newbie here,

first the questions:

-how advanced are the PID controllers used for these new consumer espresso machines? examples include the Diletta Mio from SeattleCoffeeGear, or the acosco line, I don't know all of the places to look for machines I have some common hits from google,,

- Speaking of temperature stability, there are some videos around discussing how thermoblocks are considered unstable, and generally inconsistent during water-flow , but that new PID controlled thermoblocks are much better. The Diletta Mio (which is one of the machines I'm considering) has a boiler for water, which is PID stabilized, and a thermoblock for steam, , which I don't care about at the moment, because I don't use it,, actually a selling point of the machine is a switch that if off, the steam just never powers up. ,,,,, I am inclined to like the idea of a boiler, because it is a thermal mass.

- on the other hand,,, if thermoblocks are the future, and the only real limitation is the specific stabilization loop (which once upgraded, might solve many lines to flat) ,, then it's a matter of some firmware updates to make thermoblocks fully superior to boilers, and shouldn't I just go that route? Faster warmup is pretty enticing.

So, I happen to have been interested in FPV drones for a while, and they are interesting because stabilization in drones saw an enormous amount of testing and development in the open-source community, meaning that people could use betaflight and have all sorts of experimental features, and importantly, make videos discussing them. The last bit that I paid attention to was when the Kalman Filter made it's way into the open-source stabilization offerings, and basically instantly dominated the racetracks, with all winners winning on some version of Kalman Filter stabilization. Then, shortly after this, Tesla executive Elon Musk made some press release or maybe several talking about how the Kalman Filter was also tested in his stabilization for self-driving cars ability to maintain steering and acceleration, and safety ratings went up significantly, with the Kalman Filter also being credited for making things like object avoidance and object tracking better, more capable, more safe. ,,,,,,, However, boilers are inherently stable, because they are a large thermal mass. This means that the dynamics of maintaining a constant temperature in a large boiler are much more forgiving than the dynamics of maintaining the balance of a quadcopter or the specific line of vehicle travel after a pothole or other. The boiler has so much mass, that it just heats slowly, and cools slowly, so the extreme advances that make drones so much better might not have made it into the same stabilization loops that power PID espresso machines. ,,,, However,,, thermoblocks in inexpensive home-machines are not big, stable, thermal masses,, instead, they would require very exact heating patterns to maintain a constant temperature as water flows through the thermoblock. It would need to burst power for short bursts, ramp power up and down, all sorts of things within very short moments in order to be properly constant in it's result. Also, like drones, there would need to be an amount of predictive overcompensation, which although basic PID tuning offers, generally speaking it is a dynamic that changes based on conditions, and the same tuning that works on first powerup of a cool morning is not going to be the tuning that results in a good party with friends. ,,, Here we need a much more complicated stabilization loop more similar to what's in racing drones.

Does anyone know about the Diletta Mio? It looks like a quickmill based on reddit posters, and having looked at quickmill (sold also by wholelattelove, the primary competitor to SeattleCoffeeGear in my google ad competition) and decided they are acceptable purchases,,, I am interested in the idea that the Mio offers a boiler while they will still be better for a while yet, , and until open-source firmware updates are something that my machine regularly can expect, I should go for something like that,,, and when the real future of thermoblock stabilization hits, I can expect that nothing on the market today will really be viable?? The acosco machines are intriguing, and honestly it should be as simple as an upgraded pcb,,,,, but then the heating elements in the thermoblock should probably be more aimed at burst heating rather than steady heating,, and therefore the power supply might should be different,,, in all,,, high performance heating elements aren't probably what any machines are rocking, because longevity and stability are the name of the game.

I will receive my eureka mignon notte in a week or so, I bought from seattlecoffeegear, it's super simplistic but looks capable enough for the tinker oriented, and I will be running it on my delongi dedica ec680 with the bottomless portafilter accessories from amazon, so I will not be on the market for some months,, I'm really hoping to learn more about specifically what sort of engineering is in the PID espresso machines, and if anyone can discuss the capability of todays vs. the sort of things I brought up above. (I mean,, it seems like a simple thing to make heat,, ,but on the other hand espresso machines are thousands of dollars simply because it's not easy).

The control of temperature is harder than it appears. One issue is that a typical thermoblock has a greater time constant than the speed you need it to respond (a second or less). They are also very "unpredictable" in that you're typically applying power, so the temperature gain depends on flow rate. The only commercially available, boiler-less machine that I consider "competition" for the better DB and close-coupled group machines doesn't try to manage the "thermoblock", but utilizes a hot/cold mixer. It isn't in the under $2,000 price class, by a long shot now.

As to what is in most espresso-machine PIDs, they're typically discrete-time, classic, single-loop PIDs with a way to manage warm-up so you don't get integrator wind-up. Not very sophisticated. Especially with the mass of an E61 and a time constant I'd guess at around 10 minutes, there isn't a lot you can do to improve things without several sensors and a complex algorithm.

So, I have some experience with FPV quadcopters, and watched some degree of PID evolution (from the know-nothing perspective of course!)

Quadcopters continuously benefited from getting lighter as PID stabilization advanced, and PID stabilization advanced quite a bit with open-source quadcopters, there are all sorts of considerations including microcontroller performance, tuning, and of course the use of the Kalman Filter and other filters.

Thermoblocks have a bad reputation based on my interpretation of reviews, and the impression that I get is that they are difficult to have consistency with, and also struggle to have even temperature during dispense. Thermoblocks strike me as having the basic engineering situation of: bigger heavier thermoblocks can heat more water before losing temperature, but are then big and heavy and slow.

If the heating elements were high performance, and able to rapidly heat and rapidly cool, and the PID controller were able to detect temperature accurately, probably in more than one place in the thermoblock, then the flow of water could cool the block ,and the active heating could heat the block, and then in this case, with increasing coil performance and loop performance you could have a very lightweight thermoblock and actually quite a bit of temperature stability,,,,, PROBABLY running from a modern day high performance racing flight controller, probably running from a modern day high performance version of betaflight, (free and open-source), it would run amazingly.

Can anyone comment on the validity of these ideas? I expect the espresso machine market to move slowly, and since I'm planning to buy my first "recommended" machine (I have a delongi dedica, but want an upgrade), I am looking at a boiler machine. I am looking at Either a Rancilio Silvia and to simply go with classic, and expect a learning curve,,, or I am looking for PID controlled boiler, pressure adjustment, and basically should mean smaller learning curve. (The SeattleCoffeeGear Diletta Mio is the machine that looks the best to me at present, but I'm concerned the microcontroller performance will not make me happy, and I'm concerned that a vibration pump at all will not make me happy. ,,, in this case, I should save money, buy the silvia, and just save up for a rotary vane pump,,,, but then I don't really have the space. So in this case, I should investigate other ways to stabilize the pressure)

In another conversation, I supposed that perhaps installing a second vibration pump, wiring it up backwards (there is a diode in the pump, so switching the wires on one would mean they would be like opposite pistons of a two piston cycle), and having it so that the two pumps are in parallel after the flowmeter, but before the OPV. Ideally, then, I would see more constant pressure and perhaps be able to use lower-powered springs so that instead of pumping at 15bars,, the spikes would be lower, meaning the OPV would work less (so maybe the 9 bar spring mod for the gaggia would work in it, just use them on both pumps),,, and it would be a sort of a workaround that lets me use a smaller machine, a small reservoir, something that takes up humble space, and importantly costs significantly less than a rotary machine. ,,, I did also consider just cutting the back out so I can somehow mount a rotary pump in whatever machine,, has anyone done this?? an extra vibe pump is like 30$ though.

Jeff wrote:One issue is that a typical thermoblock has a greater time constant than the speed you need it to respond (a second or less).

I did a measurement and found the same (step is 15% max heating for the first 250s, barely visible on the diagram). Really bad characteristics for a PID:


Playing around with the PID, I found due to the high time constant another approach was better. I used a fit for a second order model with delay and could control the thermoblock quite nice (<3 minutes from room temperature to ready):


Nevertheless, it was also remarkable that it is quite ok to control the water temperature of the thermoblock, as long as you know the flow. This was very long "shot" with variable flow. The fast fluctuations in the middle can probably be ignored, they will be mixed in the shower screen to the correct temperature. So I think with proper tuning (as you can see, my proportional constant was a bit so low) a 1°C error margin could be met: