I wonder if you can switch the parts yourself... the downside of being on the other side of the world. I had thought it was too easy to get the pump in the first place, normally anything connected with Hong Kong is difficult. Guess in the end I was right!

Yeah, Shawn was trying to persuade me to do it myself, actually. He described the process as 'fairly easy,' but it requires a special tool called a pin spanner to take the 'threaded cup' off. This is the lid of sorts that holds the front of the pump on, part number 11 in the bottom left of the exploded parts diagram.


TMFR Pump Parts. Source: 2009 parts list, via Fluid-O-Tec. Document is on their site (free registration required)

The process requires removal of parts 11, 10, 9, and 7, as well as I think some of the parts on the back side (15, 13, 12) to get to the liner (part 6) and the pins (part 4). It actually does look pretty easy--I took the back off of mine when I took it out to send off--I'd say you should probably be up to it if you're technically minded enough to build the other parts of this setup (I say, as I box my pump to have the FOT guys do it ). Shawn did say that one of the really nice things with the TMFR pump is that it is easier to repair, especially for the end user, compared to standard rotary pumps. It decouples more easily from the motor, and then when you're opening it up you don't have to deal with those tightly seated sealed bearings, etc.

Perfect, thanks for that. If I can't manage it I'm sure my friend who helped me hack my roaster apart will have the ability. Since I roast all his coffee, I don't feel too bad always running for help with things above my skill level. Guess it wouldn't hurt to get a couple of spare bypass pins as well in case I botch that up.

I'd love to see some videos of your rig in action!

I had a couple of thoughts when reading this thread.

First off, you already have a nice digital flow meter built into you machine. Any automatic will have a Gicar flow meter with a nice digital output back to your controller board. You should be able to tap into this line (with a opto-isolator for extra safety) and feed it our to your PIC or what ever uC you decide to use.

Secondly a basic RC filter placed on any PWM output pin will be able to produce the 0-5V signal you require for the pump controller board.

With these two simple parts you should be able to test Jim's constant flow theory with just a dozen or so lines of code.

Simple code flow-

Start pump
Run preset ramp up for infusion
set pump speed for normal flow
sample the flow pulse rate several times
average the flow rate
switch pump power to match average flow rate and repeat this line

Add one button or sample your pot for telling the uC to start and stop and you would have the basics. A couple of hours and you could test the 'flat flow' profile.

-Cecil

I'd be very impressed if an espresso machine flow meter provided remotely fast enough feedback to facilitate real-time calculation of flow rate. I think at espresso flow that a flowmeter pulses at most <10 times per second (is that right?). Even if it were a bit more than that, the latency on measuring average flowrate will be quite high. That will probably put a cramp on the accuracy of any control system. But I perhaps there are better flowmeters or even self-contained flow-rate meters out there.

... Am I crazy to think it won't work?

Hi guys,

it might be fun just to check the flow during an extraction, but I agree the accuracy would probably be quite low when using the on-board machine flow meter. I will try to hook it up just to see what numbers I'm getting.

One thing I'm not so sure about is what flow should we be aiming at? What is 'normal flow'? I'm guessing it changes quite a lot during an extraction and it might be hard to just pick one value. On the other hand it will be possible to keep the flow steady at a given level and use a PID algorithm to change the pump voltage.

Regards,
dsc.

Instead of thinking of strictly counting the pulses themselves you would get a lot more flow rate detail from measuring the pulse's position. Instead of counting the number of pulses count the time from leading edge to the following leading edge. This way you can measure the delta flow at 10 times per second, and with more code space and a look up table you can directly translate that to a ml/sec figure. But for pump adjustment the delta flow data may be a bit more simple with a uC timer module.

Code idea:

ISR for pulse position timing from timer module, stores the 16 bit timer count in a local word var. Then reset isr enable bits.

Main loop
read pot on d/a
copy 16 bit to second work var
loop to divide work var by pot var
output 8 bit value to d/a via pwm rc filter
do it all again


OFC you would still want the ramp up and infusion sections from the first code idea.

-Cecil

Hi guys,

well seeing as we are moving towards making a custom controller I'm thinking about moving from assembler and using C instead. It's much easier to do floats and big number calcs in C then in ASM and I will have to do both when implementing a PID.

For now I think I will use the PT signal as feedback and do a simple hysteresis loop for control. The idea is easy, but I'm not so sure how to handle the pot. The main reason is the variable pressure at a given voltage, which depends on the grind setting and how far you are into an extraction. I will have to think about it some more, but due to internal code calcs and the fact that the PT signal is actually 1-5V (4-20mA on a 250ohm resistor) and the pot is full 0-5V it's quite tricky. Of course ASM not being able to handle floats (well not as easy as in C) also isn't helping.

Regards,
dsc.

EDIT: problem solved:) should be easy to implement. I will try to first put together a basic hysteresis controller, just to try out manual control. If that works well I will switch on to 'memorising' (which is done anyway).

OK, just to play the devil's advocate, what is the point of any of this stuff?

I approach espresso as I do any of my other oral fixations, which include fine food and wine. I'm not too good when it comes to physics (as Andy S. has pointed out a few times), so I seldom take that approach. Rather, I use my sense of taste to try to decide what I think is "good," and then when I have found a "good" thing I try to figure out how to buy more of it, to replicate it, or to get the same sensation with some other approach.

An example would be that once finding a Bordeaux or a certain type of coffee I like, then I look for something similar that might produce for me the same sort of pleasure. Or, maybe I did something in a weird way one time, or accidentally screwed up, but the result was surprisingly good. So then, I might try to repeat the same "mistake" again, this time on purpose, to try to recreate the good result that I got by accident the first time.

I don't see any of this sort of reasoning going on in this thread, other than perhaps the assumption that one could electronically mimic the profile of a lever machine by doing this or that with pressure profiling a rotary pump. To that I would say, "why not just go out and buy a lever machine?" For one thing, it would likely be cheaper, and for another, it would probably give a real lever profile rather than an attempt at copying the profile of one.

OK, you say, that this isn't really what you are attempting to do, you are experimenting with pressure to see how you can improve your shots. But this is not really experimenting, which requires a rigorous experimental design testing some sort of hypothesis vs the "null hypothesis," which means in layman's terms that you are comparing your altered procedure with SOP, standard operating procedure. Rather, instead, what I see is what I would characterize as a sophisticated high school physics lab project, which once built will not and cannot reasonably have its results tested in any sort of real scientific manner. What this means is that at best, if one actually makes a real effort to test it, the "results" will be no better than one person's descriptive comments about the impact of any of this on his personal perception of espresso shot taste. And, it is hugely unlikely that if some other sophisticated coffee person were to then run the same testing regimen, that he would make the same observations.

So, we basically have a solution in search of a problem, with results that can't possibly be generalized to any sort of communal benefit.

In summary, this whole idea sounds to me more like a toy than like anything that anyone else here who is genuinely interested in improving his coffee, will be able to benefit from.

I am not saying that pressure is not a variable in the espresso making process, or that its effects shouldn't be tested. What I am saying is that perhaps it would be worthwhile to compare the output (e.g. the espresso) of some of what is already out there that purports to do this (Slayer, paddle group, what have you) in a blinded fashion to more simple (and common) machines with "typical" pressure profiles to try to establish if there really is any benefit at all to be chased. If there is real benefit that people can detect in blind tasting, then perhaps that would be the time to start trying to reproduce it. Before proving that there is any benefit, however, the cart is getting a bit far out in front of the horse, at least in my opinion.

ken