Hey all,

I hope all is well. I was hoping you all could give some insights on how you roast with a PID? What technique do you use when you are designing a profile? Do you go off of the last one that you like or what things do you take into account when deciding as there are so many factors?

Here's my first attempt at a PID based profile:



Any and all feedback is greatly appreciated!

This is the roaster that I used for the profile:

Tells more about your PID setup:

• What shield are you using to transduce temperature?
• Where is the PID processing happing - in Artisan or onboard circuitry (i.e. TC4)
• How are your ET and BT probes setup?
• Are you using ET or BT for SV?
• What are your P, I, and D parameters?

Starting from the last question, is the easiest to reply: other people PID parameters will never suit your own machine, if not identical build and beans.
Actually, there another simple reply to everything: don't use PID, is simply not appropriate for this kind of process.

If really interested to go deeper I can elaborate more.

As measurement setups I am using either a Mastech MS6514, either a custom build board that run a fork of TC4 software.

About the strategy what temperature using as profile, use BT only if you are very sure you take it correct and reliable, measuring as much bean surface and less air as possible. If a right BT measurement can't be done, then an ET based strategy give much better results. How to roast with ET as reference, read here:
Timing first crack heat reduction
Don't be obsessed about horror stories about FC flacks and crashes, they are specific to drum machines, not so perceivable in the hot air majoritar convection machines, as FB or hotgun.

PS: I guess the image you posted is not yours, wondering if it is even real, or some sort of simulating, as I am doing using software...
Is hard to believe that starting from such oscillating ET can result so smooth BT curve, and even more, so unrealistic smooth RoR...

I wish people would learn more about how things work before posting nonsense online about how things are fake. Just because you do not understand how it works does not mean it's "not real"... This is how it looks during the roast with zero smoothing...

Definitely we have a different understanding of what we see on that screen.
Wondering why you started this thread if homeworks so well done...

Ryan,

When you say zero smoothing, you mean on ET, BT, and ROR? Those "course-correction" ET spikes are a surprising contrast to your BT and ROR, which are indeed so impressive that they look fake.

I assume that to get this kind of rock solid response, you are controlling to an ROR curve, and not BT?

Please share more of your control scheme. I'm stunned.

Cheers!

- Jake

Ryan,

I'm with Jake. Your curves are so unusual I can see how Renato (and Jake and me) did a double take.

I am guessing that the timing of the environmental temperature bursts produced by the PID creates a steady rise in bean temperature, but you would need computer control to vary ET as much as you do in your video without having BT also dip and surge. And the declining ROR looks so very smooth, which must be your target.

Since you're trying a different control mode than most of us, please provide lots of detailed description. This is a new way at it for me, at least, and is very interesting.

Also I'd like to know more about how you're measuring BT. I do trust that you're hitting 1C since you must be hearing it. The numbers look low compared to what I'm used to seeing. I usually hit 1C at about 385°F BT with my probe placement and drop temperature is usually at least 406°F. So I can see why Renato is thinking that ET measurement may be preferable. The moisture loss figure looks low. I'm usually losing at least 12 - 13%. What do these beans look like when the roast is complete? How does it taste?

And everyone, let's stick to the data and not get personal about discussing it.

Joel, Jake, and Gary-thanks for the responses. I am going to put together a full write up on how I went about this. For now, here is another video showing a full roast in progress. This is with a 245g charge:
Pro-tip: it's fun to watch @ 2x speeds.

renatoa wrote:Definitely we have a different understanding of what we see on that screen.
Wondering why you started this thread if homeworks so well done...

There's always more to learn. That is why I love this forum to learn from others. I have only been roasting for about 3 months so I need all the help I can get

renatoa wrote:Starting from the last question, is the easiest to reply: other people PID parameters will never suit your own machine, if not identical build and beans.
Actually, there another simple reply to everything: don't use PID, is simply not appropriate for this kind of process.

If really interested to go deeper I can elaborate more.

As measurement setups I am using either a Mastech MS6514, either a custom build board that run a fork of TC4 software.

About the strategy what temperature using as profile, use BT only if you are very sure you take it correct and reliable, measuring as much bean surface and less air as possible. If a right BT measurement can't be done, then an ET based strategy give much better results. How to roast with ET as reference, ...

To be clear, these questions were directed for the OP to answer, and I do disagree with your suppositions and most of your assertions, namely to not use PID for "this kind of process", which I infer to mean "roasting". It may be in your circumstance that you've come to that conclusion, but I've been successful for years now using Artisan/PID control in with my North Roaster. If anything, automation is the "thing" now in many other disciplines and I expect it to grow with coffee roasting as PID, ML, DL, and new sensing systems take hold. I do agree with your comments about ET (which is why I asked the question about SV), for this is really the "secret sauce" to get automation right, especially for drum roasting (I have long, long overdue post on this to share at some point).

Ryan, do share with details of your setup per my earlier post. The large swings in ET are most likely the consequence of using BT for SV and falling prey to PID windup because of your integral setting, or perhaps inductive response for the temperature probe. The information you provide, the better in this regard.