another_jim wrote:In fairness to the GS3, this test is "cooked" towards the E61... So the blend should "fit" his machine to a tee.

Agreed, plus the blend is reasonably temperature tolerant (202F - 204F), which doesn't showcase the GS3's more precise temperature control capabilities. Those who have a penchant for demanding ("primadonna"?) espresso blends would undoubtedly see a more distinct difference.

Dogshot wrote:Dan, I get the sense that you have taken a stance as an HX-advocate in these days of the DB craze.

I am not strongly pro-HX or pro-dual boiler. Both can deliver great results and my goal is to help owners get the best out of each. Some adopt one or the other as "thee answer." In my view, Nature rarely conveniently optimizes things to work best one and only one way.

Dogshot wrote:In your opinion and given this perspective, does the GS3 offer the home-barista much more than the other DB machines like the Brewtus or the S1, and if so, in what ways does the GS3 make things easier/more consistent/better than these machines?

Let's be real... the GS3 easily outperforms the S1 and Brewtus on multiple fronts. Walk-up brew temperature precision, sub-degree temperature adjustability, high forgiveness factor, steaming that blows you away (2.5L boiler running at 1.8 bar). The form factor is superior, it's semi-portable, it's whisper quiet (Brewtus is among the loudest vibe pump machines I've used). And that's only a short list off the top of my head. Chris' conclusion gives more reasons why it's being lauded as one of the baddest espresso machines on the planet.

Ken Fox wrote:I think there needs to be some blind taste testing done on different machine temperature profiles before we get too convinced that there is anything magical about absolutely flat temperature profiles.

Agreed. Like you, I'm very skeptical that Nature arranges things so conveniently. It's more likely that we've optimized blends to the machine's capabilities than the other way around. Since a flat temperature profile is much easier to reproduce, I agree that it makes sense to optimize blends to it (also see related discussion of the mythical flat brew temperature).

RapidCoffee wrote:I'm no espresso machine design engineer. But (off the cuff response): couldn't you quickly and easily reduce the boiler temperature by introducing cool water?

Or automating a thermal flush? I've pretty much maxed out brew temperature reproducibility on an HX using an E61 thermocouple adapter, but it stands to reason that automation could raise that another notch. I think that the PID that I got from Eric includes a 1 amp SSR. In theory, it would not be hard to fully automate an HX cooling flush to any desired temperature, at least for low-volume use.

HB wrote:This machine puts out a beastly amount of heat.

Dan, I assume that "beastly" means even more than the "considerable heat" you reported on the Elektra A3. Heat is of some concern to me, since my home bar has a western exposure that can get a might toasty on summer days in the San Fernando Valley.

Do you have a theory as to why a 1.5 steam boiler and 3.5 litre brew boiler would put out more heat than a 6 litre steam boiler? I had assumed they would be cooler. But, my knowledge of thermodynamics is pretty limited.

RapidCoffee wrote:Re insulating the GS3 boilers:


In that case, why wasn't this state-of-the-art machine designed for optimal thermal stability with insulated boilers

As Sean Lennon mentioned somewhere, insulated brew boilers make the problem of precise temperature control more difficult. It seems paradoxical, but it's true. OTOH, insulating the GS3's steam boiler isn't going to mess up anything.

Marshall wrote:Dan, I assume that "beastly" means even more than the "considerable heat" you reported on the Elektra A3.

Perhaps not, perhaps "beastly" means the same or less than "considerable."

Marshall wrote:Do you have a theory as to why a 1.5 steam boiler and 3.5 litre brew boiler would put out more heat

You have the boiler sizes reversed. Despite that, I have no theory; I'm waiting for Dan to tame that beast.

Marshall wrote:Dan, I assume that "beastly" means even more than the "considerable heat" you reported on the Elektra A3.

Originally from the Bench writeup:

HB wrote:Even though I had seen the A3 at the Elektra booth at the SCAA conference, I was skeptical of the claimed six-liter boiler. It just didn't seem a large enough case, although the machine is definitely larger than Junior. I doubt no more, check out this bad boy:



Uninsulated... that's gonna generate some serious heat I bet.

Serious heat... beastly heat... considerable heat. I'll write it off to literary license. They both generated enough heat to noticeably raise the temperature of the back of the house (no direct sunlight). The front of our house has direct sunlight and is several degrees warmer in the summer. I haven't compared the two espresso machines side-by-side to pick the worst offender. It would be close considering the GS3's steam boiler is running at a blistering 1.8 bar. It could be incorrectly set, I don't know, that's how it came.

On a related note, I've only steamed a couple cappuccinos, but noticed the heating element doesn't keep up. The pressure drops steadily. Of course it starts out so high it wouldn't matter except for steaming buckets of milk. We measured the reservoir water temperature too - 115F. That may be a new record among pourover models.

another_jim wrote:Basically, if a machine had the capacity to put a heavy brew head at one temperature, and the water supply at another, one could produce arbitrarily (one peak only) curved profiles. An E61 box roughly approximates this spec.

I thought I was going to be able to manipulate profiles with three PIDs on my Silvia: grouphead, boiler, and feedwater. In practice, it's hard to get everything where you want it to be when you want it to be there. Surfing, always surfing....

Beastly heat, ghastly heat, hell hath no fury heat, you should see how much heat an 11 liter HX boiler pumps out into the kitchen.


But then a 4 group LM would make this look like a heating pad.

AndyS wrote:I thought I was going to be able to manipulate profiles with three PIDs on my Silvia: grouphead, boiler, and feedwater. In practice, it's hard to get everything where you want it to be when you want it to be there. Surfing, always surfing....

I did say it was a one shot only deal, then wait another hour.

In general, I don't think the time is ripe for investing a lot of time in brew profile taste experiments. Ken's dual machine setup is close to ideal for this type of experiment; but his published curves don't really show enough consistency for a short test. I thought about this and the best I could come up with is a bit of a kluge when it comes to the experimental design:

1. Do a lot of shots, rate the shot and record the temp profile.
2. For each profile, derive a curvature coefficient (in a univariate exponential curve fit, the exponent's estimated power works as a single measure of curvature -- requires a transform of the curve).
3. Correlate the curvature to the taste, statistically controlling for average temperature variations.

At least 50 shots are needed if the shot curvature made a big difference; up to several thousand, if the curvature is only a small factor. I don't have the appetite for it. If the shots are paired, and the difference in shot rating were correlated to the difference in the temperature curves, the test would be a whole lot more sensitive, and cut the required number of shots to force a valid result by a factor of 10. But it's still a huge undertaking.

another_jim wrote:I did say it was a one shot only deal, then wait another hour.

In general, I don't think the time is ripe for investing a lot of time in brew profile taste experiments. Ken's dual machine setup is close to ideal for this type of experiment; but his published curves don't really show enough consistency for a short test. I thought about this and the best I could come up with is a bit of a kluge when it comes to the experimental design:

1. Do a lot of shots, rate the shot and record the temp profile.
2. For each profile, derive a curvature coefficient (in a univariate exponential curve fit, the exponent's estimated power works as a single measure of curvature -- requires a transform of the curve).
3. Correlate the curvature to the taste, statistically controlling for average temperature variations.

At least 50 shots are needed if the shot curvature made a big difference; up to several thousand, if the curvature is only a small factor. I don't have the appetite for it. If the shots are paired, and the difference in shot rating were correlated to the difference in the temperature curves, the test would be a whole lot more sensitive, and cut the required number of shots to force a valid result by a factor of 10. But it's still a huge undertaking.

The results I've gotten so far are certainly interesting but not consistent enough, as you point out, for an accurate comparison. After extensive retesting of both machines this weekend, I'm convinced that the arrangement I've been using, a very low boiler temperature with a small flush, is not the final solution to tuning these PID'd machines. It may be that more shot temperature consistency can be achieved by splitting the difference between what I used to do and what I"m doing now, e.g. using a somewhat higher boiler temp and a somewhat larger flush. What can be really annoying is to do a whole series of shots and get maybe all but one very tightly grouped but the outlier is off by 2 degrees from the next closest shot, for no obvious reason.

An alternative approach that I think might have a higher probability of working would be to get the TC probe right up against the heat exchanger so that it more accurately reflects HX temps and gets the element responding to something a little bit closer to shot temperature than is the boiler temp.

As it is, for my own usage pattern which is to seldom make more than one drink in 10 minutes, I"ve got pretty decent actual temperature stability, but there is a fair amount of variability when shots are made at closer intervals, especially with frothing.

ken

annp wrote:My comment was in response to what Dan said:

"A few minutes with the thermofilter confirmed a three degree offset between brewhead temperature and displayed temperature"

Please correct me if I misunderstood your question or put Dan's comment in the incorrect context.

How would you know how to program it without establishing parameters manually?

The same way that you dial in any espresso machine; by taste. When we first set up our Synesso, the espresso that it was producing was incredibly acrid. My boss was clearly not happy at dropping so much coin on a machine to have it produce crappy espresso. Over an hour, we tried a whole spectrum of temperatures and settled on one 4F lower than what it came set up at. Now, with a few month's experience, it only takes four or five shots to dial our guest blends in to the right temp. We're pulling the majority of our shots in the 200-203F band, so our offset is probably more or less where it should be, but the process would be no different if we were in the 120-123F zone and the offset were programmed in monstrously wrong! Dialling in the pstat on a prosumer HX is much the same thing. Pull a shot, taste, turn pstat screw a tiny bit, let the machine settle, pull a shot ... repeat until temperature is right.

The whole 'absolute temperature' thing is something that has never struck me as all that useful as a standard measurement (with the possible exception of AndyS' contraption). There have to be a gagillion things that can go wrong in measuring temperature. Hell, in the thread Errors in Temperature Measurement, John Bicht said

The combination of the instrument and thermocouple errors are as follows.

Keithley 2700 with E type (special limits) = within 2.16 degrees F
with T type (special limits) = within 1.26 degrees F

High quality PID controller
with E type (special limits) = within 2.5 degrees F
with T type (special limits) = within 1.6 degrees F

DVM typical with E type standard = within 6.96 degrees F
with T type standard = within 5.76 degrees F
with J or K type standard = within 7.96 degrees F

... and that's not taking into account methodological differences. The magnitude of the error here is enough that you'll taste it. Now don't get me wrong; I think that temperature measurement is useful as a guideline and to ensure repeatability, but I think that there's a risk in dogmatically saying "the recommended temperature for this blend is X, so I'll set my machine to that" and not learning how to actually tweak based on taste.

Bottom line is that the machine is just a tool. Its job is deliver us a consistent temperature. It is our job to tell it what temperature to deliver, and this is not a job that we should be outsourcing to another machine ie. DVM + thermocouple.

Just my $0.02,

Luca