AssafL wrote:... The difference should be the amount of water absorbed by the puck, ...

Don't forget the volume of the brew path from the 3-way valve to the coffee bed, including the varying headspace for different doses.

jonr wrote:It would make more sense to test at something like 30 ml/min and 70 ml/min.

With ~10 ticks/sec and flow rate re-calculated at each tick (based on time stamps), it sounds like there is good resolution for flow control - it doesn't change that fast. For example, I use a scale and only adjust pump power once per second.

Yes, I agree it makes sense to try this at a lower flow rate (albeit I think the Nozzle LM chose is rather large to maximize accuracy at low flow rates).

The main problem with the way I tested the scenario was that I controlled flow at the input to the machine (with a needle valve and pump off for "low" vs. no needle valve and pump on for "high") without any portafilter in place. While it was easy for me to do - since all I really needed to do was change the timer knob on the preinfusion relay - pressure, flow rates, are all variables in the scenario I tried and not consistent with the real use case.

The correct way would have been to use a needle valve (or real coffee) to reduce the flow rate at the output of the machine (so the pressure stays relatively consistent at the nozzle side of the flowmeter).

It is on the list of things to do but for some unexplained reason takes a lower priority to the 2 year old, SO, day job, etc. I need to convince my SO to spend a spa like weekend at her mothers' house.

Jacob wrote:Don't forget the volume of the brew path from the 3-way valve to the coffee bed, including the varying headspace for different doses.

I guess the brewpath volume is mostly negated if you consistently do a flush before a pull.

As to headroom - excellent thought. My guess is once precision (repeatability) is achieved I will have to start figuring out the Volume/Mass relationships including the space above the puck.

I guess my reasoning here is that to measure the dose to provide adequately good EBF control - one needs accuracy (+- 0.x grams), but to achieve accuracy one first needs precision (repeatable results that don't fluctuate all the time).

Don't know if that is doable with the necessary level of accuracy - but on a positive note, indeed there was a clear lag between the flowmeter pulses (as preinfusion progressed) and the weight in the cup. Those pulses that take place prior to the cup should be (somehow) related to the dose and headroom.

At a certain point in time, the weight in the cup vs. the weight of the water as measured by a precision flowmeter (dose+cup+headroom) should describe a specific, target EBF. Theoretically, an Arduino should then be able to cut the shot. Again - the assumption is that the precision and accuracy are there.