Time-based flow limiting: thoughts from Maro machine review
I watched this video review of the Maro yesterday:
and one thing that struck me was the “correction” the machine can do, as well as the “grind advice” it gives. I thought about it for a bit, and I think it boils down to:
the maro makes you specify a shot time and desired weight-cup.
the built in profile sounds like “fast preinfusion, followed by flat 9 bar”, with atwist…
if the shot is running too fast after preinfusion/pressurizing, and the shot time is estimated (current flow rate/end weight) will be shorter than the goal, then Maro lowers the flow rate.
I'm fairly sure it does this by averaging the entire remaining time and weight needed.
Ie, if we're 15s into the shot, 15g is already in the cup, we want a 30s shot, and current flow rate is 2ml/s, then Maro would lower the flow rate to 1/ml, so that 15s x 1ml/s = 15g.
Naturally, this shows up as a pressure decrease (ie, lower flow, causes pressure decrease)
If the Maro does this adjustment often (say, once per second) then a pressure curve looking like a “Lever shot” should occur (ie, relatively stable flow rate, decreasing pressure).
While I can't think of any coffee reason why “always having the same shot time” would make espresso taste better, I can see that a consistent shot time would cause a pressure curve that looks like flow profiling, and that would make better coffee.
And finally, grind advice:
if the pressure reached during your espresso is low (ie, under 8 bar) then Maro can advice you to “grind finder”
and if the shot time couldn't be achieved, and ran too long, then Maro can tell you to “grind coarser”
example:
Some thoughts I have about Maro's approach
for people who want to make “lever shots”, this adaptive mechanism, followed by some simple advice, is likely helpful.
I've been growing less sure that a constant flow rate after pressurize is optimal, despite the “Adaptive Espresso” profile using it.
Londonium, for example, uses Pressure profiling during extraction, only going to Flow profiling if you have a 'gusher' to limit the damage.
And Londonium (with a well dialed in grind) for me causes gradually increasing flow rate, which helps to keep pressure up.
A gradually increasing flow rate seems like a logical thing to do, to compensate for the normal puck erosion that is occuring
I'm experimenting with this small change to Adaptive, to have a gently increasing flow rate during extraction, to help maintain pressure. I'm finding it especially helpful with lighter roasts and large flat burr grinders.
For what it's worth, I think the DE1 can effectively do the same thing as the Maro's “always end the shot a the same time” by doing a flow controlled shot.
The “flow profile for straight espresso” should always take about 18 seconds of extraction time to reach reach 36g. Adding in preinfusion time (say, 10s) and you're at a 28s shot.
I prefer to not have as a goal “total shot time” but “extraction time” instead. That's because, your first DE1 shot of the day, if you didn't flush beforehand, needs a few seconds to fill the water tubes to have water reach your puck.
I have some thought about Maro's grind adjustment advice, and initially, I liked the idea, as it is similar to my loadable “advisors” idea.
However, Bugs was very against the idea, sounding remarkably like Damian's skepticism of the idea, which I'll boil down to “no, that won't work, reality is too complicated”.
And, in the end, my thinking arrived at the same conclusion. The roast level, bean origin, bean age, basket cleanliness, burr type and size, prep consistency, and puck prep overall, are likely to misdirect a simple “peak pressure tells me how to grind” advice.
Instead, I am leaning more and more toward:
let's have a few years of giving each other advice about shots with Groupthink, where we have you enter some metadata (beans, grind) and the shot data is stored, and the chat has varied advice (your puck prep needs help, etc…) to your particular case. IE, have humans help.
and then (in a few years) let's feed that into an AI. DEREK has been quite good at digging into the Diaspora archives. I'd like to do the same for “how to make my espresso better” advice, but our training data first needs to be much better, or we will get garbage results from an AI. An AI can handle multivariable complexity about an espresso, much better than any rules I could hand-write. If I write a chat feature for Decent users, that has been designed to make clean data for AI training, we'll get much better results.
There are a bunch of manufacturers who think there is a market opportunity for a “mass market version of what Decent does”. The one that has the most people excited is Meticulous, because (naturally) a much lower price than Decent draws attention.
Maro's higher price is a challenge, but (personally) I think they have a great product for HORECA (Hotel, Restaurant, and Café/Catering) and Decent has zero market penetration there. The slightly higher price shouldn't matter for that use, though I think they'd benefit from a partnered or integrated grinder, for HORECA use.
I'm a fan of the Morning Coffee Machine, who I think combines great industrial design, simplicity, and a clever round screen UI at the machine head. It can do a number of Decent-style profiles, but with 3rd wave Nespresso capsules.