---

My bad, I was not clear. The profile above is the designed profile not measurements from a roast. The ROR plot shown is just the derivative of the blue profile curve. The red plot is the original profile you said I should reduce the ROR on near the end. You are looking at the "Profile Designer" if you will.

Trust me, my measured ROR will look much more real :)

---

Latest roast. I need to work on the PID tuning. I need to do a little work on the voltage compensation using the ZC pulse width. I think the pulse width is temperature sensitive. I am going to change to using the mean of the last minute or two as the reference pulse width. By the time I finished the second roast, the compensation was adding a lot to both the fan and heater control values. The PID did compensate, but my manual fan control settings were getting pretty skewed.

I will work on getting this data into Artisan for easier digestion for everyone who is not me

Also going to add markers for the crack beginning and ending.

I also want to try the thermocouple averaging idea. I have stainless steel probes, so I can't try the super cool twisting the thermocouple wire idea. But I have multiple probes and channels.

---

That is a nice looking curve, far better than anything I've been able to achieve manually. Are you noticing any better taste from your roasts? If nothing else one would be gaining consistency.

---

Thanks! It is a lot better than anything I have done manually as well. Still some work to do tuning things. You can see from the ROR there is still some ringing in the response.

I am not sure I have a refined enough palate to really say the coffee is better than it is from manually controlled roasts with somewhat wobbly curves. It is definitely better than the roasts when I just lost it and everything went south. Which did happen more than I want to admit.

The consistency is much better. For a given coffee, first crack always happens very close to the same time, every time. It is also much easier to pay attention to what is happening with the coffee and when to drop it.

I think it makes it easier to play with development time and final temperature also. All you have to do is watch it and stop when you want to.

I find that development and brewing method both have huge impact on flavor. It seems like the exact shape of the curve has a more subtle effect. But to be honest until now I did not have good enough control over the roast to explore subtle changes in the curve.

I think I kind of just got lucky with the curve I drew to try out the PID, because it is making really good tasting coffee. At least to my liking. I am probably roasting darker than a lot of people on this forum.

---

I've read through this thread a bunch of times and I think I understand it. Basically:

- RobotDyn dimmer is used for zero-crossing in order to time the phase angle pulses
- two opticouplers are attached -- one to the transistor used to control the heater and the other the transistor for the fan, have to use resistor for 12v on the side of the opticoupler to the roaster and different resister on the side towards our microcontroller for 5v/3.3v


Would something like an ESP32 be fast enough to run the loop to handle this? I assume you can hook up the zero crossing pulse to an interrupt? Some of the ESP32 variants actually have two CPU cores and I wonder if it might be possible to dedicate one to the lower level "talking to the hardware" part with the interrupt handler and the other to adding some kind of API exposed via wifi/BT to interact with the roaster from an app running somewhere else (ie phone/laptop/etc).

Am I right in understanding on the lower level, the zero crossing happens and the pulses of the correct duration have to be triggered every time to maintain the current fan/heat settings (or be slightly different to alter them)? I'm a software developer who can solder and tinker with microcontrollers but on a fairly superficial level so that's the perspective I'm working from (ie I made a bread dough warmer with an ESP32, temperature probe and an SSR).

After using my SR800 for a while, I'd love to be able to hook in like this with fairly minimal changes to the roaster itself.

---

It is more than fast enough.
Such things are done with Arduino Uno for more than 10 years (see TC4 project), and Uno is 10x less capable than an ESP32.
Just a thing to be aware, if using WiFi some timers and ADC logic are disturbed, so better use BT (BLE) for wireless comm, if you intend.

---

Roaster is working pretty well now. This roast was outside in a mild breeze and includes me reducing the fan from 60% to 35% in ~5% steps. I made a modification to the integral equation. I don't allow the integral to change if the error term is getting smaller. This prevents the integral from winding up when the error is large but headed in the right direction. The change got rid of the large overshoots.

---

I read through this entire thread and I’m amazing what you all have done. It sounds like you have in depth knowledge of how the Fresh Roast works. I have a couple questions. I’m new to roasting but a long time tinkerer. I’m using the SR540 w/ext tube and also getting frustrated with the stock control limitations. I don’t have the controls experience to do what you all have done, but as the OP discussed to start this thread, my inclination is that a separate manual fan control would give much better control.

1) Do you think a separate manual fan control is at all possible and would be a good 90% solution?

2) Do you know how this could be done? Would it need a different fan because of how the fan is powered?