This morning I pulled an espresso after showing 5 grams in the cup. It was not so great out of the Bosco and ran too fast. I have been using 2 gr. previously and that tasted great and pulled at a reasonable rate. To get to 5 gr. took over 2 minutes but since I was having success at 2 grams I think I have found the sweet spot for my system and will be satisfied with using that timing.

The idea isn't to time PI to 5g. Whatever you do you do as before. The 5g is just a time point.

But the point is that 5g, I think even more than 1g, is past the point of preinfusion for most of us. I don't think the end of preinfusion is measurable by what's in the cup. I think it's better marked by beginning of desired flow. What that language is around what desired flow is and how it's measured I'm not sure.

But the point of flow is hard to make consistent. One may think a few drops is flow, while others require a thick blinding pour.

5g is arbitrary since it usually happens after flow started (in your case after the 2g in drips).

AssafL wrote:The idea isn't to time PI to 5g. Whatever you do you do as before. The 5g is just a time point.

I understood that, but was just a FYI for anyone.

I thought it might help the discussion of the original topic to post some graphs of profiling on my modified GS/3 AV.

The following graph is a shot done with "traditional" espresso parameters (18g at about 25 seconds) using the default flow and pressure profiling determined by my gear pump and the GS/3 gicleur. That means: open the 3-way and turn on the motor at a speed that results in 9 BAR max at the puck.



Pressure is measured by a transducer at the pump output and flow is measured by the GS/3 flow meter. Seconds are along the top, pressure on the left axis, flow rate on the right axis. The pressure transducer and flow meter are connected to an Arduino microprocessor, which also controls the 3-way and the gear pump. The user interface to the Arduino runs on an Andriod tablet. It has buttons and sliders for commands and settings sent to the microprocessor, and both numeric and graphic displays of data sent back by the Arduino. Both the Arduino and Android app monitor the shot weight reported by an Acaia Lunar scale. Either one can be configured to cut the shot at the target weight, but in this case I used the Android app for that. What you see above is a screen shot from the Android tablet.

It takes about a second for the pump to reach the target speed, where the flow rate hits about 350 ml/min. The "normal" free flow (no puck) rate for my GS/3 is about 450-500 ml/min. The lower flow rate occurs because I've reduced the gear pump speed for brewing so it maxes out at 9 BAR. The pressure seen by the pump is about 5 BAR, which is a bit higher than line pressure due to the restriction of the gicleur, which is between the pump and puck.

It takes only 2-3 seconds for the basket to pressurize, at which point you see the pressure rise rapidly. The ramp time is about four seconds. At the same time, the flow rate drops precipitously due to the resistance of the puck.

The max pressure reads as 10 BAR because in a GS/3 the pressure drops about 1 BAR across the gicleur. In other words, pressure at the pump has to be 10 BAR to get 9 BAR at the puck.

As the puck loosens up, the pressure slowly declines and the flow rate slowly increases. This is because the gear pump has no bypass valve to force constant max pressure. The pressure curve for a rotary pump would be flat at 10 BAR and the flow rate would increase quite a bit as the shot progresses. I believe the effect of the gear pump profile is to get more contact time without diluting the drink as much, which helps with lighter roasts.

Here's a profile for a "Slayer-like shot" of a nice Geisha sample I got from Temple Coffee Roasters. It's roasted for brew, though not ultra-light. I ground the coffee much, much finer than the grind range I use for traditional shots and reduced the line pressure to 1.5-2 BAR. The 3-way is opened with the motor off, so the basket slowly fills at the reduced line pressure. I set the preinfusion time to 20 seconds. A drop or two of coffee became visible on the bottom of the basket right at the 20 second mark, at which time the pump was turned on and set to brew speed. The rest of the shot took about 35 seconds, for a total shot time of 55 seconds. Total shot times for lighter roasts can be 60-90 seconds.



Unfortunately, the chart library used by my Android app doesn't display the pressure and flow values when the shot is so long that the data points are very close together. But it does allow me to zoom in on parts of the chart, which spreads the data points out enough to see the values. Here's a zoom that shows the shot from the beginning to the 50-second mark, losing only about five seconds from the end of the shot:



I'm not sure why the flow rate starts out so slow, then peaks and slowly drops during preinfusion, but I suspect it's either because it takes a few seconds for the flow meter to get going or it's due to a lag in my app's pulse averaging algorithm. Or both. It looks like the basket begins to pressurize at about 15 seconds, at which point the pressure rises slightly and the flow rate decreases. Neither moves a whole lot because the line pressure is so low. After five more seconds I saw one or two drops emerge from the bottom of the basket (really just coffee emerging and wetting a tiny spot, not actually dripping.) At that point the motor kicked on and the normal ramp took over. Not sure why the flow rate peaks so high. Again, it could be an averaging issue with my app. Peak pressure is only about 9 BAR because the low input line pressure changes the pump's pressure characteristics, and I didn't bother to increase the brew speed to compensate. 8 BAR is fine for a light roast anyway. The remainder of the shot looks similar to the traditional shot.

The last graph, below, shows a traditional shot where I manually reduced the pump speed gradually during the second half of the shot. The idea was to keep the flow rate constant. I've also played around with gradually lowering the pump speed to zero, but that causes the flow rate to drop very low instead of staying level, which I believe results in somewhat greater extraction due to more contact time. Not the best for the coffees I prefer, but it might work well for really light roasts.



I was shooting to keep the flow rate at about 90 ml/min during the last part of the shot. The different presentation is an alternate screen I can use to monitor certain parameters and the chart during a shot. The weight is negative because I removed the cup from the tared scale after the shot.

N.B.: Assaf believes that the slight dips in pressure, and the corresponding increases in flow rate during my shots is due to channeling. He may be right, though they could also be due to inconsistency in the transducers and/or the way my averaging algorithms work.

Hope these charts help.

Dick, thanks for posting these charts. They are similar with what I'm seeing on the DE1 (see here). Even the pressure ramp up time is consistent (7-8 seconds), probably because you have reduced the pump flow rate to ~6 ml/s.

another_jim wrote:The graphs of flow rates as the puck saturates, before seeing or weighing a flow; how good are these? I'm not doubting that the flow metering is accurate; I'm wondering how many nooks and crannies in the water delivery are filling up along with the puck.

Good question and good point.

The actual saturation rate and soak time vary a lot depending on the coffee (especially roast level), grind, dose, distribution, preparation, flow rate and pump settings. With all these things being variables, I agree that we don't really know what's happening inside the puck during preinfusion, particularly with regard to the time hot water is in contact with some or all of the grounds.

But going back to your original quest for a simple measure of preinfusion, I don't think a simple measure can tell us what's really happening inside the puck. So, I think the best you can do to describe preinfusion is one of the three markers we've been discussing: time to first rise in pressure, time to first drops, or time to 5g.

That said, I think providing a recipe that reads something like "Set your preinfusion so it takes 20 seconds to get to <choose your favorite marker>" isn't useful because the result depends on the interaction of all the parameters I listed above, not just the flow rate.

another_jim wrote:I'm not doubting that the flow metering is accurate...

Well, you should be. Due to systemic errors and lagging realtime response of turbine flow meters, the DE1 (and apparently this interesting new machine) have moved from flow meters to pump modeling.