AndyS wrote:

So far, though, IT ROCKS, BAYBEE!

Yea, that's what I thought

Day 10:

OK.

This is where Dan starts calling me crazy. It's going to be like the "haunted portafilter" episode all over again.

So...

I don't know if this has been implied by people in the past or just inferred by everyone, but the assumed idea about brew temp for espresso has been that there is a single "sweet spot" and that distribution outside of that sweet spot forms a bell curve. In other words, quality falls away from that sweet spot gradually at first and then rapidly as one gets further from the target.

When I started testing out the results of incremental changes in temp I noticed something that made me think that this might be wrong. Given that it was going to be a controversial statement, I double, triple and quadruple checked my results.
So I just finished yet another run through the various temps.
I've done this with two different coffees (thank you Stumptown - thank you Artigiano; I can't imagine what I would do if I had to pay for all this coffee).

The results are illuminating and somewhat shocking to me.

While there is a bell curve distribution at a gross level, at the center of the bell curve and at a granular level the model fails.
What do I mean?

If you're looking at the results of changes of brew temp at a 3F level of granularity, there is a clear bell curve. Even at a 1F there is a bell curve of quality. But at a tighter resolution, I start to find multiple sweet spots - each with their own distinct flavour profile.

Here is an illustration.
Using the Stumptown Hairbender, I found that shot quality fell off dramatically once one hit 196.7F on the low end and 199.4 on the high end. So let's look at that "center of the bell curve". I incremented the machine by 0.3F steps. For each one, I scored shots on Body, Clarity, Sweetness, Mouthfeel, Balance, Flavour, Finish and Aroma. I threw away the lows (barista error) and averaged the remaining scores for each category. I then totalled those scores to get a final result for each temp.
(Maximize your window to see columns aligned)
full results here

You will notice that there are a couple of distinct sweet spots with the central target zone of the bell curve - each with their own distinct flavour profile. Depending on your personal taste, one of these will most likely be your own favorite - but that doesn't mean there is only one sweet spot or that the bell curve holds at this level of granularity.

For example - I found shots pulled at 197.3F to be lovely with great sweetness and fruit and a lovely finish. These are lighter shots, less concentrated and "intense" in flavour. If you're looking for an intense and powerful shot, 199.1F yields bold and powerful flavours. If you're looking for balance and polish, then 197.9F is probably going to suit you best.

This has implications when it comes to drinks consumed as well. Shots pulled at 197.3F are less suitable for use in milk drinks I find. Shots pulled at 197.9F are wonderful in cappuccinos whereas shots at 199.1F are great in lattes.

Perhaps of even more interest to me is the "bad spots" on the temp curve.
There are noticeable issues with certain temps - even though they might fall between the desirable temps.
This was what drove me to test and re-test and to duplicate the results with other coffees.
At first I figured it had to be barista technique issues but it became clear that it was either incredible coincidence or reality. As I re-tested, it became clear that it was consistent and repeatable.

I'm sure this has some incredible implications when it comes to brew temp profiling. But right now I can't get my head around it.
Does it mean that a very, very stable and flat brew temp profile really is the answer? This would allow you to avoid those "off" temps and stay within a discrete sweet spot.
Or does it mean that a changing brew profile is the answer as it will allow you to combine positive attributes from various sweet spots across the top of the bell curve?

Right about now I really badly want to be able to control brew temp profile.


And, for the first time in a long time, I have consumed enough caffeine to give myself the sweaty creepy crawlies.
Yikes.


Andy... Greg... if you have a chance please please check and see if you can replicate this.


Insane.

(Please note that the scores above are not in relation to some absolute scale - i.e are not compared to other coffees - but rather purely self-referential - i.e. compared to other shots in this experiment - and are thus totally arbitrary)

Great post! This is really an endless search: once we have a machine can control profiles, the new area of exploration will be ascending v. descending profiles, and gauging ideal temps at various points in the extraction profile - maybe a coffee is best at 197.6 in the beginning of the pull, 199.2 in the middle, and 198.4 at the end.

In your tests, when you indicate a particular temperature at which you're brewing, is this the temp that you have programmed into the PID, or the temperature you are indicating at the grouphead? I recall earlier in this thread someone mentioning that there was a consistent 3 degree difference between the PID setting and the water exiting the grouphead (or maybe that was on CG - all the GS3 threads are blending together for me...).

Interesting pattern, every other .3 change was a decrease in quality. So in theory, if I wanted to get a chocolate toned espresso with a bit of fruit I would need to adjust the temp during the shot to span the 197.9 to 198.2 to get the best qualities of both profiles, or would reversing it, high to low, yield the best cup. Just what I needed, another variable, my brain hurts...

So that suggests that a controlled bell curve is not necessarily a bad thing, and Chris can actually be over caffeinated.

skyryders90 wrote:In your tests, when you indicate a particular temperature at which you're brewing, is this the temp that you have programmed into the PID, or the temperature you are indicating at the grouphead? I recall earlier in this thread someone mentioning that there was a consistent 3 degree difference between the PID setting and the water exiting the grouphead (or maybe that was on CG - all the GS3 threads are blending together for me...).

My prototype has the software version that allows for programmed compensation (allowing you to correct for the offset and thus displaying the actual brew temp when you set your target).

Of course... you need to keep in mind that with this machine a temp of 197.6F should actually be stated as 197.6F ± 0.15F.

malachi wrote:Andy... Greg... if you have a chance please please check and see if you can replicate this.

Chris, I've got a Scace device coming from EPNW. Should be here Tues. Hopefully I'll get calibrated up and coffeed up and give this a try by next weekend. Meanwhile, I found your post fascinating.

At first I really thought I was going crazy.

I'm really interested to see what your results are like.

AndyS wrote: So far I do notice a really heavy, smooth mouthfeel. I feel the flavors seem to be more recognizable and distinct, maybe this is the "clarity" that Chris talks about. Today I tasted a really neat butter flavor that I'd normally associate with a California Chardonnay, but there it was in the espresso. Also, changing the temp a degree seemed to make much more of a difference on this machine than with the Siliva.

Theory on the mouthfeel... a combination of the low water debit and the temp stability (increased emulsified oils).
Are you running with the auto mode (and thus pre-infusion) or continuous brew?

The distinctness of flavours is, indeed, what I mean by "clarity". It's pretty cool, isn't it? You get to find all these varietal characteristics that are usually lost and blended into one monolithic flavour. Love it.

You're right about the massive difference a single degree temp change brings. I'm actually noticing huge differences at even 0.3F changes. And that is an eye-opening statement.

how many shots did you pull at each temp, and was the taste profile consistent across those shots?

I pulled shots over three sessions for each coffee.
In each case I pulled four shots at a set temp (unless I had a shot that either channelled or looked wrong in which case I'd discard and repeat). In most cases I was able to count all four of these shots though, as noted earlier, every once and a while I would get an "off" scoring shot which I would then discard when calculating averages.

I probably should have pulled more shots at each temp but I don't think I could have coped with that amount of coffee. Plus, as it was, I burned through something like 3 pounds of each of the Hairbender and the Artigiano Espresso.