The distinction between fast ramp, high final temperature and slow ramp, low final temperature heating profiles is taken from the ongoing research literature. Schenker notes somewhere that roasters use much more complex proprietary profiles; but he's making measurements to prove or disprove hypotheses made about these extreme ramps. This is good science, you find out what happens at each extreme, and assume, until there's contrary info, that what happens in the middle is somewhere in between.

I must be more pragmatic than academic. I can't imagine that 3 minutes at constant 500F or 9-12 minutes at 428F would produce a drinkable cup. Am I mistaken about this? (...So the pragmatist asks, why study a poorly roasted bean?) On the industrial roast profiles, are the beans pre-dried before they enter the roaster? He discusses the moisture issue at length, and pre-drying before roasting (and humidity is all good info too) but it is not explicitly stated for these roasts(or I missed it ... Edit: 3 of 4 profiles start at 150C/302F). I'm always wondering how they execute a good roast so fast. There is a lot of good data here on these 'industrial' roasts as well.

Iirc, the profiles had a soak at the start temperature, then a ramp to the final temperature, and a final soak there. The HTST was roughy 425F to 525F, the LTLT was 325 to 450. These profiles may be schematic, but they are fairly close to the actual 3 minute roasts prevalent in the 80s and the 8 minute ones used now.

I have very little doubt that the vertical drum convection roasters used for mass market beans can produce a better roast than the shop roasters used by specialty roasters. They are fixed cost items, and any quality increase, no matter how expensively obtained, will be repayed instantly by being able to produce the same cup quality with even cheaper beans. After all, it was the move from HTST to to LTLT profiles that allowed roasters to switch from Brazilian Arabica/Robusta mixes to Vietnamese Robusta.

On pages 34-35 (46-47 of pdf) he describes the temperature profiles of the lab and 'industrial' roasts. His Lab tests (HTST an LTLT) used a constant air inlet temperature. Perhaps it was to be more consistent for the lab tests(?) Your LTLT profile looks to be a more workable starting point. (I could get something drinkable from my popper using this as a starting point)

Sorry, I think I confused his lab roasts with what he describes as generic industrial profiles. A straight line heat source, e.g. like most hot air roaster, including the ones Sivetz sells, will result in the usual asymptotic, convex curve for bean temeprature (as shown in the thesis). A soak/ramp/soak profile will straighten that curve out a but reatian the same shape. A pure ramp profile will get a straight line bean temperature.

Many craft roasters, including acknowledged experts Carl Staub and Willem Boot, emphasize that bean profiles should be "slow start/fast finish," i.e. not at all like the asymptotic curve one gets from constant supply temperature roasting. The move to slower, lower temeprature roasts shows that the big roasters seem to agree with this basic observation.

Andy Schecter and I tried straight line ramp supply profiles which gives a very exaggerated slow start/fast finish roast. The results were interesting, but they were mostly overly or green or slightly baked. The sweet spot for a profile of this sort is very, very tight. So the profile for supply temperature is soak/ramp/soak almost by necessity to get a slow start/fast finish roast profile with a usably wide sweet spot.

A lot of us are only controlling one primary variable on our roasts, usually the heater over time. With PID control, we can achieve good profiles with an environmental temperature that never drops. The bean mass temperature profile typically follows a similar profile at a substantially lower temperature. But wouldn't we gain more control if we had controls on the air flow as well as the heater? Perhaps even the bean/drum speed?

Setting aside the bean/drum speed for the moment, I found some discussions where it was proposed that we retain the enviromental temperature - heater loop and then control the air flow based on the bean mass temperature. Lots of room for experimention of course. But are there any general profiles or principles that might be plausibly derived as a starting point for the second, bean mass temp/airflow profile?

I've been thinking of trying it. In essence, the PID control of the environmental temperature would be designed to get roughly the right bean profile, and the air control would fine tune it. I haven't tried it: tuning a cascade of PID loops is tricky, the SCR for the fan is expensive, and I'm skeptical if there's a big quality bump, and mostly, I absolutely hate doing cut outs in project boxes.

I've been doing this with my homebuilt. I can adjust my bean bat speed and convection air flow to speed up or slow the roast without having to effect the ET much. It becomes a lot to keep track of, BT, ET, Bat speed, fan speed and the upper and lower heating elements variac while looking at and smelling the beans.
Ed

I spent the evening starting to review some of the literature on roasting, beginning with Schenker's roasting thesis that was recommended earlier in this thread. Lots of interesting & useful stuff there that I haven't begun to really digest yet. One other paper that caught my eye was a fairly recent patent assigned to Kraft, one of the big industrial roasters. The patent is essentially for a roasting profile, or rather a number of similar profiles, as patents invariably try to make the broadest possible claims. Successful results were claimed for a wide range of beans, including the infamous Vietnamese Robusta.

Several interesting points. FWIW, the patent claims that the profiles can be successfully applied to conventional, commercial drum roasters as well as the large industrial, fixed drum convection roasters. And the profiles are extreme. Or at least it seems that way to me. The patent suggests profiles with a very long (9 - 13 minutes) drying stage, at a low, fixed environmental temperature. Then more or less doubling the ET and running the roast from approx. 350F (bean mass temp) to completion in about one minute. This is a profile that I will try firsthand in the near future just to see if it is drinkable. But does this type of roasting profile make sense? Is that really industrial roasting practice?

Interesting, I have no clue what the object of this is; but that won't stop me from guessing: If they are doing the long dry at around 300 to 325F bean temp, rather than 275 to 300F, they are in the beginning of the Maillard reaction zone, and the whole thing may be more like a malting process, or much talked about flaw of baking, rather than a conventional roast. You create lots of fairly sweet, bready, and malty flavors, then finish the roast to get an acceptable color and knock out the chlorogenic acids. Since there's no acidity to kill off in the first place, this may be one way to get something drinkable from robusta. The high ETs required for such a fast finish may even create enough char on the outside to add some Starbucks flavors too.

I may be wrong, but it's hard to see this as anything except dealing with the coffee beans as a lowest common denominator raw material, then creating some brandable product from them.