI, water slowly infiltrates and soaks the puck, which expands, compresses less at full pressure due to resistance by the water it now contains, and provides greater permeability. Thus, you get the greater surface area of the finer grind and longer contact time during PI without the machine choking.

Hopefully without sounding like a suck-up, I'm really impressed with you've been able to build yourself and the insights you've had. I agree with everything you wrote, and note that it's rare thing to read what you've written, outside of our DE1 owner's forum.

Peppersass wrote:Anyway, I'm wondering whether trying to implement blooming would be worth it. I can do very slow PI, so I imagine that blooming would be necessary only for really, really light roasts. Comments from those who do blooming shots?

Can I recommend you give the "Allongé" style espresso a try for ultralight roasts? From what you describe as your setup, I think you can do it.

Here's the Decent description of that:

An amazing long espresso for light roasts, this is the biggest fruit bomb of any brewing method we know. 5:1 ratio, 35-40 seconds, coarse espresso grind. If close to the right pressure, make 0.5g dose adjustments to get to an 8-9 bar peak. The very high flow rate means small grind adjustments cause big pressure changes. An advanced technique, allongé averages 24% extraction.

and a graph:



Don't make the water too hot (not as hot as normal espresso), because the fast flow rate means there's almost no cooling of the input water.

As to Blooming, it's great for expensive beans, where you really want to extract a lot and get a lot of water contact time, and don't mind a (controlled) bit of acidity. But the Blooming is for when you want a concentrated drink, not the dilution you get from the Allongé.

Peppersass wrote: Anyway, I'm wondering whether trying to implement blooming would be worth it. I can do very slow PI, so I imagine that blooming would be necessary only for really, really light roasts. Comments from those who do blooming shots?

The major benefit (for me at least) is that the blooming preinfusion is so effective that it allows an even finer grind than a very low debit start. For some of the lightest stuff, this can be really helpful. Unless something is so light that I just can't get it flowing at a rate I like, I use a low debit start until I'm getting flow from the puck. I wouldn't bother blooming if I didn't have to.

FWIW, I love how concisely you point out the reason profiling allows for higher extractions. There's so much misunderstanding about the reason it works, so finding someone else that agrees about the cause is nice.

Just found this thread and finding it interesting. I worked for years in critical care in ICU's, and while lungs are not coffee pucks, many of the same challenges of flow, pressure and volume have been present in mechanical ventilators for years. Even today, manufacturers are still tweaking algorithms to see if they can positively impact patient outcomes. Bravo Decent in your efforts. If you find yourself seeking continued inspiration for algorithm improvements, have a look at ventilators!

The pace of understanding prep and extraction has been quite rapid with the data the DE1 provides, along with its repeatability and control. You might want to follow https://coffeeadastra.com/ and perhaps his Instagram and Patreon for some of the insights being developed.