Jim's experimentation with the dry coffee dose has sparked a lot of interest among the terminally obsessed (you know who you are). I decided to play with the dose and grind variables to see how they affected the flow rate and subsequent shot timing.

In this testing session, for every gram additional coffee in the portafilter, the shot time increased about 4 seconds. And on the Mazzer Robur, for each notch that the grinder was tightened, the shot time increased about 3 seconds.

Fun to play with, and useful info. YMMV, of course.

AndyS wrote:I decided to play with the dose and grind variables to see how they affected the flow rate and subsequent shot timing.

In this testing session, for every gram additional coffee in the portafilter, the shot time increased about 4 seconds. And on the Mazzer Robur, for each notch that the grinder was tightened, the shot time increased about 3 seconds.

Very interesting results, especially the relationship between dose and shot time. Were you using brew ratio to determine when to cut the pour? Has this been demonstrated before? An increase in pour time with increasing dose is to be expected, but a linear dependency is pretty cool. BTW, I'd guess the linear relationship only holds for a "sweet spot" in dose, and breaks down for severe underdosing and overdosing.

Nice work, Andy!

I got roughly linear changes too when I systematically changed dose and adjusted the grinder for similar timed shots.

This actually surprised me. I was expecting an exponential curve if one charted the weight of the shot after a fixed time to the grind level. That is, one gets a weight of zero if one sets the grinder to zero, then slowly increasing shot weights at first, and then ever faster as one loosens up the grind. I still think this has to be the case, but that the "reasonable espresso range" is so narrow, that everything is close to linear in it.

RapidCoffee wrote:Were you using brew ratio to determine when to cut the pour? Has this been demonstrated before? An increase in pour time with increasing dose is to be expected, but a linear dependency is pretty cool. BTW, I'd guess the linear relationship only holds for a "sweet spot" in dose, and breaks down for severe underdosing and overdosing.

jim schulman wrote:the "reasonable espresso range" is so narrow, that everything is close to linear in it.

I agree with youse guys; since we're working within a very narrow range, the relationship looks linear even if it truly isn't.

Although on some machines, hitting the shower screen ends the linearity fast!

John, I was trying to keep the brew ratio fairly constant for the purposes of this test.

another_jim wrote:I got roughly linear changes too when I systematically changed dose and adjusted the grinder for similar timed shots.

This actually surprised me. I was expecting an exponential curve if one charted the weight of the shot after a fixed time to the grind level. That is, one gets a weight of zero if one sets the grinder to zero, then slowly increasing shot weights at first, and then ever faster as one loosens up the grind. I still think this has to be the case, but that the "reasonable espresso range" is so narrow, that everything is close to linear in it.

That sounds right. I was actually thinking in terms of flow rate, which has a well-defined maximum (no coffee) and minimum (completely choked). I'd predict an inverse sigmoid relationship between dose and flow for a fixed (and reasonably fine) grind setting:



Both dose and flow rate have been normalized to [0,1] on this hypothetical S-curve. Andy's measurements would presumably be near the center of the curve, with dose in the [0.4,0.6] range. Pure speculation, of course.

RapidCoffee wrote:That sounds right. I was actually thinking in terms of flow rate, which has a well-defined maximum (no coffee) and minimum (completely choked). I'd predict an inverse sigmoid relationship between dose and flow for a fixed (and reasonably fine) grind setting:

<image>

Both dose and flow rate have been normalized to [0,1] on this hypothetical S-curve. Andy's measurements would presumably be near the center of the curve, with dose in the [0.4,0.6] range. Pure speculation, of course.

Yeah, it has to be sigmoid -- the holes in the basket and the flow/preesure limits on the pump limit the maximum rates just as the grinder zero limits the minimum one. When I said exponential, I was forgetting about the maximum speed limit.

I wonder if there's any sense normalizing the shot time/volume data on every machine by using the zero flow at one end (no normaliztion required) and the water debit flow/volume rate at the other (as the 1)? In other words, if the water debit works out at 90 grams in 25 seconds, and one gets a shot that's 18 grams in 25 seconds, is that a flow rate of 0.2?

another_jim wrote:Yeah, it has to be sigmoid -- the holes in the basket and the flow/preesure limits on the pump limit the maximum rates just as the grinder zero limits the minimum one. When I said exponential, I was forgetting about the maximum speed limit.

The sigmoid is indeed an (inverse) exponential function:

another_jim wrote:I wonder if there's any sense normalizing the shot time/volume data on every machine by using the zero flow at one end (no normaliztion required) and the water debit flow/volume rate at the other (as the 1)? In other words, if the water debit works out at 90 grams in 25 seconds, and one gets a shot that's 18 grams in 25 seconds, is that a flow rate of 0.2?

I think I see what you're getting at. This might also be another way to quantify shot characteristics. The flow ratio helps incorporate pour time (which Andy's brew ratio does not address). The problem is, 25 seconds may not be an optimal brew time for all pours...

RapidCoffee wrote:I think I see what you're getting at. This might also be another way to quantify shot characteristics. The flow ratio helps incorporate pour time (which Andy's brew ratio does not address). The problem is, 25 seconds may not be an optimal brew time for all pours...

The 25 seconds was an example. If one has the water debit as X grams/second; then the flow-ratio one wants is:

shot_weight/(shot_time*X)

Not sure this will be any use for standardising among machines.

In my research, I settled on dose/filter_area as the predictor for extraction ratios. However, this is distinctly a second best; this ratio will also affect the flow characteristics. The hidden variable in all this is some true measures of grinder setting. One would be the grind fineness of the coarse particles, which would determine the soluble yield. The other would be some, or the proportion of fines, or whatever else it is that determines the flow rate through the puck

another_jim wrote:In my research, I settled on dose/filter_area as the predictor for extraction ratios. However, this is distinctly a second best; this ratio will also affect the flow characteristics. The hidden variable in all this is some true measures of grinder setting. One would be the grind fineness of the coarse particles, which would determine the soluble yield. The other would be some, or the proportion of fines, or whatever else it is that determines the flow rate through the puck

I think there's a typo in your last sentence.

In any case, predicting the flow rate based on dose, filter area, extraction pressure and grind setting is impossible. The rampup rate to full pressure at the beginning of the extraction is a huge factor. Assuming one grinds, doses, distributes and tamps in precisely the same manner each time, I bet the shot time needed to extract 1.5 oz/25g of espresso could vary from 25 seconds to 90 seconds depending on how fast or slow you brought the pressure up at the beginning of the shot. As I've said in many boring posts on alt.coffee, that is the significance of water debit. After the first few seconds of the extraction, your machine's water debit is irrelevant (or very nearly so).

This also may be the reason for Petracco's mystery graph in Figure 7.1 of "Espresso Coffee -- The Science of Quality." He claims that the flow rate is less at 7 bar than it is at 5 bar. I think he's wrong. I believe he's confusing the initial rampup to pressure with the subsequent steady-state pressure.

Can't blame him though, the poor guy's undoubtedly using one of those obsolete fixed pressure pumps instead of a modern profiling pump. If only he'd ask, I'd send him a Schectermatic.

I don't start timing until I see the first drop -- that takes care of all the ramp up stuff.

The reason I went to this is as I dropped the dose in my experiments, keeping the flow characteristics (after a drop appeared) about the same, the dwell time got longer longer and longer. At 19.5 in the triple, I averaged around 4 seconds, at 12 grams I averaged around 8 seconds. This is a machine with a 0.9 gicleur (may as well be none) and no other preinfusion control.

Here's the graph:

another_jim wrote:I don't start timing until I see the first drop -- that takes care of all the ramp up stuff.

The reason I went to this is as I dropped the dose in my experiments, keeping the flow characteristics (after a drop appeared) about the same, the dwell time got longer longer and longer. At 19.5 in the triple, I averaged around 4 seconds, at 12 grams I averaged around 8 seconds.

Argh. I would have predicted just the opposite. Is this due to faster pressure ramp up with larger dose?

RapidCoffee wrote:Argh. I would have predicted just the opposite. Is this due to faster pressure ramp up with larger dose?

Me too.... but you have to grind finer to keep a comparable time... I guess grind trumps dose in dwell time.

I haven't figured this out yet, and I haven't posted the data before. It is a solid fact; I've observed it countless times in regular shot making. None of the doses I charted touched the shower screen, so that's not part of the effect.

I also think it is simply finer grind. My hypothesis (until someone tells me something different) is that the flow is determined only by the amount of fines. A lower dose requires a disproportionally finer grind to produce the same number of fines, and that should mean the water gets through the puck more slowly.

The alternative is that it's simply a question of filling up more headspace over the puck. If someone will give me set of baskets, identical except for depth, I'll test which one it is.

another_jim wrote:I got roughly linear changes too when I systematically changed dose and adjusted the grinder for similar timed shots.

another_jim wrote:I also think it is simply finer grind.

Sorry, I missed that. Thought you were just changing dose. Duh.

another_jim wrote:I haven't figured this out yet, and I haven't posted the data before. It is a solid fact; I've observed it countless times in regular shot making. None of the doses I charted touched the shower screen, so that's not part of the effect.

I also think it is simply finer grind. My hypothesis (until someone tells me something different) is that the flow is determined only by the amount of fines. A lower dose requires a disproportionally finer grind to produce the same number of fines, and that should mean the water gets through the puck more slowly.

The alternative is that it's simply a question of filling up more headspace over the puck. If someone will give me set of baskets, identical except for depth, I'll test which one it is.

jim

i realize i'm a month late replying, but could you mimic "same basket/different depth" by either using a gasket of a different thickness OR using a smaller gasket along with something to wedge under the "ears" of the portafilter, thereby making the basket effectively shallower?

scott

bigabeano wrote:jim

i realize i'm a month late replying, but could you mimic "same basket/different depth" by either using a gasket of a different thickness OR using a smaller gasket along with something to wedge under the "ears" of the portafilter, thereby making the basket effectively shallower?

scott

That could work; but I advise wearing a mask while you work out the bugs