Here is my profile for a roast I did today of 2 LBS of Ethiopian Yirge - Adado
This is my first roast after I changed my ET probe by adding a 90* bend, it is now 3" into the drum not straight across at the bottom of the bean chute..
I also cut the tip of the probe tubing so the bead is bare, this works much better with my 3/16" probe..



Question:
After reading Scott's Book I'm not trying to get a declining ROR and I'm having trouble at the middle of the finish.. @ 1:20 sec I get a slight drop and 20-30 sec later I get a rise..
How do you all handle this??

I'm struggling with the same thing, JK. My TJ-067 is set up similar to yours, bead thermocouple in a 1/8" L-shaped stainless steel tube with just the tip of the bead exposed, sitting in the bean mass.

I read somewhere, not in Rao's book, that to achieve a smoothly declining ROR, the heat input should decrease throughout the roast. I found that for my setup, that definitely does NOT work. Just as the beans reach 1Cs, I need to decrease airflow (= increase heat input) or exothermic flash, as Rao describes it, causes ROR to drop dramatically. At 1Ce, ROR bumps up again unless I cut the gas and increase airflow substantially until drop.

Here is a roast of 20 oz of a dry-processed Ethiopian. Not the best roast - drying was too fast and I didn't quite tame all the bumps in ROR, but I was able to draw out the development time to hit 25% of the total time, ending at a low finish temperature (too low, actually) without stalling.



As you have no doubt discovered as well, the smoothing settings in Artisan make a huge difference on the shape of the ROR curve. With a very thin, responsive thermocouple, the curve is unusable without some degree of smoothing. How much is too much? For the posted graph, I used these settings (TOOLS>EXTRAS>):



In any case, playing with the roaster to achieve Rao's 3 commandments is an ongoing learning experience and a lot of fun.

Dregs

Opps you do have your projection on .....

I'm not sure how worried I'd be about that blip, JK. The smoothing makes all of that data slightly suspect. Marko is incorporating some changes to Artisan for the Phidget 1048 that will make the ROR calculation potentially more accurate.

It's all good and you can certainly try to nail it down, but there's nearly always a point of diminishing returns where the return dwindles to zero.

Or not if you are having more fun.