Once again, you nailed it.

AssafL wrote:Okay. So at first I was confused. And then - Ahhh....

So a bit of disambiguation for your story:
OPV1 - The OPV that you are referring to is the recirculating OPV in the pump (or otherwise). It is typically set to open at around 9 bar (feinschmekers set it at 8 or 8.5 bars). It can expel water back to the reservoir, down the sink, or back to the mains.

Note that that setting a lower brew pressure requires the OPV1, (I like that terminology...) to bypass even more flow to the tank because the pump must provide more flow at the lower pressure, by design.

AsaafL wrote: If the recirculating OPV is against mains, one would set it against mains pressure (so if mains is 3 bar, the pump would add 6 bar). That is the main reason to add a regulator before the espresso machine.

I totally agree. Also note that if you were to apply 3 bar to the inlet of the Ulka Pump, you would reference 6 Bar on the pump curve to understand how much flow the system must allow in order for the pump to operate at that point. Suddenly, we have 300-400cc/min to deal with when all we want is 120cc/min for a normale, and maybe only 60-80cc/min for a ristretto. Our recirculating OPV is going to be BUSY. Especially since this guy has to not only flow more through it than a pour over OPV would, but it has to dump it back to the mains, which is pressurized to 3 bar. The recirculating OPV (generally referred to simply as the bypass on a rotary pump) has a very active life, indeed.

AsaafL wrote: Pump graphs: These are okay for constant voltage pumps. if one is to use a variable speed motor with a pump, the graph is free to move up and down and becomes the maximum flow per pressure (or maximum pressure at a flow rate).

Point taken. By design, the Ulka runs at 50% duty cycle with constant voltage applied, with the large diode built into the coil cutting off half the 60Hz (or 50Hz, outside the US) waveform, allowing the spring in the pump to pump the water. Not sure how many folks already knew this, but the power you supply to the pump doesn't do the pumping. Just like a lever machine, where releasing the lever, allows the shot to flow, the power compresses a spring and pulls water into the pump and the spring generates the pressure as it pushes the piston (and water) towards the puck. The diode cutting the current is the only way the piston is allowed to move at all. Without the diode, the pumps just buzz, helplessly. Below shows what would happen if you were to somehow decrease the duty cycle of the same Ulka vibe pump (I think the most effective way to do this would be to change the frequency rather than the duty cycle, since they run on AC, but my point remains the same...).

Rotary pumps behave in much the same way when puting them on a variable speed drive. Important thing to note here is that it is the system, that dictates what the back pressure will be on the pump for a given flow rate through the system. The pump then operates on its curve at the system pressure. It is the job of the recirculating OPV (or bypass valve) to deal with any additional flow that the system cannot handle at the desired flow. Pressure-profiling machines can (and do) shift the pump curve to match the demand of the machine, to the extent that they can.

AssafL wrote: The moral is that there are many layouts for Espresso machine hydraulics. What reads well for one machine may make little or even no sense for the owner of another machine. The common denominator is that levers, pumps, orifices, gicleurs, FLB/metering valves, springs, PWM controllers, gears, puck headspace & timers are all there to profile the introduction of water to the puck. Without them pulling an espresso becomes a very finicky system...

Bingo. And consequently with them and all the variations therein, trying to predict the outcome of one machine based on the experiences of another machine is very nearly impossible...

I'm glad you find my ramblings a good read! I am having a very good time putting these thoughts together and have been blown away by the warm welcome I've received from the HB Community. Thank you all for contributing to such a fantastic place. Hopefully, there will be a few more "ahhh..." moments as the story progresses...

Cheers!

- Jake

Chapter 9: What's going on in there, anyway. Part 2

Ok, quick recap: We have an espresso machine with a bunch of stuff in it. In the case of Junior, we have a reservoir, followed by a pump, a recirculating OPV (which was the only component really mentioned thus far as "the system") and a host of other downstream bits. We've established that the pump must operate at a point on its curve and that the system is in charge of determining the actual pressure at which the pump will operate. Remember that the system as a whole is everything downstream of the pump. Also remember that it is currently early September of 2017 and I just sold Junior and have been reading and learning and testing as much as I can to try and understand why the S20 behaves so differently and what I can do to try and emulate the flow characteristics of Junior in the big guy.

It's easy to oversimplify things and just imagine that the pump pressurizes the entire espresso machine up to 9 bar and delicious espresso floweth forth. Of course, it doesn't actually work like that at all. Below is an image borrowed from this thread and edited by yours truly.


Starting at the pump:
P0 is the inlet pressure. It might be 0 bar on a gauge for a pour over machine, it might be 3bar or higher on a plumbed machine.
P1 is the pump outlet pressure. It is limited to be no higher than what the recirculating OPV (or bypass on a plumbed rotary) is set at, but could logically be any value less than that, depending on what the flow rate of the rest of the system is. It also doesn't drop much at all through the system until it gets into the group, aside from a nominal amount to open the check valve.
P2 is a new pressure for us, and the topic of this post...

P2 is defined by 2 criteria. First, it is defined as P1 (pump outlet pressure) minus the pressure drop across the gicleur. Second, it is defined as the pressure at the puck, which is defined as the pressure drop across the puck. These two pressures are equal by definition. In the case of rinsing the screen with no portafilter, P2 is arguably zero bar (there is some miniscule pressure drop through the mushroom, brew valve and passages but I maintain that this is negligible.) The other pressure I have annotated is P2i. This is an odd duck, but represents the cracking pressure of the preinfusion pressure relief valve found in E61 groups. P2 is limited to be no higher than P2i until the preinfusion chamber is full. Based on the text of Valente's original patent, the preinfusion valve is designed to lift at a pressure of 1.5 bar. However, most home users report that their brew pressure gauge shows preinfusion at 4 bar. What gives?

(Note that while manufactures may have played with the spring constants and chamber volumes over the years and that all my assertions below may be inaccurate in their details that the concepts and principles are still directionally sound and valid. The numbers themselves are inconsequential, it is the relationship between them that matters.)

Behold the power of the gicleur... If we were to let the pump run unrestricted at the preinfusion relief valve's 1.5 bar, we would have a lot of flow and the preinfusion chamber would fill up quicker than we would like. At 1.5 bar, we're seeing around 550cc/min on our trustee Ulka Pump curve. Thats just shy of 10cc per second and if we had a 30cc chamber (I have no idea what it actually is) it would be full in just over 3 seconds.

The gicleur placed at the top of the mushroom restricts the flow from the pump and creates a pressure drop of about 2.5 bar across it when doing so. 1.5 bar for the preinfusion valve, plus the 2.5 bar pressure drop across the gicleur gives us the 4 bar that folks see on their brew gauges while the "alternately seating valves" do their thing. (Note on the picture that the gauge is reading P1) This results in the pump operating on a different point on its curve while the PI chamber is filling:

So, 425cc/minute flow through the gicleur when P1 is at 4 bar, creating a pressure drop of 2.5 bar, so that P2 is 1.5 bar while the PI chamber is filling. Our 30cc imaginary pre infusion chamber would now take just shy of 4 seconds to fill, once P2 has reached 1.5 bar, and lifted the preinfusion valve.

Great. What does this mean and why do I care?

Well, remember how I said a while ago that for a given system, the pressure drop through that system is proportionate the flow through the same system? Probably not. Thats ok. Well, we all know that we, as the barista, design the puck to deliver 120cc/min or less when the pressure drop across the puck is 9 bar or so. While the relationship between flow and pressure drop across the puck is not linear, due to the non-homogeneous nature of a coffee puck, it is relatively safe to say that the flow through the puck is much less at 1.5 bar than it is at 9 bar. So, some portion of that 425cc/min flow, coming through the gliceur that is filling the PI chamber is slowly permeating the puck, and we have good reason to deduce that this portion of flow is decidedly less than the 120cc/min of a full extraction. I suspect this flow through the puck during preinfusion is in the 40-50cc/min range, but I haven't found an accurate method of measurement to back that claim yet. However, this estimate does pass the sniff test in terms of how long it would take to fill a 30cc chamber with 375cc/min flowing into the chamber, looking at various pressure ramp graphs of E61 machines. Obviously, armchair engineering never actually gets you accurate results, just directionally valid concepts that can be tested...

So, what happens when the PI chamber fills up? At that point, the flow through the puck must rapidly increase since it no longer has anywhere else to go. The pressure at the puck starts climbing, due to the increased flow and the whole "flow vs pressure drop" relationship until it reaches the point where the system flow and pressure reach equlibrium. Typically, this means that the flow and pressure rise until the recirculating OPV opens. When the pressure stabilizes, the spring in the PI chamber pushes the valve closed, but that doesn't really influence the extraction. But what about the gliceur? It's still causing a pressure drop between P1 and P2 in our diagram, right? How much pressure are we losing at the puck? Well, since the gliceur is a true orifice, it's pressure drop is directly proportional to the square of flow through it. So, what was a 2.5 bar pressure drop at 425cc/min is only 0.2 bar pressure drop at 120cc/min. Basically, not a big deal.

So, as Valente noted in the original E61 patent, the gliceur, coupled with the volume of the PI chamber, and the opening pressure of the the PI valve, all work together to give the barista repeatable control over the preinfusion of the grounds with his design of alternately seating valves. The "barista control" portion on the original eclipse was a needle valve, but in subsequent variants, the needle valve was replaced with a simple fixed orifice to provide a nice gentle startup curve when the pump kicks in. Furthermore, once the flow is established, the gliceur has hardly any measurable effect on the overall flow of the system.

Next up? Let's see how the Rancilio hydraulic circuit differs from Junior...

Chapter 10: Unexpected results

So I'm working hard on chapter 10 right now, getting all of my facts and figures in order, getting ready to tell you how the 200lph pump in my S20 is VASTLY different from the Ulka we all know and love when I decide to do something idiotic and take some measurements to lend my post some credibility.

Below is the chart I put together, ready to discuss; as this is what I "learned" was in my machine back in October... The Procon data is corrected for 1360 RPM and translated to cc/min and bar from this site.

Please disregard the typo. The pump should be a Procon 1510, based on everything I've read. So I hook a short hose up to my spouted portafilter with no basket on my right group, which has no gicleur, no jet breaker and no shower screen, and time a 60 second purge of the pump to witness engineering majesty as 2700cc flows from the group like clockwork.

Except instead of that, I got 771 grams of 63 degree water in 60 seconds. Which is Not what I expected. The machine was cold, so I did expect the temperature, but I'm missing most of my water...

So, here I am, delaying any real update on my topic to chat about the little things in life, like being really wrong and being mostly ok with it. The big question is what pump did Rancilio put in there? This flow is actually less than what you would expect from a 15gph Procon, which the above referenced link suggests should be around 945cc/min, wide open with no back pressure. According to the Procon chart, the 1521 pump would need over 85 psi pressure differential across the pump to drop the flow down to 771cc/min, (or 15.5gph). What I can tell you from the Procon charts is that their output is a function of pressure differential across the pump and appears very predictable at 1gph lost for ever additional 25psi across the pump. This translates to ~36.6cc/min less flow for every additional bar of pressure without the bypass valve doing a thing. I suppose it is entirely possible that this is the flow profile of a 24 year old 1521 pump, and I am totally good with that if that is the case. Not what I expected but actually not a negative thing at all. I need some more instrumentation if I'm ever going figure out what is actually going on in there.

Alright,

Back on task. I'm on a mission, it's now October of 2017 and I have a fairly decent grasp of what's going on in a vibe-pump E61 group when you lift the brew lever. What did the engineers at Rancilio have in mind with the S20? Well, as I mentioned before, it's different. Very, very different.

First off, I have 4 bar inlet pressure on my procon pump (I know, I know, it's too high!). As soon as the brew switch is pressed and the solenoid opens, water is immediately motivated by 4 bar mains pressure to work it's way through the solenoid cavity and back up into the group. The pump hasn't even started yet and we're already more aggressive than the gentle ramp up of our Ulka pump and PI relief valve in Junior.

There's no preinfusion chamber here. The inside of the solenoid valve, the tiny conduits leading up to the group and the headspace between the screen and the puck is all you get. Furthermore, the 350W, 1360rpm motor has a massive starter capacitor and the motor is up to speed, now. So, whatever pump is in there, it's already riding the bypass valve the second I hit the brew switch and the forgiveness factor is basically non-existent.

If that wasn't enough, after reading Randy G's excellent EASY GUIDE TO BETTER ESPRESSO AT HOME, I realize that I am hopelessly over-dosing with the 14 gram baskets I ordered from Stefano no matter what I do. First step, I pull the jet breaker and get it machined to accept a flat head 6mm shower screen screw so I can get rid of the bolt head that has been fracturing my pucks for the last 4 years . I still can't pass the nickel test and accurately control my dose. Time to order scale so I can figure out how much is too much and get it right. Turns out my headspace is so minimized in my machine that (depending on the beans and roast level) I can only fit 13.5 to 13.8 grams in my 14g basket without hitting the screen when I lock the portafilter in. Any less than 13.5g, and my tamper can't reach the coffee due to the taper of the basket. I'm turning into a basket case...

With the scale on hand, I clean my grinder out spic and span and remove the finger guard and tape the doser paddles to sweep the doser clean. I determine that roughly 3.5 grams hang out in the chute after grinding. I learn to sweep the burrs from the top after all the beans are ground as the grinder stops turning, then pulse the grinder motor while I smack the double ridgeless basket that is covering the neck of the grinder with the palm of my hand. The combination of the pressure spike coming through the basket and the momentum of the rotating burr blasts the the chute mostly spotless. 0.2 grams still hang out and can be swept out if I really need them, but I usually leave them for Heidi . I find that once I finally get my dose appropriate for my basket and headspace, the grind is far more fine than I've ever had it before. All this time, I have been dosing by volume, scraping the grounds flush with the top of the basket using the doser lid. I'd tamp to about the mid point of the ridge and then lock it on. Come to find out, I'd been giving the puck a nice 300 pound tamp against the shower screen when I did this and the grind had to be loosened up to get a decent (that's being generous) pour. Things are slowly improving...

At this point, I am bound and determined to create a mechanical preinfusion device for a solenoid-actuated commercial group that utilizes the fundamental principles that guided Valente over 56 years ago as is able to handle the flow characteristics of a rotary pump. My search for prior art will lead me down the rabbit hole; to a place from which I may never fully return. The land of pressure profiling...

TomC wrote:Between October till now, I hope you've picked up some better baskets. VST, IMS or HQ from Espresso Parts. Having tapering baskets and not weighing each dose can be a real PITA as you've no doubt learned.

I have, learned. Believe me, I have learned... while I haven't picked up VST, IMS, or HQ baskets yet, I did pick up an updated 18g basket from Stefano along with a gicleur, stainless flat head shower bolt and a new group gasket. I love being able to go pick stuff up in person!

Not earth shattering, but a definite improvement from the heavily tapered 14g basket. I can fully tamp 16g in this basket with no issues and can pass the nickel test with 17+ grams. Not that up-dosing is an automatic win, but at least now I have options . My old ridgeless basket is a nightmare WRT hole distribution and consistency by comparison. Christmas is coming, so there may be a nice set of precision baskets in my future

I have been weighing every dose ever since the scale arrived and with my 18g basket I pulled a fantastic shot this morning with near perfect uniformity using a 17g dose with 6g of Finca Samaria Columbia and 11g Fazenda Pica Agudo Brazil, both expertly roasted at Equiano Coffee in Eugene. Okon knows his stuff...

AssafL wrote:As to your missing water, I am not acquainted with the Rancilio, but try to account for "spurious" gicleur-like orifices in the hydraulic path:
1. Flowmeters have real gicleurs in them
2. Valves (solenoid and others) have orifices and can be ordered with different sized orifices in them
3. Mixers have gicleurs in them

So even a degicleured machine is not really free flow...

This is really challenging my preconceived notions of what's going on when flushing. I had always assumed that the pressure at the pump outlet would be very low in absence of a puck, but my opinion has been changed but my findings and your well-founded observations/comments. I had to do some more research, but I came across this graphic, originally posted by Eric S here.



I believe the Procon data above assumes 1725 rpm, but even at 1360 rpm, it is clear that when flushing the group and getting 771cc/min flow, the bypass valve is active in reducing the volume sent to the group. A smaller gicleur results in more flow through the bypass and less flow through the group. I always knew the gliceur played a huge role in vibe pumps, since the output of the pump is greatly influenced by the back pressure, but I figured (mostly correctly) that the rotary would deliver the same volume regardless of the back pressure until the bypass opened up. What I totally didn't expect was that the bypass is open almost as soon as the pump starts turning . That said, it's entirely possible I do have the 200lph pump in my S20, and the bypass valve is sending nearly 2 liters per minute back to the pump inlet. I don't know why, but I'm really surprised by this... I shouldn't be, but I am.

Your next reply made me smile ear-to-ear

Chapter 11: Prior Art

Ok, it's now mid-October, 2017. I've decided that mechanical preinfusion is a must-have on my rotary pump-driven S20. Doing preliminary engineering nerd stuff, I try to determine the factory gicleur size, as it is a critical parameter for determining the volume of chamber needed to optimize the preinfusion duration. I find the parts diagrams for my S20.

There is something missing from the diagram... Item 7 is the "gicleur cap", but where is the gicleur? I search the group diagrams for other Rancilio machines and find a single part number for most other machines for a 0.5mm gicleur. No matter, just a typo on the parts diagram, I'm sure. Just to be sure, I pull off the top cover of my machine and see if I can get a part number off my the gicleur in my machine. I shut the water supply off, power down and pull the gliceur cap. Oh... nope. Not a typo. My machine has A THREADED 8MM HOLE instead of a gicleur... no beuno. Luckilly, Stefano has what I need and the price is right, but it will be another month before I figure that out...

Anyway, I've done some preliminary sketches of how to go about getting Faema-style preinfusion with a solenoid actuated group. Nothing fancy, but with the rearrangement of the water path, the simple spring-behind-a-valve technique of the E61 design won't work with a solenoid. The key to any PI "device" functioning is that it must be empty prior to starting the pump. Valente accomplished this by using the cam on the brew lever to unseat the PI valve and using the PI valve to open the drain valve. It really is a thing of simplistic beauty. With a solenoid controlling the exhaust of the group, emptying the chamber is proving to be a bit of a challenge. Even so, I come up with a functional schematic of how to do it. It will use a standard E61 valve and spring, oriented horizontally, attached to the group port of a sandwich plate between the group and the solenoid. Below is my first unfinished sketch of what I'm thinking of:
When the solenoid connects the pump to the group, the pressure will build to 1.5 bar (or whatever adjustable pressure I want) and unseat the valve, just like an E61. When the chamber is full, the pressure will climb up to 9 bar and extraction will begin in earnest. Here's the challenging part. When the solenoid dumps the group pressure to zero, the PI chamber is full of hot water and the relief valve is sealing it tight from the inside out. To solve for this, I locate a supplier of miniature ball check valves. There will be two. One will be located at the bottom of the chamber and allow the water in the chamber to bleed back through the sandwich plate to the exhaust through the the same port that pressurizes the chamber when the pump is on. The second will be located and the top of the chamber and will function as a vacuum breaker valve. The schematic has the check valves called out, but they are not located in the chamber sketches, as there will need to be some clever galeries machined into the chamber to get everything into a tidy package. The chamber will be located such that it will gravity drain through the solenoid. Neither miniature ball checks will have springs. They will be a simple SS ball sealing against a brass seat. They're cheap, reliable and come in metric threads. Awesome.

Now that I have an idea of what I'd like to do, it's time to do a search for prior art and see:

1) If it's already been done.

And

2) What else has been done instead?

This process is time consuming, and I get lost in the details, but let's look at a few of the highlights. Early on in my search, I come across this La Speziale preinfusion kit. I read on, I find patents for LM showing a similar chamber sitting above their saturated group. I find LSM machines that use nothing more than an a sealed chamber mounted above the group, and this CG topic covering water hammer arrestor preinfusion on a Silvia. Ok. Its been done. Ish. One thing consistent about all of these devices is that they utilize what i would call (I know this term is already loosely defined, but bear with me) progressive preinfusion. What I mean by this is that all of the devices listed above hit the preinfusion device with some amount of initial water flow and the pressure progressively rises as the chamber fills. Whether the water progressively compresses a spring as it does in the La Speziale and LM designs or progressively compresses a gas (air) as it does in the LSM and water hammer arrestor designs, the end result is a more gentle ramp up to full pressure while the chamber fills. And the spring, or compressed air forces the water out of the group when brewing is complete. KvdW uses this same design principle, so it must be good enough, right?

These designs all look totally doable and effective, but at this point in time, I'm still hung up on being able to hold a constant pressure for some amount of time and am less interested in these progressive approaches. I also familiarize myself with the E61 middle position line pressure preinfusion techniques and the solenoid equivalents used by LM and others. When reading the Silvia water hammer topic in more detail I stumble upon this post and discover for the first time a little espresso company recently acquired by Gruppo Cimbali.

Crap.

Down the rabbit hole I go. I find this topic and get sucked in. Badly. It immediately pulls me here and I spend the better part of a week trying to wrap my brain around all this not new stuff that seems really cool and new. While plotting my final plans (at this point, I'm all about doing this with pressure/flow profiling, sorry Valente ), this post catches my eye, and the following ensues towards the end of October... Don't worry, I know this a terrible shot from most standards, but this is the FIRST TIME EVER I've prepared a shot that I wanted to drink all by itself. This is kind of a big deal... If you open the video in youtube you can read my notes of what's going on here. Remember, I dont have a scale yet, and my group doesn't have a gicleur...

I think I have one more chapter of this story before we're all caught up and then I will launch my modification topic in the repairs area. I'm getting excited...