La Marzocco GS3 Thermal Performance Quantified

[gscace]

Hi there:

I finally got time over the last weekend to gather some temperature data on the GS3. From my perspective, the GS3 is unbelievably consistent and easy to use. The data shows why. In addition, I have a 2-group PID La Marzocco Linea with which I am intimately familiar. The comparison shown here demonstrates how the GS3 elevates the standard for espresso machine performance.

Here's the data on the GS3 compared to my PID'd Linea 2AV. One graph shows the brew temps recorded using the WBC measurement method, which does not use cooling or heating flushes. This exposes machine weaknesses that must be compensated for in actual use. Notice that the GS3 requires no group heating at all. By contrast, my Linea requires heating of the group, and while the temperature comes to an asymptotic value, it takes a bunch of shots to get there, which makes the machine harder to use in intermittent duty usage (my usual) than the GS3 by far.

I picked two shots at random, one from each machine during fairly continuous duty cycles, to show the brew temperature profile from each machine. I haven't examined a batch of them to get statistics on each machine's ability to reproduce the brew temperature profile shown in the graphs, but I'm guessing from looking at the graphs that the GS3 is a bit more consistent than the Linea.

These results don't surprise me one bit. They underscore the level of ease with which one may brew stellar espresso using the GS3. Given that one's technique is sound and that one has good raw materials, the GS3 cranks out shot after shot with ridiculous levels of consistency.

Results for the GS3 - Brew temperature reproducibility is +- 0.48 degrees F. Stability is +- .87 degrees F. Compare to Linea's respective specs of 1.9 degrees and 1.2 degrees.

-Greg Scace

[malachi]

Wow!

That's amazing. Very cool.

[cinergi]

Results for the GS3 - Brew temperature reproducibility is +- 0.48 degrees F. Stability is +- .87 degrees F. Compare to Linea's respective specs of 1.9 degrees and 1.2 degrees.

-Greg Scace

Greg,

Are those reproducibility and stability scores based on one or two standard deviations?

Are these scores based on just one shot or shots 5-14?

Thanks.

[gscace]

Greg,

Are those reproducibility and stability scores based on one or two standard deviations?

Are these scores based on just one shot or shots 5-14?

Thanks.

2 std deviations and shots 5-14.

[cinergi]

very impressive. The best I've seen.

[AndyS]

Results for the GS3 - Brew temperature reproducibility is +- 0.48 degrees F. Stability is +- .87 degrees F. Compare to Linea's respective specs of 1.9 degrees and 1.2 degrees.

Very very impressive, thanks for posting this.

Still, it makes it hard to see how malachi's tasters can easily pick out a 0.3 F change in setpoint. Given the reproducibility figure, it seems like it would take several shots before they could have confidence in their tasting.

Not doubting your observation, Chris, I'm just looking at the data and asking questions. Your machine may perform better than Greg's for some reason. My GS3 is mostly in the range of what Greg is showing here, with occasional unexplained excursions outside this range.

[gscace]

Very very impressive, thanks for posting this.

Still, it makes it hard to see how malachi's tasters can easily pick out a 0.3 F change in setpoint. Given the reproducibility figure, it seems like it would take several shots before they could have confidence in their tasting.

Not doubting your observation, Chris, I'm just looking at the data and asking questions. Your machine may perform better than Greg's for some reason. My GS3 is mostly in the range of what Greg is showing here, with occasional unexplained excursions outside this range.

Hi Andy:

Don't forget the effects of free convection in the brew boiler. Free convection can show up as temperature differences at the sensor pretty easily. It's of course caused by feedwater influx, heater produced temperature gradient, and cooling near the walls of the boiler. Whether or not the effects are seen at the cake are a different story. There's plenty of mixing that goes on inside the plumbing that oughtta help damp out the excursions you see. And one way to damp out the indicated excursions is to do some signal integrating for the display. Makes folks breathe easier, but hides information. I rather like the fact that there is some noise. I'm used to it and don't get too hung up on it. I'd prefer to do my averaging after getting the raw data.

-Greg

[another_jim]

That's very impressive; best I've seen, and quite probably a new record for espresso machines (unless the Synessos show up better).

It looks, if anything, the GS3 runs a tad hot in idle rather than low -- wonder how that works with a dual boiler and that very lovely exposed head?

The intra-shot stability may be better than stated if that particular slightly rising profile repeats itself; since right now, you are computing it so that only a horizontal straight line graph would be perfect.

I'm trying to think of a statistic which is insensitive to any stable non-horizontal profile, i.e. registers that as perfect stability; but it turns out to be a fascinating problem (in translation that means it's got me running to the time series gurus in the stats department). The machine can be modelled as two sets of differential equations or auto-regressive time series, one for idle, one for shots, with the boundary conditions for each determined by the other one. My gurus tell me there may be simple and robust stability statistics based on how well the machine "wipes out" all dependence on its previous states when doing shots. A result of zero dependence would imply perfect stability.

[AndyS]

Don't forget the effects of free convection in the brew boiler. Free convection can show up as temperature differences at the sensor pretty easily. It's of course caused by feedwater influx, heater produced temperature gradient, and cooling near the walls of the boiler. Whether or not the effects are seen at the cake are a different story. There's plenty of mixing that goes on inside the plumbing that oughtta help damp out the excursions you see.

Hi Greg:

I don't understand. I thought you made these measurements using the Eponymous Device. The device IS "at the cake," so all these effects ARE seen at the cake. [EDIT: At least, the effects are seen on the top surface of the cake.] It's too late for them to be damped out inside the plumbing, because the effects are seen AFTER the plumbing.

If the first shot averages 202.2, the third shot averages 201.3, and the fifth shot averages 200.7, aren't they going to taste significantly different (assuming 0.3F is a tastable difference)?

[AndyS]

That's very impressive; best I've seen, and quite probably a new record for espresso machines (unless the Synessos show up better).

Or that mysterious machine from Colorado.

It looks, if anything, the GS3 runs a tad hot in idle rather than low -- wonder how that works with a dual boiler and that very lovely exposed head?

The GS3 takes its brew water from a spot near the top of the group head, which is said to be the hottest point of the entire boiler system. During idle, it seems to get a wee bit hotter up there.

[Abe Carmeli]

Thanks Greg. It is good to finally see some machine performance data here. Chris' palate reports leave me sometimes confused, but mostly flat out envious . This one, I can relate to. At last, my world is back in order. Now, where did I leave my keys?

[malachi]

If the first shot averages 202.2, the third shot averages 201.3, and the fifth shot averages 200.7, aren't they going to taste significantly different (assuming 0.3F is a tastable difference)?

I have a feeling my PID settings are going to be different from Greg's as I saw tighter averages than he saw. Of course, I have a feeling that he may well have NOT been using a stock Scace device and Fluke which would mean his results are more accurate than mine.

[gscace]

Hi Greg:

I don't understand. I thought you made these measurements using the Eponymous Device. The device IS "at the cake," so all these effects ARE seen at the cake. [EDIT: At least, the effects are seen on the top surface of the cake.] It's too late for them to be damped out inside the plumbing, because the effects are seen AFTER the plumbing.

If the first shot averages 202.2, the third shot averages 201.3, and the fifth shot averages 200.7, aren't they going to taste significantly different (assuming 0.3F is a tastable difference)?

the device is at the cake and is the Ewenohoo device for sure. I was talking about the variance one see on the display, which seems to get pretty well attenuated by the time the water comes out'n da group.

And yes, the measurements suggest that the limit of taster sensitivity to temperature might be more than the 0.3 degrees. I'm guessing it's more like .6 degrees or so (root sum of the squares of 0.3 and 0.5 actually).

-Greg

[gscace]

That's very impressive; best I've seen, and quite probably a new record for espresso machines (unless the Synessos show up better).

It looks, if anything, the GS3 runs a tad hot in idle rather than low -- wonder how that works with a dual boiler and that very lovely exposed head?

The intra-shot stability may be better than stated if that particular slightly rising profile repeats itself; since right now, you are computing it so that only a horizontal straight line graph would be perfect.

I'm trying to think of a statistic which is insensitive to any stable non-horizontal profile, i.e. registers that as perfect stability; but it turns out to be a fascinating problem (in translation that means it's got me running to the time series gurus in the stats department). The machine can be modelled as two sets of differential equations or auto-regressive time series, one for idle, one for shots, with the boundary conditions for each determined by the other one. My gurus tell me there may be simple and robust stability statistics based on how well the machine "wipes out" all dependence on its previous states when doing shots. A result of zero dependence would imply perfect stability.

Jim:

I'll send you the data today.

-Greg

[another_jim]

That's very impressive; best I've seen, and quite probably a new record for espresso machines (unless the Synessos show up better).

Actually, I just checked the graphs of Sean's ueber-Brewtus with the brew boiler/head circulating pump, and it should give the GS3 a run for its money.

Andy, I think he's built a few more than the Colorado outfit.

[another_jim]

Jim:

I'll send you the data today.

-Greg

Thanks

[MOSFET]

I'm trying to understand. And I really am rooting for new ground to be broken. But I'm still confused based on these results. Schomer's "Using this technology our brewing water temperature does not vary more than 3/10th of a degree Fahrenheit during brewing." statement, assuming I didn't miss any retraction to this, would imply more stability than what we're getting here.

The representative brew profiles above both vary by more than 1 degree from the knee to the end, the worse coming from the GS3. What am I missing?

'fuzed,
Keith

[malachi]

Schomer throws out the first 8-10 seconds of measurements (i.e. only measures the temp once temp has stabilized).

[gscace]

I'm trying to understand. And I really am rooting for new ground to be broken. But I'm still confused based on these results. Schomer's "Using this technology our brewing water temperature does not vary more than 3/10th of a degree Fahrenheit during brewing." statement, assuming I didn't miss any retraction to this, would imply more stability than what we're getting here.

The representative brew profiles above both vary by more than 1 degree from the knee to the end, the worse coming from the GS3. What am I missing?

'fuzed,
Keith

It's methodology. We're not ignoring the data in the front half of the shot.
Schomer chooses to ignore the front part of the data becuz the later stuff looks better and lets you talk in tenths of a degree. Data's data.


-Greg

[lennoncs]

Actually, I just checked the graphs of Sean's ueber-Brewtus with the brew boiler/head circulating pump, and it should give the GS3 a run for its money.

Andy, I think he's built a few more than the Colorado outfit.

Thanks for the vote of confidence Jim

Unfortunately, the ueber-Brewtus just pee'd all over my kitchen floor with a gen-2 pump

Heather left a box of cat litter next to the machine for me this morning


As soon as I slide in a new shaft seal I will do a set of tests with a genuine Scace PF to the current WBC protocol to make sure we are all talking apples to apples.


Sean