another_jim wrote:So you watch the RoR twitch around and don't do anything except make mental notes for adjustments to your next roast? What kind of changes do you make to prevent flicks and crashes?

Yes, exactly. When my roast is over I look at it and decide if I want to tweak something or if I think it's good enough. I don't have the palate to be too fussy, so I'm perfectly happy with a "generic" profile and I don't worry about tweaking things for different beans.

If I see a crash I try increasing the temperature setting from 30 or 40 seconds before the crash by a few percent to see what that does next time. If I see a flick I decrease the temperature setting from 30 or 40 seconds before the flick. If the times when my alarms hit to change temperature don't seem to line up with what the beans are doing then I move them back and forth a few seconds to see if I can get the RoR to smooth out better. The hottop is electric so there's a lot of inertia in the burner which keeps my temperature change smooth so I assume that if I just make small tweaks I can't knock anything too far out of skew. It's a slow, iterative procedure.

It took a long time to get to alarm settings that gave me a reasonably smooth RoR curve but I'm patient. Like I said, my palate isn't really discerning at all so whatever comes out of the roaster makes me happy. Sometimes it needs a bit more sugar than other times.

Thanks, that makes a lot more sense.

But I still think that the jiggles in the RoR plots are mostly noise, i.e. not electric noise, but local and transient changes in air temperatures around the BT sensor, rather than anything in the bean mass as a whole. The only way to actually test this argument would be to install a second bean temp sensor, and compare the two RoR plots. If the jiggles in the two plots correlate, I'm wrong, if they don't, I'm right. Any volunteers?

Sure. Not even considering the noise inherent in the sensor you're measuring a constantly changing air/coffee bean mixture. The truth of bean surface temperature is there somewhere but you definitely can't measure it accurately. I'm really interested in the Bullet and its infrared sensor but still I'm sure that's only an approximation as well since the beans are constantly in motion.

Nevertheless, the question still remains as to what the RoR graph ought to look like. As I said before, it seems to me that a reasonable goal is to shoot for equal development throughout the bean. Since the roaster applies heat to the outside of the bean which then has to travel in to the center, and the rate of heat transfer is proportional to the difference in temperature, it seems clear to me that a maximum temperature difference at the start of the roast which steadily decreases as the roast progresses will transfer as much heat as possible to the center of the bean before the exterior reaches the desired roast level. That leads directly to the steadily decreasing RoR idea. If the RoR rises then the outside of the bean is going to reach the desired roast level sooner that is optimal and if the RoR crashes then you're not going to get as much heat into the center of the bean as is optimal.

So in my view the "straight line RoR decrease" optimizes bean development and the slope of that line optimizes flavor.

If someone is going to just say "RoR bad" then I think they have to propose an alternative and back that up with why they say that. Certainly the straight line RoR isn't everything, but from a thermodynamics point of view I think it's the solution for even development from the outside to the inside of the coffee bean.

rmongiovi wrote:I confess I nodded off before the webinar was (half) finished. Does he ever propose an alternative to steadily declining RoR, or is he just arguing that a straight RoR all by itself doesn't say anything about the flavor you're going to achieve? If you don't at least try to avoid the crash and flick then what is the alternative? Don't worry be happy?

I also got about halfway through, so maybe I shouldn't be speaking about this, but it seems to me that one of Morten's main points is that if something has gone wrong in the roast that you have worked out from the graphs, and it is so bad that it is going to cause a taste issue, then it is too late to fix it meaningfully. What you should do is to follow the plan that you had from the outset and let the roast fail, so that you fully understand how badly it has failed.

So, to give a worked example, I am currently roasting with an inlet temperature profile on ROEST, a paddle-stirred hot air roaster. My profile is essentially a flat air inlet temp for the first part of the roast, after which it declines. Occasionally I put in some green, and there just isn't enough power. It is only when the BT reading is near, say, 185C and the programmed inlet temp decline is already well-and-truly underway that it becomes apparent to me that this particular green simply needs more environment temp put into it than the last green that worked well on this profile. At this point, I can put in more heat, but basically, the roast has already stalled. Morten's argument seems to be that instead of flailing around trying to frantically increase the inlet temp at this point in time, I should continue the roast plan and accept that the roast is a disaster, so that I can see exactly how bad my heat inputs are.

I have a graph that sort of demonstrates this. It's a little hard to see with the resolution, admittedly, but you can see that on batch 167 I didn't have enough heat, saw that I wasn't going to get there at about 5:15 and took manual control, which is why the inlet temp (top lighter green) doesn't continue the programmed orderly slope ie because I intervened at that point to see if keeping the power constant might help. So the roast finished at 193.5C BT at 6:03. That roast was light and underdeveloped. So you can see that for batch 168, I corrected for this by moving later in time the point in time at which the inlet temp decline starts. And I got to BT 195.8C at 6:01. This roast was much better, but probably would have been better if I had got it to a higher finishing temperature. That would have required either a higher inlet temp or an even later inlet temp decline starting point, or both. Morten would probably say that if I had simply let the first profile run and let the first batch underdevelop even worse than it did, then I would have probably had a better appreciation of how far out of the ballpark the first roast is, and I would have made a larger correction - probably both increasing the inlet temp and making the decline start even later - and that would have resulted in a better second roast. Morten would probably also observe that my heat changes the second time around didn't really make first crack start all that much earlier, which is a non-ROR indicator that they didn't have that much of an effect.

another_jim wrote:Thanks, that makes a lot more sense.

But I still think that the jiggles in the RoR plots are mostly noise, i.e. not electric noise, but local and transient changes in air temperatures around the BT sensor, rather than anything in the bean mass as a whole. The only way to actually test this argument would be to install a second bean temp sensor, and compare the two RoR plots. If the jiggles in the two plots correlate, I'm wrong, if they don't, I'm right. Any volunteers?

It isn't really about small jiggles. There is plenty of noise in a probe. A crash can last for 10s or until drop, a flick can last from when it starts to the end of the roast, and a plateau/rise can also be 10-30s long. These aren't flash-in-the-pan features. They are overall trends in the graph.

Take for example MakoMo's graph. The plateau before first crack lasts a full 45s and then the change in slope to a relatively quicker decrease in RoR lasts over 1min. We aren't analyzing the jiggles but the trend.

I personally find that roasting arguments that tout themselves to be backed by Science™ and present their argument in a "anyone who has a bare minimum understanding of thermodynamics should understand this" manner to very often cherry-pick scientific principles that support their position while ignoring other ones that could explain an opposite position, making simplifying assumptions, etc. I'm not very far in this one yet but I'm getting that familiar feel from the tone. Not saying this guy is necessarily wrong or Rao is wrong or anything, but I try to take all these things with a grain of salt and see what makes sense in my roasting.

MaKoMo wrote:Not sure it this is a good example ... image

It's an excellent example. The two traces are tightly correlated if you smooth them over 20 seconds; the residuals after that smoothing would be uncorrelated. That's not as noisy as I thought; but you still have to wait 20 seconds to see if a bump or dup is real or transient. This might be a little faster than just eyeballing the trend on the BT readout, but not by much. For a post roast analysis, it might help in locating a change more precisely.

In my roasting model, I allow the thermal mass of the beans to change over time, so that the same amount of heat can change their temperature more as the roast progresses. When I fit the data to determine this, there is a fairly drastic step change at the outset of the first crack, when the beans start steaming, otherwise the thermal mass stays constant. The added activity in the RoR traces just ahead of the first crack may indicate such a change. On the other hand, the RoR amplitude also picks up at the yellowing stage, where my roasting models do not indicate sudden thermal mass changes

Trjelenc wrote:I personally find that roasting arguments that tout themselves to be backed by Science™ and present their argument in a "anyone who has a bare minimum understanding of thermodynamics should understand this" manner to very often cherry-pick scientific principles that support their position while ignoring other ones that could explain an opposite position, making simplifying assumptions, etc. I'm not very far in this one yet but I'm getting that familiar feel from the tone. Not saying this guy is necessarily wrong or Rao is wrong or anything, but I try to take all these things with a grain of salt and see what makes sense in my roasting.

So I finished the whole video and really I didn't find his approach objectionable like I thought I would. I do think that going back to fundamentals of time, temperature, and color before worrying about the shape of an RoR curve is much more helpful in learning from your roasts, especially for those of us who have only been doing it as a hobby for a few years. Lately I've been paying closer attention to those like Hoos that are emphasizing similar aspects over RoR nuances

I feel like it all links together and I don't believe Rao ever said RoR is an end-all-be-all but an indicator of a good roast. You still have to hit your phase, temperature, and color and such. For example, you can have a great RoR and drop at 320F. Not a good roast. I do know one thing, data logging has given us a bunch of stuff to fuss over