Almico wrote:The software I would want would be able to
read ambient temp and humidity,
hit DE, 1C and drop milestones for time and temp,
while paying respect to RoR.

There would need to be a PID function following the RoR to keep it always descending.

Adjustments need to be made in advance, so it would need to have an AI feature where I could input the particular coffee and batch size and it could learn how it reacted to heat and air changes. Parameters would need to change automatically to batch size changes.

Probe sensitivity would need to be matched to the roasters thermodynamic personality and software.

I would like to be able to input:

coffee
batch size
time/temp DE
time/temp 1C
time/temp drop

and have it execute that while maintaining a straight line declining RoR slope. And maybe max RoR and finish RoR.

This may be OT (so maybe should be a different thread) but perhaps one doesn't need the exact coffee... Perhaps one can assign a particular coffee 5-10 prebuilt templates based on their behavior around 1C... Do they flick or crash? Are they exothermic or endothermic (not as a bean but as a bean/roaster combination). That way a farmer can give 4-5 parameters that could tell a roaster how it behaves during roasting.

Just an idea...

BTW - you can add a humidity sensor. Arpi did try it once and discovered it peaks (obviously) around 1c. Humidity sensor

Coffee is an agricultural crop where every crop is different and every bean from a given crop is different. You could computerize and automate Folgers roasting as the blend is derived from the sample roasting and tasting rooms. This will be a real flop with artisan roasting.

You could run sample roasts to build a specific profile unique to a specific coffee and crop year and then use that to automate larger weight runs. This would still require continuous roast master observation.

ira wrote:Designing software and algorithms to control a roaster does not require knowledge of how to roast coffee, it requires knowledge of how coffee roasts and what matters.

I have no doubt that software to fully automate is here today and very capable at duplicating roasts. But as you can see from the suggestions of the last few posts there is vastly different data and control people want when they roast. There is no recipe or algorithm worth locking in proprietary code which allows us to push a button and be happy with results. It is this type automation that manufacturers seem to be pushing out at the world. I'm not for it. I am not against algorithms and code. The best software would inform users of all available data, what is happening and allow them to change it. It is flexible.

My cruise control on my car does a much better job than I could with maintaining a speed. But I wouldn't call it automation, because I still need to keep a close watch on my speed. Maybe partial automation? I think of it as a knob called speed instead of the traditional accelerator pedal. Some day, you will get into a car, put in an address and sit back while it takes you there. That is fully automated and something I'd rather not have.

Ikawa is a very popular fully automated roaster. And maybe some folks are dialing in roasts to perfection. But I think most are not. Its popularity is not due to it making good flavorful coffee. It is primarily used by professional roasters to sample roast a variety of beans. The samples all are run through roughly the same profile. Push a button for each sample and cup.

EddyQ wrote:I have no doubt that software to fully automate is here today and very capable at duplicating roasts. But as you can see from the suggestions of the last few posts there is vastly different data and control people want when they roast. There is no recipe or algorithm worth locking in proprietary code which allows us to push a button and be happy with results.

I'm was not and would never suggest that software could be designed to automatically roast all or any coffee correctly, I'm suggesting that software can be designed that will let you define what you want, and figure out how to roast it according to your rules.

Ira

Almico wrote:
The evening started out watching a Sumatra roast and discussing how the Loring works. I asked to be able to see the RoR curve and was told that the Loring does some special algorithmic calculation in "real time", but cannot display the curve. That was very disappointing. RH also depreciated the significance of the crash/flick phenomenon, calling it a temperature probe anomaly, and that the way a Loring roasts bypasses the effects. I found that disappointing as well, since I've be proving over and over again to myself just how important a steadily declining RoR curve is to sweetness and dynamics in a roast. More on that latter...

RoR is merely a calculation of bean temp data. While the software in Loring cannot display RoR curve on the roaster's screen but you can (and you should) hook it up with Cropster or Artisan.

I thought so too. And they had a laptop right next to the machine. Curious why they declined to show RoR curves.

Not a difficult calculation either, bean temperature vs time. The lack of this display is strange.

Maybe they feel that it isn't as important as the zeitgeist says ROR is.

Declining ROR was not called this in the past but I remember long ago that Willem Boot said that roast should slow down during development. So I guess the idea wasn't new - even if the rigorous administration of the rule is.

The screen on the S7 displays ROR numerically but not graphically.
Obviously roasting software will draw curves for ya. I use Artisan myself which makes it easy to see everything. T'was actually vary easy to connect, almost plug and play.

Having to work two screens while roasting is not very convenient. I have a hard enough time with one!