Jonk wrote:It should be easier for a PID to control an air roaster like that. But even those seem to struggle a bit (Ikawa, Kaffelogic). I believe Scott Rao posted on Instagram not so long ago about how he'd never seen an automated roast work well. But hey, if it's close enough I'm all for it.

Well, I have to disagree. Pleas look at the curve for a PID roast with Artisan 3e at the moment 19:00. Here it is:

As you can see the BT curve follows the background curve exactly.

Now, see at the PID roast curve of Kaleido:

The BT curve is unable to follow the background curve, And look at this Kaleido RoR curve. It goes all the way down, but then all the way up. It is, actually, a terrible roast.

However, I think that it is just because the Kaleido technician is not experienced. May be by playing with PID parameters they can fix that, but at the moment there is no proof that PID works well with Kaleido. But there is such a proof for Artisan 3e. Sorry for being blunt.

Also, I apologize if I am wrong, and if I am misstating anything. Please correct me if I am wrong! I am trying to choose between 3e and Kaleido, and so far 3e look more convincing ...

Thank you.

Kaleido example profiles are funny, but here I am attaching mine using only 200g of Ethiopia washed Yrgacheffe, have in mind that I am still learning coming from a convection(hot air roaster). The biggest advantage of KR is that they use medium wave infrared heating directly from the source to bean.
For the previous question about the bean source, yes, I buy green beans and from Netherlands and we will exchange experience and profiles.


Greetings!

L&R wrote:Kaleido example profiles are funny, but here I am attaching mine using only 200g of Ethiopia washed Yrgacheffe, have in mind that I am still learning coming from a convection(hot air roaster).

Thank you, Victor! Your example is convincing. But is it a PID roast? I do not see a background curve in your picture.

Cheers!

I enjoy my manual roasts still, not bored enough to use background curve, I am learning and experimenting. In the future will play with background curves for known beans when I have collected good amount of alog files.

Hi Daumier, I think you got the wrong focus. Just look at the hook-shaped BT curve, it follows the background curve all the time. As for other parameters of the curve, it might be caused by different roasting volumes, bean types and even the ambient humidity during roasting.

DaumierS wrote:Well, I have to disagree. Pleas look at the curve for a PID roast with Artisan 3e

I don't understand what/if there's a disagreement. I will say that was a good video and it looks pretty good, seemingly avoiding much of the overcompensation that often happens in automatic roasts. Hard to validate further without actually tasting the coffee.

vulne wrote:Hi Daumier, I think you got the wrong focus. Just look at the hook-shaped BT curve, it follows the background curve all the time. As for other parameters of the curve, it might be caused by different roasting volumes, bean types and even the ambient humidity during roasting.

Yes, but I posted two screenshots. For 3e the curve follows the background curve exactly. Two curves coincide. For Kaleido the curve is below background for a long time, but then it is above the background. One can try to argue that the difference is small, but hey. There is a difference, while for 3e there is none.

Moreover, when PID roasting with Kaleido is trying to catch the background curve from below, it is achieved via a huge jump in the RoR curve which is considered to be bad per Scott Rao's book. As a result of increasing RoR curve the graph for the BT curve is no longer concave down. What I conclude is that PID example for Kaleido is really bad. Trying to follow the background curve, the Kaleido PID algorithm produces a terrible RoR curve even though the background RoR curve does not look bad. More on this: even though the BT curve and the background curves are close, but in order to compensate this small difference that PID algorithm causes the RoR curve to jump from the very bottom all the way to the top.

I think all this can be fixed by proper tuning the PID parameters, which was done for 3e.

I am still considering M10 since it is infrared, but it would be helpful, if the manufacturer tunes the PID parameters and uploads a video with a good PID roast.

You can tune PID using manual rules such as Cohen-Coon https://blog.opticontrols.com/archives/383 or software such as a matlab https://www.mathworks.com/discovery/pid-tuning.html

What I think is that manufacturers advertise the PID feature, and by that they mean that it is possible to push the button "PID" in artisan. But they never show how to produce a good PID controlled roast.

In my opinion, PID roasting is a very important feature for consumers, while they may not realize it at the moment since there are not so many good home PID roasters. But I want to come up with a curve in the Artisan designer, programming the overall and the development time, points where the first (and possibly the second) cracks occur. And with monotonically decreasing RoR curve (which just means that the BT curve is concave down).

At this point I want just to hit the button "PID roast" and leave the machine alone.

This would make possible to find the desired profile at home with just one or two trial roasts. Such a machine would be attractive to those who do not want to invest their time into learning the art of roasting. The majority of people who design the machines and the members of HB are different - we love manual roasting. We enjoy controlling our machines, it is fun. But the general public is different.

I have to confess that I am an amateur, I got my Hottop 2K+ many years ago, and was never able to produce a very good manual roast. You roast only once a week, and the learning progress is slow. So, I stopped roasting.

This is why I am now looking for a machine with a fully working PID control. What I see is that Artisan 3e PID control is working, there is a proof for that. There is no such a proof for Kaleido yet.

It is mentioned everywhere (i.e., in the Scott Rao's book), that for a decent roast we need a monotonically decreasing RoR curve. From Calculus I we know that
1) decreasing RoR curve => the graph for BT is concave down.
2) increasing RoR curve => the graph for BT is concave up.
We certainly see it in the example with the Kaleido PID roast.


We see that from 7:30 till 9:30 the RoR curve is increasing substantially. Bad! Bad! Bad! Why it is increasing? The reason is in that the BT curve from 7:30 is below the background curve, So, the PID algorithm wants it back. The only way to bring it back is by making the curve concave up, so it could catch the background curve. But when the BT curve is concave up, it means that RoR is increasing.

Hence the increasing RoR curve in this example is the result of the PID algorithm. The PID does not really work well.

Some people argued that the BT curve is still very close to the background curve, which might be OK. Nothing of the kind! It is not true that if the two BT curves are close then the two RoR curves are close as well. In fact, we see that the background RoR looks OK, but the actual RoR curve is terrible.

Why don't you contact the manufacturer? They're quite responsive at Kaleido and eager for feedback. You'll likely be texting/messaging with the person who built it in less than an hour.

I'm pretty sure you're right that most of the mechanics are not expert coffee roasters (or even interested in coffee at all). If they can experience the problem or it is pointed out, they likely have the capacity to fix it or figure out which parameters to use to dampen the effect. Overall the machine is designed pretty well (chaff collector, the exhaust fan, batwings cooling/cleaning to name a few). There must be someone walking around there with a bit of knowledge, trying to make the machine/workflow as good as it can possibly be.

For me I'm just using manual mode. There's too many factors involved (ET, outside temp, pre heat etc.) and my quantities are too small to really benefit from the auto roast feature. Eventually I may get to the point I want to just load a profile and get it over with, but that's still months out. Maybe if I get lucky with a good roast profile, I might just do a PID-run in a couple weeks, but there's still too much thing to learn/improve for me now.