'Pressure Profiling' With The Fluid-O-Tech TMFR Pump - Or, Wholesale Copying Greg Scace's Ideas

Need help with equipment usage or want to share your latest discovery?
User avatar
shadowfax

#1: Post by shadowfax »

0. Introduction
For some time, I've been intrigued by the idea of 'pressure profiling' espresso shots, no doubt on account of my sheepish nature and the hype that's surrounding the Slayer and the upcoming La Marzocco profiling machine. From my perspective, Greg Scace and Andy Schecter did most of the early work (i.e., they actually published it... can't say anything about industry insiders who preceded them without publishing) in terms of taking a look at brew pressure profiling, each of them coming up with their own systems for accomplishing dynamically adjustable brew pressure. There's much reading available on this, including this thread and this one. There've been some other vague attempts at accomplishing this more simply, such as Tom Jagiello's thread on pressure profiling with a needle bypass valve.

Anyway, some months ago Tom and I got in touch with Fluid-O-Tech USA and started inquiring about pricing and details on the TMFR pump that Greg used, and after some mild consideration decided to buy a pair of pumps for ourselves. I posted about our acquisitions in this thread; however I'd like to post some details about the process of building an interface to control the TMFR pump, caveats that I ran into (some graciously from Greg), and discuss different options for controlling the thing in a better way than I have, as well as pitch in my own experience about the effect of brew pressure profiling on espresso shots.


The TMFR Pump has 2 parts: the pump, which is a rotary pump magnetically coupled to a 3-phase AC motor, and a small black controller that converts single phase 110V/220V power to 3-phase power for the pump.

As Greg explained in his response to Marshall about the TMFR-based profiling system, this setup has a lot of hurdles before it could be used in a mass-produced machine. However, one can cobble together the parts and hook it up to just about any espresso machine for a matter of a few hundred bucks (say, $4-800, depending on how fancy you're looking to get). For starters, I opted to shoot for about the cheapest setup I could build.

The TMFR pump controller has a number of different options for input signals. The most basic one just has a set of DIP switches that allow you to set the motor speed. They're all pretty useless for profiling, except the one that takes a 6-pin input, one of which is a 0-5V input: the controller sets the motor speed between ~1100 and ~3500 rpm proportionally based on the voltage on that input. This means you can control pump speed with a simple 5V AC-DC adapter and a potentiometer.

1. Conquering the Bypass Valve
However, varying pump speed will only get you so far. Rotary pumps as they are used in espresso machines are extremely high-flow pumps that actually bypass most of the excess water that they pump back into the inlet side via an adjustable bypass valve that's typically protruding 90° from the inlet side of the pump. This bypass valve opens when the force of the pump outlet water exceeds the pressure of the inlet side combined with the pressure of the spring pressing against the valve. You can adjust the effective output pressure by compressing the spring via the adjustment screw on the pump. In this setup, the pump's operating speed actually has little to do with the output pressure, provided the speed is fast enough to produce the maximum operating pressure: In fact, the pump's speed will only change the ramp-up rate in this configuration, giving you an interesting method of "variable-rate preinfusion." Once you get to operating pressure as set by the bypass valve, you'll have little control to decline the pressure effectively: Greg explained to me that straddling that point where the bypass valve opens and closes causes very erratic pressure fluctuations, something you really don't want in such a system.

So that leads me to what is probably the most challenging part of building the TMFR pump, at least from an equipment perspective: defeating the bypass valve. To do this, you need to drill a hole through the bypass; the hole's diameter sets the rate of backflow from the pump outlet to the inlet, and it changes smoothly and proportionally with the pump motor's speed. The diameter of this hole is simple enough: the hole should be the right size to produce 9 bars of pressure at the maximum desired motor speed, given espresso flow rate at the output of the system (i.e., when the group is outputting water at a rate of ~100-120 mL/minute, either with a Scace[-like] device or pulling test shots). In practice, there's a pretty good amount of fudge factor here: if you drill the bypass such that it maxes out at 12 bars, you can still get plenty of adjustment, just avoiding running at top speed. This means that you can test your setup so that it maxes out around x bars at free-flow, and then just guess that you'll get about x+1 or x+2 bars more when pulling a shot.

The real challenge to this, however, is drilling the bypass. They're made of brass/stainless steel, so you will want a drill press. Also, we're talking about a hole about 1-2 mm in diameter, and you will need a set of drill bits with very small increments, such as a numbered drill set. I started out with a #51 hole (~1.7 mm), and ended up widening the hole to a #45 (~2.1 mm), through repeated adjustments (#51 -> #49 -> #47 -> #45). Since the hole diameter has a non-linear relationship with flow rate (squared or cubic? I can't recall), it's important to be able to change diameter in very small increments. However, you may be able to get in the ballpark with a 3/64", 1/16", and 5/64" drill bit. Naturally, start with the smallest diameter, unless you have a lot of spares! If you screw up and drill too wide, you can raise the pressure a bit by increasing your line pressure--assuming you had it dialed down with a pressure regulator to begin with (recommended). For reference, here's what the bypass looks like:


The bypass valve seals in the pump around the conical section; note that this image depicts an already-drilled bypass. (photo courtesy of Tom)

Tom made an excellent CAD drawing of how the bypass is to be drilled, based on Greg's suggestions:


Original on the left, drilled bypass on the right. (photo courtesy of Tom)

... but I'm getting ahead of myself. Before you can even begin testing this out, you have to get the pump running. Stay tuned ...

DISCLAIMER: All information/advice posted in this comment and others on this topic are provided as a matter of interest ONLY. The author assumes that anyone who uses this commentary as guidelines for building a similar system has sufficient expertise for this endeavor and willingly assumes ALL RISK associated with constructing that system. Working with electricity and water is inherently dangerous and entails risk of life and limb. It is the reader's responsibility to identify and prevent all possible accidents that may occur as a result of undertaking a project like this.
Nicholas Lundgaard

godlyone

#2: Post by godlyone »

Nicholas,

I am confused why you needed to drill out the bypass.

I have a general understanding of how it works.. you set it to a certain resistance, any water pressure over that will just get cycled back to inlet side.

So if you set it to produce 9 bar at outlet, and for some reason pump speed increases and would theoretically be making 11 bar, the extra 2 bar is shunted back to the inlet and you still see 9 bar at the outlet.

But if the speed is less... and the pressure is less, why wouldnt you see less pressure on the output? It is my understanding that the bypass will only be "activated" when the pressure is MORE than whatever it is set to??

Flair Espresso: handcrafted espresso. cafe-quality shots, anytime, anywhere
Sponsored by Flair Espresso
User avatar
shadowfax (original poster)

#3: Post by shadowfax (original poster) »

2. Building the Control Box
As I mentioned before, the TMFR pump controller you want is the one that takes a 0-5V input. For reference, this happens to be the "analogue input (0-5 V/1100-3500 rpm) interface board (p/n 329111)." A wiring diagram for this board can be found in the TMFR pump manual, available on Fluid-O-Tech's site. Here's a tiny excerpt of the important part, the 6-pin input:



So, you have to build a system that powers the interface board with 5V input, signal it to turn ON by putting +5V on pin 3, ground pins 2 (FAILURE line), 4 and 6, and put 0-5V on pin 5. Simple enough.

As I also said before, the pump controller box takes 120V/240V power (choose when ordering), and you will need to supply this to the controller, preferably from a central box that can be powered off. Note that there can be some delay to starting the TMFR pump by delivering power to it, and that doing this is not recommended. Instead, it's preferable to power the pump controller all the time, or at least whenever the machine is on.

You'll need a method, then, to convert the (probably) AC signal from your espresso machine that turns the pump on into a +5V signal to pin 3 of the pump controller. The best thing for this job is a silent, reliable AC input-DC output SSR, but you can just as well cheap out and get a $3 mechanical relay with an AC coil. The current across the contacts is virtually nothing, so you shouldn't have to worry about ever burning them out.

Finally, you need a device to vary the voltage on pin 5, if you want to change the motor speed and thereby change the pump pressure. The easiest way to do this is to wire in a potentiometer from the 5V power supply unit that you have to get to power the interface board. Tom suggested that a 10kΩ wirewound potentiometer is perfect for this application. He's suggested that a linear pot is best, and that's what I'm using, but at some point I'd like to wire a logarithmic one to see if it mimics the squared (right?) relationship between motor speed and pressure.

That's the $5 option, of course. You can control the system very elegantly with a pressure transducer and a PID controller that has a ramp function and can read the pressure transducer's signal as psi (preferably) and output 0-5V as the control signal. Note, a cheap Auber PID isn't going to do this, and I believe such a controller is likely to cost hundreds of dollars. I'm hoping to upgrade to something like this in the future; if I do I will post about it. I hope that Tom will also chime in with his own ideas about controlling the pump. For now, however, this post will cover the use of a potentiometer for control.

... That's basically it. Put it together in a box, and you have a friggin' mess!


My box; see the Flickr link to see the original, annotated photo that identifies all the parts and input/output cords and how they are used.


Here's another angle on the box that shows the pot on the box and the power switch a little better.

Of course, it may seem kind of strange to have the potentiometer on the box, which goes under the cabinet. However, this one isn't the main pot; it acts to set the maximum voltage on the potentiometer that attaches remotely to this box, and sits on the counter. I found that I was bothered A LOT by the pump's noise at its maximum speed, and I chose to drill my hole such that the maximum pressure is ~13 bars at free-flow and max speed. I use this bottom potentiometer to cap that to 9 bars during a shot. This way, I get the full range of ~line pressure all the way up to 9 bars matching the minimum and maximum settings on the countertop potentiometer that I use to actually adjust pressure during the shot.


The installation. Again, please see the original Flickr link to view the annotated image with labels for all the components.

After screwing down the lid for the control box, things start looking a lot cleaner. The system is certainly not as simple as a typical rotary pump, but the advantages are many: no need for a start capacitor, much lower power draw, and a much longer pump life because there's no physical coupling between the motor and pump--it's magnetically driven. The seal is much better than the bearing seals on a typical rotary pump. And of course there's that whole dynamic pressure variation thing. What's that about again?

Finally, A view up top.


Not quite a Slayer paddle, but it works. Too bad it's too small for a 'GIT'R'done' sticker!

DISCLAIMER: All information/advice posted in this comment and others on this topic are provided as a matter of interest ONLY. The author assumes that anyone who uses this commentary as guidelines for building a similar system has sufficient expertise for this endeavor and willingly assumes ALL RISK associated with constructing that system. Working with electricity and water is inherently dangerous and entails risk of life and limb. It is the reader's responsibility to identify and prevent all possible accidents that may occur as a result of undertaking a project like this.
Nicholas Lundgaard

User avatar
shadowfax (original poster)

#4: Post by shadowfax (original poster) »

godlyone wrote:Nicholas,

I am confused why you needed to drill out the bypass.

I have a general understanding of how it works.. you set it to a certain resistance, any water pressure over that will just get cycled back to inlet side.

So if you set it to produce 9 bar at outlet, and for some reason pump speed increases and would theoretically be making 11 bar, the extra 2 bar is shunted back to the inlet and you still see 9 bar at the outlet.

But if the speed is less... and the pressure is less, why wouldnt you see less pressure on the output? It is my understanding that the bypass will only be "activated" when the pressure is MORE than whatever it is set to??
Ilya, that's a good question, and one that took me a good bit of time on the phone with Greg to wrap my head around. You're right, if the pump can spin slowly enough to produce low pressure at espresso flow, then you can control the pressure in that range. However, I believe you won't be able to control declining pressure very accurately with a configuration like this. Also, you would really have to buy a very low-flow pump to be able to do this at all, which I made the mistake of not doing. I'm under the impression that even the smallest one has a flow rate that's rather too high to do what I'd assume is possible, if a pump with the right basic flow rate existed. I believe Greg's TMFR pump is a much lower-flowing one than mine (the lowest, Greg?), and he definitely found that he needed to drill the bypass out as I've described.

I hope that helps a little. I'm still not entirely clear on why it's necessary; I'm mostly regurgitating what I've learned from Greg's experience.
Nicholas Lundgaard

User avatar
shadowfax (original poster)

#5: Post by shadowfax (original poster) »

3. First Impressions
I completed my setup over this past weekend, making my first shots on Sunday afternoon and evening. It's been a busy couple of days at work as well, so I've only had a little bit of time to play with it. I've been using Intelligentsia Black Cat, which is just what I happen to have on my 'rotation.' It's perhaps a sub-optimal choice, since it's relatively easy to work with as is. I've only used one basic profile, which is I think the 'popular' one that loosely mimics a lever machine--a quick ramp to full pressure and then a slow decline over the shot duration. Typically I've gotten to ~9 bars or slightly less (on the machine's gauge) by 7-8s, and then held that till 18-20s. I've been declining the pressure from there to the end, around 26-30s in. I only today installed the second "capping" pot that I described in the previous post; prior to that it was more challenging to control. I started out with a single 3-turn potentiometer, and found it to be extremely user-unfriendly. Switching to a single-turn potentiometer with a second to cap it makes the system MUCH easier to control consistently.

But I digress. So far, I'm liking a modest decline the best, and pulling sequences of shots at full pressure and then declining pressure certainly yields a notable muting of the low-end flavors, and by comparison a sweeter, brighter shot. Sometimes this isn't spectacularly better if at all. Other times, it's quite delectable. Not exactly a blind test, but a nice first showing for me. I'm looking forward to using this basic profile on some 'edgy' single origin coffees to see if this can be used to balance them out well.

That's it for now, I guess. I'll try to post some videos of the system in action later this week. However, I would like to thank Greg, Andy, and all the other guys who've studied this topic and come up with really interesting solutions. Kudos to you guys, and thanks for the help, Greg!
Nicholas Lundgaard

User avatar
gyro

#6: Post by gyro »

Thanks for taking the time to post this Nicholas. (and thanks Greg, Andy, Tom as well of course) Fun stuff!

User avatar
michaelbenis

#7: Post by michaelbenis »

Hats off to you Nicholas, for some great work and reporting.

And look forward to hearing how this all pans out.

Cheers

Mike
LMWDP No. 237

CafelatStore: home of Cafelat products online
Sponsored by CafelatStore
darrensandford

#8: Post by darrensandford »

Wow, I read this and my brain says "PIC controller! PID control! Graphic display with profile graph!"

Please, shoot me now or hide my wallet.

User avatar
dsc

#9: Post by dsc »

Hi guys,

good write up Nicholas, this is actually one of the reasons I'm glad you did all of this before me:)

Again a big thanks to Greg, Shawn at FOT, and my friend Jon from TMC. I know I can be annoying with all the questions:)

Anyways, I've not finished my setup yet, I've ordered some missing hydraulic parts and still have to order some electronic bits and pieces. I'm guessing it all won't arrive before the weekend which means I'll probably start testing next week. At least for now, my config is going to be the same as shown above, so a pot to control the pump speed and no fancy controllers. I actually need this rig to check the characteristics of the pump, before I move on to more complicated control via a PIC. The software for the uC is actually written and sort of tested. The final version will look similar to the test version but will allow auto-replays of the profiles done in manual mode. So it's going to be like a pot with memory, oh and the pot will let you set the pressure not the speed, so it will be more user-friendly.
Wow, I read this and my brain says "PIC controller! PID control! Graphic display with profile graph!"
Ah yes, profile graphs:) something I thought about. Setting the profile on a small screen via some buttons and seeing how it progresses during the extraction. Sounds cool, but I'd rather put together something simple and cheap and perhaps add an LCD later on if it works.

Regards,
dsc.

User avatar
mhoy

#10: Post by mhoy »

The Atmel Butteryfly AVR prototype board is what you need to control your system. Inexpensive micro-controller with an LCD screen. Various buttons for user input and fairly simple to program. Some flash memory to store your profiles, etc.

Mark