Jake_G wrote:My point is that it would interesting to see what happened with a decent machine if it just ran the pumps and let the shot do what it wants to do. You'd need a pressure ceiling setpoint, since there is no OPV (that I know of), but I think there would be value in understanding what all the tech is buying us by having a "Stupid" mode, where the pumps run at a fixed (adjustable) duty cycle and the temperature is set by a fixed (adjustable) bias between the pumps.

It occurs to me that there isn't really such a thing as a truly fixed duty (or at least not one that makes sense) for a Decent machine, due to the two-pump water mixing that used to regulate temperature. I think what you're really talking about is a "fixed input force." But here the input force is a combination of the force used to drive water into the water heater (the hot water pump) and the force used to drive cold water (the cold water pump). Moreover, I suspect the additive force of these is difficult to model.

For example, running the hot water pump at full duty probably yields a different result than running the cold water pump at full duty, both due to simple pump-to-pump variance, and because the hot water pump has to force water through a much longer path (through the thermocoil).

And, even if one knew a CP (cold pump) duty that was equivalent to a full HP (hot pump) duty, I'll bet that choosing 50% of each of those duties for the two pumps yields a different result - that is, it's not clear to me that the output varies linearly with the duty cycle.

So, assuming one wants a predictable brew temperature, there might not be a simple way to have the Decent machines behave the same way as a typical vibe pump machine.

There are various options that might theoretically approximate what you call a "stupid mode" (e.g. use a fixed duty cycle for the hot pump, and let the firmware adjust the cold pump as necessary for temperature control), though.

MrEd wrote:As you can see, wetting occurred after pressure rise began (at about 2 bar), but a second or so before the end of preinfusion was triggered at 4 bar (the vertical line).

This is great stuff - thanks!

Visible wetting of the basket should lag actual saturation of the puck, because it takes a moment for the espresso to make from the edge of the puck through the basket holes. (I would guess that the lag is slight, but don't really have a good guess of how slight.)

The thing I find notable about the graph is that pressure stays flat only until about 6.5 or 7 seconds into preinfusion. That seems to me to be the interesting point - the point at which flow is being disrupted by hydraulic forces. From that point on, you're continuing what the machine says is preinfusion about another 4-5 seconds at 4.5 ml per second, and you finish using about 61 ml for that "preinfusion." If you look only at the "pre-hydraulic" period, you are using about 30 ml of water.

What was your dose weight? My eyeballed numbers look tantalizing close to 2x dose to saturate the puck (if you assume any pressure blip indicates saturation), but assuming a modern dose of around 18g or so, you're probably shy of 2x dose.

My theory is that if you have a faster flow rate, you will get drops on the bottom of the basket before the puck is fully wetted - that is, hydraulic pressure will force some espresso out before capillary action can spread the water out across the entire puck. This makes me wonder what the chart would look like if you hold everything else constant (same coffee, same dose, same tamp) but lower the preinfusion rate to, say, 1.5ml. Would we see a pressure blip after a higher volume of water into the puck?

MrEd wrote:I discussed making this process easier with John Buckman, and suggested having the machine optionally make beep sound at profile transitions. This way you could video the basket directly, and sync the chart to the video later. I suspect some variant of this will be incorporated in a future software update.

There are probably a number of additional optional audio feedback options that would be useful. For example, having the machine signal that it is up to temp after being woken up, or during a shot that pouring volume has passed a set threshold (to let you know that it will soon be time to end the shot).

An option to let you know when Hal is unable to open the pod bay doors would be good, too.

roastini wrote: So you could just pump water as fast as possible. But doing that likely will result in filling the headroom long before the puck is full saturated. Pressure will rise, and likely you will disturb the uniformity of the puck, leading to a suboptimal shot.

So I did something related to this idea recently, and the results are interesting.

I set the profile with no preinfusion, just hold 8 bar flat pressure. I expected the pressure graph to basically ramp quickly, then hold a line. What the graphs show instead is that there is minimal pressure for the first several seconds, then a rapid ramp spiking to just over 8 bar at about 10 sec, followed by correction down to 8 bar then holding pressure as expected. (Contrast the dotted green pressure profile line with the dark green sensor pressure line.) The flow profile rapidly ramps to 6 ml/sec, then drops to less than 1 ml/sec as the puck compresses coincident with pressure ramp up. As the shot proceeds the flow gradually increases to over 1.5 ml/sec due to puck degradation.

These graphs make sense in retrospect, from the perspective of the relevant sensors. Since the pressure sensor is in the manifold, and effectively reading puck pressure, there is not much rise until the puck approaches saturation. During this early phase, the pumps are at maximum flow of 6 ml/sec. This looks very similar to any other pressure profile shot with preinfusion. What is most interesting to me is that this says there is effectively no such thing on this machine as no preinfusion. Rather it is just the amount time/flow allowed for preinfusion to occur.

Contrast this, for example, with a manual lever machine, where you can pull the lever and fill the group piston in about a second, and then immediately pull the shot, as opposed to waiting several seconds for the puck to fully saturate.

roastini wrote:My eyeballed numbers look tantalizing close to 2x dose to saturate the puck (if you assume any pressure blip indicates saturation), but assuming a modern dose of around 18g or so, you're probably shy of 2x dose.

This is not right, as I think about it. You won't see a pressure rise until the headroom is filled, too, so however much water was consumed at this point equals "water in puck" plus "water in headroom."

MrEd wrote:These graphs make sense in retrospect, from the perspective of the relevant sensors. Since the pressure sensor is in the manifold, and effectively reading puck pressure, there is not much rise until the puck approaches saturation. During this early phase, the pumps are at maximum flow of 6 ml/sec. This looks very similar to any other pressure profile shot with preinfusion. What is most interesting to me is that this says there is effectively no such thing on this machine as no preinfusion. Rather it is just the amount time/flow allowed for preinfusion to occur.

This graph is also interesting. I find myself wanting to know the total flow as of 5s, which is when we first see the pressure go up (maybe 4.5s, actually).

This is hard to determine, due the machine taking a few seconds to get up to 6ml flow. My best estimate is that it was around 27ml. (It would be nice if there was a way to see a "cumulative flow" graph.) If so, and assuming the dose was the same as in your previous shot graph, that does suggest that basket wetting probably occurs with less water in the puck when the flow rate is higher. But it's hard to be sure, both due to shot-to-shot variance in prep and because I'm not very confident in my 27 ml estimate on this shot.

roastini, you are going to burn through a lot of coffee when you get your machine...

roastini wrote:It would be nice if there was a way to see a "cumulative flow" graph.

You are asking to calculate the area under the curve between specific time points, which should be doable with the appropriate software. All of the trace data are stored on the tablet, and can be downloaded. Short of going through that, the easiest way to do this on a small dataset is to print out the charts, cut out the area of interest and weigh the paper. It is surprisingly accurate.

MrEd wrote:I've done a little investigating of this question. Unfortunately right now it is not trivial because it is hard to watch the charts and the basket at the same time. The only way I've found is to make a movie from an angle that captures the charts on the top of the screen, and the basket in the mirror at the bottom of the backsplash...

Ed, until John will incorporate a screen recorder into his app, there are screen recorder apps that might help you to simultaneously record both Decent chart and basket. No mixing and syncing work is necessary.
I use AZ Screen Recorder, for instance. The basket gets recorded in a small screen, like the Picture in Picture TV.

MrEd wrote:So I did something related to this idea recently, and the results are interesting.

I set the profile with no preinfusion...

Ed,

Did you happen to notice when first drops appeared using this method?

Thanks!

- Jake