Firstly, as I have written before I am quite happy with the results I have been getting from my Coffee-Tech Maggiolino. The roaster does have a characteristic that I have to compensate for and the thought keeps gnawing in me that perhaps there is a more elegant solution to the problem.
As you can see in the upper diagram the temperature at any given setting undergoes a sine-wave variation as the electric heater is cycled from full on to off and on again. The recording was made using GreenBean's OCR monitoring software, the drum was empty and the thermostat was set to 150°C. The picture is so small you probably can't read the scales, but the swing is from 139°C to 159°C. The peak-to-to peak time is 8 minutes. This very large hysteresis is probably the result of a number of factors; the slow thermocouple, the slow reacting temperature controller, the slow reacting heating element, placement of the thermocouple above the bean mass and so on. I understand that with this type of setup some hysteresis is unavoidable, but this does seem extreme.
I compensate for this problem by increasing the temperature to 190°C while the temperature is on the upswing, dump in the beans and then the increase is smooth (heater remains on). At higher temperature the bean mass keeps the temperature more stable, but of course the reaction time is still very slow. The curve at the bottom shows a typical roast and as one can see it is quite smooth and predictable. The heater gets shut off at 189°C and I let it coast into second crack before ejecting the beans.
My question is obvious, I guess: How can I improve this situation? Using the same temperature probe in the same position, would a PID be able to reduce the amplitude of the swings? Or would controlling the voltage to the heater be a better solution? Or do I really need to go for a different type of thermocouple or probe, perhaps positioned differently (not a simple job!)? Or should I just leave it as it is and stop thinking of it as a problem?
If anyone has any thoughts on this or suggestions I would be most indebted.