I currently roast using an iRoast which I have chopped all of the electronics out of and control using a dimmer for the fan and an Omron E5CK-T PID for the temperature/process (the -T allows 4 ramp/soak programs of 8 steps each; regular E5CKs don't have programs, near as I can tell).

I feel that PID roasting is really the way to go; logging is great, but when you have a solid roasting setup with enough power to do what the PID says to do, logging the temperature curves is pretty much fitting a curve to a bunch of pre-plotted points. You plot the points when you program the PID program, then watch as the bean temperature follows your program. Big whoop.

However it changes the game - instead of battling a complex system to follow your intentions, you now have a control system that will do it all for you. The new game is figuring out the best place to place those points on the graph - how fast to ramp, at what temps to start/stop ramping, exactly where to stop the roast.

While measuring temps and ror are pretty, I think a more useful metric to graph for a PID is the duty-cycle of the pid output. While running a program, the PID modulates the output power, which can be expressed as a percentage. This shows how hard the roaster is working to maintain the temperature curve. If you're pushing 100% for any length of time, then you've hit a rail, and you know that you're asking more from the roaster than it can deliver.

But output power, I feel, will show more. I wish I had the ability to easily graph the output power from my PID, but I don't have the serial option card for it (I have considered hacking together a line driver - I believe the signal is present on an internal connector, but I've been too busy to find the time to play with it).

While one might determine where the sivitz bumps are for a given coffee by applying a fixed amount of heat and watching the temperature curves, trying to map the coffee with the PID by looking at temperature is a meaningless exercise.

On the other hand, by looking at power applied by the PID (on my PID, it's the Manipulated Variable, MV, described as % duty cycle; I can watch the value on the PID display, but no graph) on a linear rise program will show the sivitz bumps - graphing it will show the power applied rising then falling as it goes around the bump.

On the Sweet Maria's coffee list, we were discussing profiles for the Ethiopian Jimma coffee, which was an incredible lot, and a roaster I respect very much posted his profile - it was oddly precise, but following it yielded incredible results. Watching the MV on my PID, I noticed that the stopping point coincided with the tail end of one of these bumps.

I have no idea if this theory is sound, but my thought is that the "sweet spot" for a given coffee might be determined to be just past such a bump. Not necessarily any bump; damn, I need to read more about the specific chemistry of roasting, but I've got a day job

Anyway, just throwing this out there for discussion - and Arpi, I love your project - you should add the MV to the graph, if you can - I'd love to see how it looks on the graph.

Hi Allon.

Thanks for the idea. I added it with an LCD called PID %. I can also graph it but I won't roast anything till the weekend.



I have noticed an increase of Delta BT once the first crack starts (even before audible cracks). It changes with beans. With the new indicator, it'll be easy to tell as the duty cycle should go down.

Thanks for the idea