Hi Jake

I'm sure there is still lot of room to improve, yes. And I will continue testing with prop. valve.
This was just a kind of "fast prototype".

Settup was:
watertank -> Ulka vibra EP5 (Type E) -> Prop. Valve -> Ulka nonreturn bypass valve -> OPV -> Boiler


Probably the Ulka valve is the problem in line. I guess it needs a certain flow to close the air relief (airbleeder) valve. And when it closes the flow is to high for extraction. This might shows this hysteresis of flow / non-flow.
And I don't know, if I need the Ulka valve because of back pressure (safe the pump).
And after cool down of the single boiler, it may recreates atmospheric pressure (vacuum)... guess not neccesary..

May the OPV also acts as an back-pressure valve.


Does anyone know the exact function of this parts?
I think I really have to test part-by-part...

Here my "technical expertise":

Ulka-airbleeder:
. Works as a back-pressure valve
. Recreates atmospheric pressure in a single-boiler-dual-use after switch off
. Needs a certain flow to close the air-relief-valve (I did not measure)
Means, for low Flow or Pressure it is not a good option. At very low water rate, it just fills up your water-tank by return line.

The OPV acts only as on OPV. No back-pressure valve inside.

The proportional valve does not work as a back-pressure valve.


I did test following setting:
Water tank -> Topsflo -> OPV -> Prop. valve -> back-pressure valve -> (Boiler)


Without boiler:
The flow rate with only the pump can be adjustet low. When I narrow the pipe eith the prop. valv in addition... I can adjust down to drops.
It is a play between pump and prop-valve. What makes most sense... I don't know yet.


With boiler:
At first, recreating atmospheric pressure can be done over steam line. no worries about that.
With prop. valve open 50% and low pump speed I could reach a flow of 20ml/min.
But the settling of the flow after opening the solenoid valve is not very stable. This might be caused by the single boiler. I should test with a HX, hence without boiler, the flow could be adjustet very accurate and easy.


Next step ist to get used to this many parameters.
Handling a machine with one button is easier than one with additional two potmeters.
...cognitive abilities

Chapter 16: The calm before the storm

It's nearing the end of February, 2018. I've completed the design of my needle valve integration into the group of my S20 and have a few sketches on how to manipulate the valve during a shot for pressure profiling, but the exact method will be determined by the actual sweep of the shaft needed on the valve to attain the results I'm after. If the needed sweep is 120 degrees or less, I can simply affix a lever arm onto the shaft, manufacture an upper group cover with a corresponding slot in it for the lever to poke through and attach an attractive paddle handle on the lever. This would be a good outcome.

If substantially more or less sweep is required, I will need to devise a linkage that translates the available range of paddle motion into the appropriate twist of the control shaft on the needle valve. I've done all the math I can without causing permenant brain damage and I should be really close to 120 degrees with one of the two valves I've selected, but the one thing I've learned over the years is that benchtop engineering calculations are a great way to precisely determine what an actual outcome won't be.

So instead I will order my parts, install the needle valve and figure out what the real world tells me with respect to how it's all going to work out and what my interface will be. But here I am, two months after getting my parts list together and I still haven't ordered my parts. Why am I stalling?
Two reasons:

First, I am annoyed that I will be spending 40% of the value of my shopping cart on shipping to get my parts. What the heck? I've been searching (and finding no luck at all) for online coupons, available local distributors, etc... where I can knock a few bucks off of the total price of all this stuff. Yeah. I'm cheap.

Second, I am surprisingly satisfied with the current state of the machine. Line pressure preinfusion has opened up a world of consistently great shots with a depth of flavor I had not previously been able to get. I constructed a homemade puck grooming tool to help with getting even distribution, as the low headspace design of my S20 necessitates that I dose well beneath the rim of my 18g basket. Dosing anything more than ~17.5g results in hyper-compression of the puck by the shower screen and severe channeling.

Below are two back back shots I pulled this morning. Even though they both pulled short at 17.2g in and 31g out in 32-33s including 15-16s of preinfusion before turning on the pump, they both tasted great and the distribution and extraction were rock-solid. I would have tightened the grind further and pulled a third shot but I drank both of these after my morning Cappa and didn't really feel like dying.

First shot: And second: Here's a quick look at the result of the first shot above:

Life is good...

So, I'm spending some time to get to know that machine in its current state before moving along with the surgery. I'm fully committed to seeing it through, but I think there is sound logic in getting a firm grasp of the machine as it is now so that I understand what each stage of the modifications are bringing to the table on their own before I have them all available at once. I've also been doing some thinking about how to better characterize the puck with respect to available pressure and compliance to flow at various stages throughout the pull. Once I have my thoughts put together, I'll publish part 3 of "What's going on there, anyway?" and see if we can't drive further discussion into the dynamics of our machines and firm up a better understanding of what it takes to get certain results we might be after...

Cheers!

- Jake

Looking good Jake!

You're inspiring me to film a few videos of my Silvia in action, too.
I've got two distinct pressure profiles setup for it now...

Chapter 17: What's going on in there, anyway? Part 3

So, with all the discussions surrounding Part 1 and Part 2 and the subsequent conversations regarding the measurement of flow rate and pressure, the voltage and current analogy, and the compliance of the pump to deliver a given pressure to the system, we haven't spent much time discussing what is generally considered to be the overall governing factor determining the relationship between pressure and flow when pulling a shot:
The Puck

Let's talk about that. OPVs are great. Needle valves? Neat stuff. Mains pressure can preinfuse. Pumps can pump. But literally beneath it all lies the elusive puck. Loosely defined, it's a given dose of coffee, ground to some degree, distributed in a basket and tamped to some extent. Want less flow or a longer shot? Tighten the grind or increase the dose or both. Want more flow or a shorter shot? Loosen the grind or decrease the dose or both. Piece of cake, right?

Well, sort of.

Here's what gets me: We generally think of a shot in terms of averages. In general, we want 30-ish grams of beverage dispensed in 30-ish seconds at 9-ish bar. This is beat into us from our espresso infancy. So I would be tempted to describe a "good" puck as being this thing "defined" above that gives us a good tasting shot under those parameters. Going back to our electricity analogy and Ohm's law, V=I*R we get 9bar (V)=1g/s(I)*R. So, the "resistance" of the puck is equal to the applied pressure divided by the flow rate. This gives us "puck resistance" units of bar/g/s.

Ok. So what? A "good" puck should have a "puck resistance" of 9bar/g/s, where g is the mass of the water (30g) and s is the shot time (30s). Right?

Right-ish.

The ish comes from the fact that these are based on averages and we all know that the puck compresses at the beginning of a shot and erodes as the shot progresses. As such, the "puck resistance" starts high and falls over time. Also, the beginning of the shot is a hectic time for life on the puck. I've often thought of the shot progression as a relatively simple 3 step process wherein first the water fills the head space, then the water pushes the air out of (saturates) the puck and then espresso floweth forth from the bottom of the puck. Nice. Linear. Easy.
Wrong.

I'm going to base much of the rest of this opinion piece from data and discussions plucked from others. Particular credit is due to Assaf, Dick Green, John Buckman of Decent, and the Slayer team (for writing a patent and excellent owner's manual that can be read by all...). Not that others aren't to be credited for driving the concepts, but these are the sources I've most relied on to inspire my own curiosity and get some wheels turning in my head.

Before I can get too much further, I need to define some terms I'll be using so that my head hurts less as I piece this all together. I'm up for challenge on anything I get wrong and I'm happy to make corrections to this post to migrate towards existing conventions of terminology, wherever I may stray from them...

Here goes:

Water Debit: The rate of water flow from the group with no puck slowing it down. This is generally adjusted with a change in gicleur (needle valve adjustment in Slayer), mains pressure or pump pressure.

Puck Compliance: The rate of water flow through a puck at a given pressure at a given point in the puck's life. This is effectively the instantaneous flow rate of the shot once flow is established and the rate at which the puck absorbs water before flow is established.

Puck Saturation: The extent to which a puck has absorbed all the water it can. This value can only be measured once flow is established by comparing the flow into the puck with the flow out of the puck. A puck is 100% saturated when every gram of water added to the top of the puck results in a gram of water (plus some dissolved solids) exiting the portafilter.

Puck Capacity: The mass of water that a puck can hold at 100% saturation. Experimental results of others suggest that this tops out around twice the weight of the coffee in the basket. It is somewhat difficult to measure accurately...

Puck Resistance: This is simply the inlet pressure divided by the puck compliance. Puck resistance has units of bar per gram per second.

Puck Compression: A puck is considered compressed when it begins building pressure against the water debit to which it is subjected. The rate of puck compression has a dramatic impact on puck resistance.

Alright. We have our terms defined. Let's start an imaginary shot in slow motion. All the pictures below are based on this imaginary shot. Good old excel, no data acquisition. Not real. Let's just get that out there right now... Through whatever process that is currently in use, water enters the group, is dispersed by the jet breaker/shower screen and begins falling through the headspace and lands on the surface of our puck. The initial flow rate at this point is equal to the water debit of the machine. The puck, which is totally dry until these first drops start pummeling it, starts drinking up. More drops fall, and the puck keeps on drinking until drops are falling faster than it can drink them.

At this point, water begins accumulating on top of the puck. I had always envisioned that the water comes down like a piston, pushing the air out of the headspace through the puck, but in this slow motion shot, it seems readily apparent that the rain drops are falling through the air gap between the shower screen and the puck and that once a pool of water accumulates on top of the puck, the air begins to be compressed at the top of the group. As more drops fall, this compression continues, and a pressure gauge mounted in the group would start registering pressure. In the picture below, representing a 15 second preinfusion at 3 bar on my S20 with a 0.5mm gicluer, this starts happening after 2 seconds, and the pressure rises until supply pressure is reached as long as the supply flow is greater than the puck compliance.


As more and more water rains down on the now saturated top layer of the puck (the bottom of the puck may very well still be dry), the headspace becomes more and more pressurized and the flow rate decreases until some point at which we consider the puck to be compressed. I'm foggy on when this point is, but it's an easy mental exercise to envision the point when a Slayer hits 9 bar in pre brew to be a good representation of a puck becoming fully compressed. By this point, the headspace is all but flooded with water that is at "full" pressure. On the picture above, it is obvious that the puck is as compressed as 3 bar can compress it at 12 seconds. It should go without saying that puck compression is a bit of a continuum; when I kick the pump on at 15 seconds, I guarantee you the puck gets compressed more than it is at 12 seconds with 3 bar line pressure pushing down on it. Flow through the puck is at its minimum right when the puck is compressed, as shown from around 12 to 18 seconds below:


Even with the flow rate at its minimum, the water does keep moving through the puck, with much of it becoming absorbed by the ground coffee. An important realization made from comparing gravimetric dosing curves against calibrated flow meters (by others) is that the shot can (and generally does) start flowing into the cup long before the puck is 100% saturated. That is, some water passes through the puck, taking dissolved solids with it while simultaneously some water is retained and accumulates in the puck. It's not necessarily linear... This process continues on until the puck becomes completely saturated, and the flow meter reading water into the puck comes in sync with the weight of espresso falling into the cup. The picture above shows first drops landing in the cup at a slow rate starting at 16s. At 30s, the flow into the puck equals the flow into the cup. This is the point of 100% puck saturation. This point of complete saturation could occur at any time between the start of the shot and the end, with evidence supporting that preinfusion tends to saturate the puck earlier in the shot than a quick ramp up to full pressure. The picture below shows accumulated volume from the machine, into the puck and into the cup. Note that the puck tops out at 30g at 30s for our imaginary shot, which I designed with a 15g dose so the math would work out nicely...

So, why am I talking about any of this?

You may remember that I have plans to replace the gicleur in my Rancilio S20 MIDI CD with a needle valve that I can adjust on the fly. My early goal was to be able to replicate the slow pre brew of a Slayer and then simply open the needle valve to go wide open, rather than needing an extra solenoid valve and all that jazz. Then it dawned on me that I could potentially achieve full MP style paddle-controlled flow profiling of the shot if I played my cards right. But achieving full control involves knowing what to expect during the shot and how to respond.

The first "ah ha" I got from this mental exercise was that it takes not 30g of water but 60g to extract a 15g dose at 1:2 brew ratio, assuming the puck absorbs twice its weight in water during the shot. This may very well explain in large part why slow preinfusion results in faster shots and finer grinds; when you get 50% of the flow taken care of before any drips hit the cup, it's pretty reasonable to expect the other half to go much quicker than when you do them both at the same time... Go back to the beginning with the puck resistance calculation... for 60g in 30 seconds, you get 9bar/60g/30s, which is 4.5bar/g/s. Apply that resistance to the second half of a long preinfusion shot and you get a fast shot, indeed... finer grind, anyone?

I have much more rambling to add to this post, but it's pretty darn long already, so let's see if anyone even cares before I keep going

Cheers!

- Jake

Jake - I am not sure about the flow rate. This is from the Digmesa sent by Piezo - taken from this morning of an Ethiopean. Oh - as of this morning Chimera has telemetry thanks to Tobias' showing me how to add an ESP8266 to Chimera (See more on the Chimera thread).

Cyan is Flow Rate (ml/min), Blue pressure (bars), Grey is shot weight.


It looks as if flow rate is constant - but I think the flow rates during PI and the pull just happen to be that way (as flow during PI is inhibited by the Digmesa Flow Meter - 1mm, Gicar Flow Meter - 1.15mm, HX gicleur (?mm), Needle Valve (?mm) and "Gicleur" (0.6mm)). But during the pull and erosion - it seems fairly constant (perhaps a bit of a slant of 5-10ml/min over a 45 second period... @ an oscillating 9 bar).

HTH. Or maybe HTDH?

Something looks amis with your flow data, though. On top of being really noisy, you've got 45g in the cup, 45 (60?)second shot, but an average flow rate of 15ml/min.

As a sanity check, just take the derivative of your scale reading and this would be the flow rate into the cup. Looks like you have an average flow rate of 85ml/45 (60?) seconds (assuming 20g shot, 40 grams water to saturate and 45g in the cup).

Since you're using the FLB, and pre-infusing to 9 bar vs me using an "old school" approach, I could see your flow rates being a tad more steady than what I have but even so... what comes out must have gone in, no?

Cheers!

- Jake

**edit**
While I don't "like" the noise or scale of your flow meter, (picking up pulses?) I do find very interesting the clear puck compression that occurs after the peak pressure, but it rebounds so quickly it's hard to know if that isn't an artifact of removing the FLB from the circuit...

**edit #2**
Just read your update to Chimera. Noisy flow meter may vary well be an overactive PID loop, eh?
Drop you k_p way down and kill k_d and k_i until you get a nice coil spring oscillation output and then temper that by adding k_d to limit overshoot. You will then have a steady state error that you can nix with k_i. Then it's iterative to bring k_p up, while keeping the response in check with the other two until you're satisfied with the response speed.

This is the same pull - except it shows the pulses from the Digmesa flow meter. Each pulse is (according to the datasheet) 0.025ml.



4036 pulses means 4036[pulse] x 0.025 [ml/pulse] = 100mL overall volume. Obviously my ml/min calculation is - ahem - off....

I can't find why this is off: flowRate = (float)mlPerFlowMeterPulse * 60.0 * 1000.0 / (float)g_averageF.mean()
But since there are 20x pulses than the Gicar - I may be missing pulses in the interrupt. Which would explain the noise as well. Gicar needs 1/T to get resolution. With Digmesa I have enough pulses and should be doing num_of_pulses/time. Time for some more coding....

Shot weight 45.9 grams and dose was 17.8 grams.

Which means that 54 grams of water stayed in the puck. Or for each gram of coffee - 3 grams of water. Or it may mean that the calibration is off or non linear. One of the reasons for the telemetry is trying to figure out these numbers.

Aren't you glad you keep me around to bug you?
Check my edit #2 above your post for notes on PID tuning. I've used this process many a time to get good results on many different systems.

Cheers!

- Jake

Indeed - running the pulse numbers sans the interrupt based 1/t calculation yields the following:



I am not sure I know where the noise comes from. Accuracy on the Gicar came from capturing millis in the flowmeter interrupt. But since there are too many interrupts we are losing cycles (hence i/t is now wrong). Worse, the 500mSec time between cycles isn't timer driven. Hence the sampling point wanders a bit.

So I think I need to program one of the Arduino's timers and run the entire system off a 250mSec or so clock.

Oh - and Jake - it sort of does prove your analysis... sort-of-ish.