Decent Espresso Machines (DE1) - Any thoughts?

[>decentespresso]

I know it's a bit premature when your espresso machine isn't even in production yet, but it almost sounds like you'd be keen to look at doing for home grinders what the DE1 is going to do for home espresso machines. I like your upgraded Tontile grinder and it'll be a more than adequate option for anyone buying the standard DE1, but fit the real coffee geeks that are looking at the + and Pro they're likely to be keen on something better.

You're very good at reading tea leaves!

Yes, it is premature for me to talk about our next major product when we haven't yet shipped our espresso machine.

The slogan on our web site is "Rethinking everything about espresso", and that's all I will say at the moment.

And yeah I agree that Anfim are great grinders. As the company that is essentially the Italian arm of Mahlkonig you'd expect nothing less. Not sure that they can be thrown fully in the 'lower end' category though. The SP450 is over $4k here in NZ and equal or better than most other modern electronic commercial grinders.

The ANFIM Super Caimano is under USD$2000, and I think it's great for that price. It's very "basic" in a way that I really like. I own one, and it's paired with my GS/3, which shares that same great quality.

[>decentespresso]

Some good news on the schedule (for a change)



Last week, we picked the company that is going to make all the custom metal parts (about 50 different parts per DE1). As this comes to $319 per machine (and thus represents $100k in business for 300 machines) and is one of the most crucial suppliers in order for us to make a quality product, we're pretty keen to develop a very good relationship with this company.

So far, their salesperson's communications (via chat software) have been really fast and clear (and hungry for our business, despite our being currently a very small order).

Their turnaround speed is impressive: 25 days to get a single final sample to us, and then 10 days thereafter to produce 300 more of that, once we give the thumbs up signal.

That puts these parts arriving end-of-October, and since everything depends on the chassis, it's good news.

Their engineer today sent 3 questions in, and I was impressed by the clarity of the communications and also that a proper audit was done on their end of what's needed. Good signs.

One question, for instance, was about the type of rivets. Another was confusion as to whether they were supposed to supply the rubber feet (they're not).

Attached above is the "questions document" we just received from them, in case you're curious.

[>decentespresso]

3 new Decent products

3 new Decent products today:
https://decentespresso.com/new

- the Scent One Aroma Standard: 100 perfumes found in coffee and a book describing the chemical composition of each. This is an interesting new coffee description system from Korea, and exceedingly hard to find in stock at the moment. Integrates into the DE1+ espresso machine software. I met Jake, the CEO, a few months ago at SCAE in Hungary. We have a limited stock of 16 kits, which Jake hand-delivered to me last week when we brainstormed all Sunday about the tablet software I'm working on. https://decentespresso.com/scentone

- New "latte" double walled glass. The unique shape opens up wide to let your milk jug enter deeply into the glass. This makes latte art significantly easier to do. We've sold out of all our other glasses and are only focusing on this shape I designed. The "doppio" glass for espresso has a new handle shape, redesigned to more closely follow the tradition. I first showed this idea for a glass 16 months ago in this video: https://www.youtube.com/watch?v=xt5Lel1aD3g Also, we're now selling glasses in quantity: one, whereas we used to only sell them in four-packs. More info at https://decentespresso.com/glass

- New portafilter baskets in sizes of 7 grams, 10 grams, and 24 grams. I was in Italy two months ago, where they mostly use smaller sizes, and last month at Berlin Coffee Fest Bugs also got feedback that we needed to provide these traditionally Italian sizes. We've also added the super-huge 24 grams basket for those who like to "overdose" (in a good way). https://decentespresso.com/basket

[woodhouse]

damn, that's pretty decent value for an aroma standard. the le nez du cafe is around the same price and only has 36 samples.

[TampPolice]

I was disbelieving for years, but what Rao does (gentle side taps, and nothing else) produces the highest pressure according to this test. Rao also angles the portafilter while he taps, which I don't do, but otherwise I've ended up at the same technique he uses (and that was not my intention).

Customer Demand: I want a decent distribution tool (DDT) with distribution profiling system so I can emulate Rao distribution method.

Jokes aside, a distribution tool is probably the only thing you don’t sell as far as the barista accessories go. FYI Matt Perger is currently in the middle of a crowd funded experiment (which I am proudly a part of) to compare ONA and Pullman distribution tools. My guess is that he will come up with a low cost tool of his own design following the experiment, that may outperform the other two designs.

The Mythos is fantastic, my only wishes for improving it would be (a) weighed doses and (b) a locked position for the portafilter which results in a centred mound.

Is it fantastic in both cold and warm states? This is, if you were to grind straight after switching it ON, even if you had it dialed in at a slightly warmer state?

The race for weight based dosing is definitely ON. Let’s see who comes up with it first.
My guess: Compak will win the race with their PKR series but with extremely high prices, then Mahlkonig will release their Barista model which will outperform PKR series in terms of PSD. And then about a year or two later, a fella named John will put the aforementioned products out of business by releasing the version 2 of his decent grinder with auto-dial-in functionality that will detect the type of beans in the chamber and grind to the desired PSD level.

Our current grinder is OEMed from a subdivision of Mahlkonig (from HeyCafe, and then we modify it further inhouse) and my goal with it was to have a very acceptable under USD$1000 grinder with dose weighing.

The Mythos is in a different league.

I've had some Peak shots and they were excellent, but reliability in a production setting is at the moment, not great with those grinders. Once they solve that, the Peak will probably be in the same league as the Mythos.

The grind retention in the decent grinder is the major thing holding me back from getting one. If the burrs are similar to Robur, what is the other assembly like? More specifically, how does it perform with less than 100g beans in the hopper? Is it safe to assume that the particle distribution would be similar to Robur?

Roburs are to be avoided.

On the lower cost side, I think very highly of ANFIM.

I hear you. Back in my uni days, I worked as a barista (part time) in a busy CBD cafe with a Robur and a Linea for over 3years. What I learned in those years was that the Espresso machines cause very little to no grief at all to the baristas if taken care of, by baristas and the owner. It’s the inconsistencies from the roast and the grinders that give you the most amount of grief.
Using Roburs that’s what I learned:
1) Dealing with a Robur in a busy cafe is as difficult as taming a wild horse. After a few days you think you’ve worked out how to tame the monster and then the roaster decides to come up with a different profile/blend and you start all over again.
2) At times, I had to purge up to 100g of coffee after the quiet periods before I would start seeing what I liked and by that time, bloody thing had warmed up again
3) Always have the Robur loaded up to at least 1kg of beans if you want better tasting shots.
4) There is no solution to clumps (if dosing directly into the PF)

Someone who had visited a Mazzer manufacturing facility, once told me that they definitely have all the technology they need to get whatever they want out of their grinders but they deliberately make them produce more fines because it causes a slower pour and that’s how they like it in Italy where almost no distribution method is used with little to no tamping.

As much as I hate Roburs for these reasons, I do have a Mini E at home purely because I can get 20-22% EY out of it with little effort and I only have to purge around 2g of coffee.

[>decentespresso]

damn, that's pretty decent value for an aroma standard. the le nez du cafe is around the same price and only has 36 samples.

It's a new product, they only made 3000 of them, and they weren't aiming to make any money from this version of the kit. My understanding is that they're working on a bigger kit (more scents) and that it'll be about 50% more expensive than the current one. From talking to their CEO, it's pretty clear that this is a very expensive product to produce (100 separate scents) but they're also in this for the long term. They're about to release a new product, a scent kit for tea.

a distribution tool is probably the only thing you don’t sell as far as the barista accessories go.

There's a good reason for that: my experiments indicate that distribution tools greatly increase the variability of shots. Some shots turn out amazing, but the average is worse than no distribution tool.

I have seen talented baristas achieve consistency with NSEW distribution but I think it's a hard technique to master
https://www.fivesenses.com.au/blog/p...-distribution/

and I haven't tried it myself as I'm pretty committed to weighed doses.

FYI Matt Perger is currently in the middle of a crowd funded experiment (which I am proudly a part of) to compare ONA and Pullman distribution tools. My guess is that he will come up with a low cost tool of his own design following the experiment, that may outperform the other two designs.

I talk to Matt a fair amount and we've talked about his distribution tool, and I'm very happy to leave the R&D to him on this one. If he manages to "crack this nut" Decent will probably end up reselling his tool, but first it needs to be shown to objectively work in controlled tests.

You probably all have read the Socratic Coffee post on the OCD:
Examining the Impact of the OCD (and OCD 2) on Total Dissolved Solids Extraction – Socratic Coffee

Is it fantastic in both cold and warm states? This is, if you were to grind straight after switching it ON, even if you had it dialed in at a slightly warmer state?

I've only used the Mythos in cafes where it's been warm and loaded up with a lot of beans.

The grind retention in the decent grinder is the major thing holding me back from getting one. If the burrs are similar to Robur, what is the other assembly like? More specifically, how does it perform with less than 100g beans in the hopper? Is it safe to assume that the particle distribution would be similar to Robur?

Our current grinder is much closer to a Super Jolly, with similar burrs and grinding speed.

On our current grinder, grinds retention is typical of this style grinder, i.e. it's not a zero-retention, single dosing grinder.

I don't make any claims to changing the world with our current grinder. It's a "good value, good quality" espresso grinder. There's not much under USD$1000 that makes competent espresso, and I also wanted to have a fast workflow for weighing doses.

---

Going back to the topic at the top (distribution tools) my experience is that if your grinder delivers a fluffy, not too 'peaked' (nor clumpy) mound in the center of your portafilter, then nothing but tamping is required to get a really even shot. Our current grinder does that by having a X, Y, Z adjustable stand for the scale and portafilter https://www.youtube.com/watch?v=O2mEWG2S7SY

Back when we were trying to design a distribution tool, Matt dismissed our work because we were only affecting the top of the mound. I (eventually) agreed with him.

This thread on HB is really interesting, as it integrates a distribution tool into the grinder chute:
https://www.home-barista.com/grinder...ds-t12954.html

We're investigating this idea as well.

[>decentespresso]

it's the little things

We've hired the 3 best summer interns to work 1 day each week during the school year.

Each time they come, they scrutinize a different component and post a report to our Basecamp discussion group of their finding.

...

Charlotte, a 3rd year mechanical engineering student, yesterday reported on her measurements and inspection of 10 insulating rings that are used where the steam wand bolts down to the front panel. We added these to the steam wand assembly because we found that without this, the ball joint of the steam wand transferred heat to the front mirror panel, and made it unpleasantly hot.




She wrote:

Highlight of Result:
- Actual thickness (about 0.6mm) is slightly deviated from that of part drawing (0.5mm)
- Small black spots appear on washer

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Tommy, also a 3rd year mechanical engineering student, worked on the rubber spacer that fills the gap on the group head handle (so that naked threads aren't visible). We didn't specify the durometer hardness of this piece, because we didn't think it mattered, and when the handles arrived last week, we found that the rubber was soft enough that it compressed unattractively when the handle was screwed in. We paid US$120 to have 1000 of this piece made in ABS plastic, which arrived a few days ago, and now we need to use a knife to cut the old rubber part out (it's glued in) and put the new ABS part in.

Tommy checked this (spacer), and also the thread size and depth.

Here are his results.





. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Stanley, a 3rd year electrical engineering student, is figuring out how to use our very expensive (USD$1500) "Hi pot" tester.
https://en.wikipedia.org/wiki/Hipot

This is a machine which checks for any electrical leakage, even slight. This is useful to see if the insulation is everywhere thick enough, or if some naked connections (such on the PCB) are too close to something else and bleeding a bit of current.



. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

No exciting news in any of this, but you guys have indicated interest in "how our stuff gets made" so here is a bit of the gritty details.

[>decentespresso]

Describe your Espresso



I've been working on a feature to describe (and rate) the espressos you make on the DE1+.

As you're fiddling with various settings, this lets you track what your experiments are doing to the coffee flavor. And because all settings (and live data) is automatically saved to disk with every espresso made, you can load up anything from your history and recreate that God shot you made.

Above is today's iteration. It's been quite a struggle to reduce complexity and the data-entry burden yet still ask the vital questions.

The goal with this iteration is to make it clear that you need only answer the ENJOYMENT question, and hit save. That's enough and very useful.

The "your setup" tab is where relatively-unchanging data is entered. The type and brand of coffee beans, grinder setting, and dose are entered there. I'll post a screen picture of it in a few days when I've gathered and incorporated feedback from this "your espresso tab" I'm posting today.

Many thanks to Mat North and Scott Rao for their extensive help on this.

[>decentespresso]

Now with fewer features!



From the discussion that emerged around my proposed "Describe this Espresso" feature, it became obvious that 5 competent baristas would propose 5 different ways to describe an espresso. No single choice I might make would please more than 30% of the audience.

Following on how I've read Switzerland's governance to work, I prefer to "do nothing unless there is a consensus".

As such, I've simplified the GUI to only have one subjective measurement, and that's "Enjoyment", and "Notes" you can write to yourself. Everything else the DE1+ collects is hard data.

As mentioned yesterday, I now show you the "Your setup" page. Here, things that are relatively unchanging should go. That keeps the data entry load down for you, but hopefully still gathers useful information.

Also, from a data analysis standpoint, I (the programmer) can do a straightforward analysis of which parameters lead to espressos that you enjoy more. Before, with acidity/sweetness/aftertaste in the mix, that sort of analysis was problematic.



To further simplify the main page, I'm now hiding the sliders for "drink weight" and "TDS and Extraction Yield" unless you "Enable these features".

I've also done away with the hideous "rate this espresso" button that was on the main GUI. Yes, I designed it, so it's my fault, and I hated it.



It's now been replaced by a discrete "heart icon" that appears on the "data card" once your shot has been finished. FYI if you steam milk and come back to the espresso tab, the heart icon is still there (for those of you who want to describe your milky drinks).

I'd love to get your thoughts on the "Your setup" page. I've not done any visual grouping yet (via photoshop) as I'm expecting some good suggestions in the next 24 hours...

[level3ninja]

I like how the Acaia app has what they call a "bean stash" where you can list beans you have on hand that you can then select from a list when you brew. I think that would be a good idea, also to add to that because since some people run multiple grinders it could be good to have some preset "setups" you can edit that store the grinder and beans (and other things like basket size used? target dose / yield? etc) to enable quickly switching between different groups of data commonly used together. Could this also be integrated with shot selector at the "before shot" menu down the track (probably PRO territory?).
The end result of this would be being able to pull consecutive shots on 3 different grinders running 3 different beans, having all that info automatically recorded by which shot program is run without having to remember to type in the grinder and bean every shot when it changes. Grinder setting could be carried over from last time and highlighted or have a question mark next to it until you tap to confirm or similar?

Which brings me to: more that you've freed up more space on the first screen (notes could shrink significantly) you could move a couple of things to the first screen, like grinder, setting and bean. Tapping on them could bring a pop up list to select from with a "new" option at the bottom that takes you to the relevant screen to add one, tapping done take you back etc.

[>decentespresso]

I like how the Acaia app has what they call a "bean stash" where you can list beans you have on hand that you can then select from a list when you brew. I think that would be a good idea, also to add to that because since some people run multiple grinders it could be good to have some preset "setups" you can edit that store the grinder and beans (and other things like basket size used? target dose / yield? etc) to enable quickly switching between different groups of data commonly used together. Could this also be integrated with shot selector at the "before shot" menu down the track (probably PRO territory?).The end result of this would be being able to pull consecutive shots on 3 different grinders running 3 different beans, having all that info automatically recorded by which shot program is run without having to remember to type in the grinder and bean every shot when it changes. Grinder setting could be carried over from last time and highlighted or have a question mark next to it until you tap to confirm or similar?

I really like this idea, and (for example) for a cafe this could be super useful as they're likely to have a decaf grinder, with a different extraction method being optimal for them than for the regular beans.

I will have to think a bit more about how to present this in the user interface, because currently the "describe your espresso" GUI is only available AFTER you make an espresso, whereas what you're describing is before.

However, the way the tablet software currently works, when you save a shot profile, the "your setup" currently in memory will get saved with the shot profile. In most cases, you won't want that but once your shot programs have stabilized, then you would want that. ie, have a "decaf shot profile" that also loads up the grinder, grind settings, dose, etc... hmmm...

Perhaps there needs to be a way to optionally save "your setup" when you save a shot profile.

Thanks for the suggestion, really helpful. No quick actions on my part, though, I need to cogitate a bit on this.

Which brings me to: more that you've freed up more space on the first screen (notes could shrink significantly) you could move a couple of things to the first screen, like grinder, setting and bean. Tapping on them could bring a pop up list to select from with a "new" option at the bottom that takes you to the relevant screen to add one, tapping done take you back etc.

Yes, I agree that the first screen has a lot of empty space (especially if you're not using a refractometer) and things could move around between the tabs. I'll probably leave that work until version 2, giving people time to actually use this software, gather feedback, and decide how this page needs to change.

[MrJack]

Correct: we know that our virtual flow rate numbers are not yet correct. I mentioned in a previous post, that this in fact was the reason for integrating a scale into the DE1+ at this time, in order to be able to fix the model.

The main mistake in our model seems to be over-estimating flow when there is pressure, and under-estimating it when there is not. Vibratory pumps move a differing amount of water depending on what they're pressure they're "fighting against".

In a few weeks, we'll have improved our virtual model, but for now we'll only be using to display quick, transient changes in flow (to help notice channeling), but using the hardware flow meter for accuracy of reporting flow.

Had a bit of a play about with the ulka pump curve, to see if I could predict actual inflow based on the virtual flowrate and the measured pressure.





Obviously not a perfect model (esp. given it's based on the pump curve for a different pump), but it's an interesting result. Quite sure my earlier thoughts about there accumulation of water in the puck during extraction were right - it's likely not a real effect.

I'm looking forward to applying the model to the "182 shots" data, as I have a feeling a lot of the strange behaviour will go away...

[>decentespresso]

Had a bit of a play about with the ulka pump curve, to see if I could predict actual inflow based on the virtual flowrate and the measured pressure.
Obviously not a perfect model (esp. given it's based on the pump curve for a different pump), but it's an interesting result. Quite sure my earlier thoughts about there accumulation of water in the puck during extraction were right - it's likely not a real effect.
I'm looking forward to applying the model to the "182 shots" data, as I have a feeling a lot of the strange behaviour will go away...

We've received a pump curve chart from ODE, for the pump we do use. Let me ask around here and see if I can get it to you.

And.... just received it from a coworker via whatsapp...

[MrJack]

We've received a pump curve chart from ODE, for the pump we do use. Let me ask around here and see if I can get it to you.

Great! Is it a curve for a single speed only?

Fairly straightforward to measure one yourself, provided you had access to a set of bluetooth scales, a pressure logging espresso machine and a portafilter with a needle valve attached... [emoji1]

[MrJack]

Putting the "low" pump curve through the same algorithm (with a few tweaks), and here are the results. The profile is not wildly different from the Ulka pump, but more linear.

A few details on my method:
1. Estimate the "relative speed" of the pump (essentially % of max speed), based on the virtual inflow.
2. Model the pump curve at various relative speeds (just using a simple polynomial expression).
3. Use multivariate ordinary least squares regression on the flowrate data to estimate the derated flowrate (dependent variable) from the measured pressure and estimated relative speed (independent variables).


The result is a predicted flowrate (in mL/s) that is slightly lower than the measured mass outlet flow.



Just for fun, I thought I'd subtract the water inflow from the mass outflow (i.e. assuming water density of 1 g / mL), to get an estimate of the extracted coffee mass, which I then used to estimate the TDS. It was right in the ballpark (and even follows a decay profile, as I would expect)...



I'm surprised by the high initial flowrate, but it's a pretty basic empirical model, and it could be an issue with how I estimate the relative speed. Could only be confident through validation against real data, to determine the pump curve profile (i.e. relative speed) that corresponds to a particular value from the virtual flowmeter.