The previous installment in this series outlined some challenges that presented themselves while using our Ikawa roasters here at Royal and The Crown. Below, we’ll look at the journey of creating roast profiles and making adjustments to suit many different purposes.
After getting our machines back from their most recent round of servicing, I needed to create new profiles: all my old roasts were created using miscalibrated heating elements and fan speeds. I wasn’t quite starting from scratch, but I was close.
I’d long been considering the viability of recreating a manual sample roast, beat for beat, on the Ikawa, and figured this was as good a time as any to give it a try.
Profile transfer is probably one of the least tangible aspects of coffee roasting, and anyone who’s roasted on more than one machine will be quick to point out how challenging this can be. Complications may include thermocouple placement, drum rotation rpm, air speed, burner placement, material (steel, cast iron, etc.,) and any number of other tangible and intangible elements that prevents an operator from simply copy/pasting a temperature curve.
My first roast tied as closely as possible to the sample roast time/temp benchmarks, but came out darker than expected, so my first 4 iterations focused on first crack temperature and fan speed, the best of which included the highest fan speed after crack began (in green, below).
Using this as the starting point for round 2, my next few roasts began to play with the shape of the curve, primarily during Maillard.
Unable to achieve favorable results either in color matching or cup profile, I eventually decided to abandon the idea of matching end temperature, and lowered my drop temp.
I also tried one roast at the end of the day where I deleted all my data points except the drop temp. This “curve-less” roast (in purple below) didn’t even hit first crack, even though the profile had the same end temperature programmed — proof positive that end temperature is hardly a reliable guidepost if it’s the only metric used to discuss a roast profile.
After another round of tasting using a Colombia Huila as the control green, the unanimous and resounding affirmation was that roast 7 (in yellow) was the clear winner, and even beat out the drum roast in a blind taste test side by side. The roasts hallmarks are a sharper decline into the turnaround (about 25 seconds, down from about 45 on earlier versions) a sharp incline during drying phase, and a very mellow Maillard and first crack. I lowered the end temperature and lowered the fan speed at the end in tandem, and this seemed to give very positive results.
I tossed a few additional coffees into the Ikawa, and ultimately the good results didn’t hold up. It seemed that this profile was tailored too closely to the individual green coffee I’d been working on, and didn’t have enough versatility to serve as a universal sample roast.
Ultimately, the constraints had become too convoluted, the interaction between curve and air and temperature all more aesthetic than practical. So I gave up. I’d fallen into a classic trap; I tried to force the coffee to bend to my designs rather than allow it to roast the way it wanted.
The 6 Minute Sample Roast
While the majority of my focus on prior profiles was often washed coffees, plenty of dry process coffees come through the Crown and I made an adjusted profile specifically for Natural Ethiopias at some point. The combination of processing and characteristic low moisture / high density along with small screen size presented an interesting challenge, and I’d stumbled upon a very workable sample roast profile without a lot of effort. As it happens this was the same profile that brought my attention to the out-of-sync calibration of our Ikawas.
The profile was based on a moderately high charge temperature with as high as possible airflow at the turnaround temperature, followed by a gradual decline of airflow as the coffee approaches color change. Once Maillard reactions began in earnest, the airflow gradually increased in tandem with a slight reduction in the rate of rise, effectively running a 40% / 40% / 20% profile ratio, with about 60 seconds of post crack development and a downturn in airflow throughout first crack to reduce the loss of the roasted coffee (which at this stage is rapidly decreasing in density).
This was also the profile I’d modified to roast with in the hotel room in Shanghai on a V3 at 220 volts, so I figured why not try and perfect this, the legacy roast that had caused so much trouble and still somehow made great tasting sample roasts.
After a number of small adjustments that included raising the charge temperature and the turnaround temp, smoothing the slope during drying, and lowering the drop temp a few degrees, a roast emerged with a 3-2-1 ratio of Drying-Maillard-Development, with a near-universal application.
I’ve called it Crown Standard SR 1.0; it appears in red below, charted against the original blue profile. I decided to make another, slightly longer roast (in yellow here) by dropping the charge a few degrees and adding about 30 seconds to Maillard. It’s called Crown Maillard +30 SR 1.0.
We cupped the profiles above blind against a Probatino trial roast as the control sample. Using the Crown’s cupping form for production/discrimination, each of the roasts was almost indistinguishable in flavor profile from the control sample across a number of different origins and processes.
I think I learned a few things from this process about profiling coffees. One is that perfection takes a long time. In the four months it’s been since we got the machines back from service, I’ve been through a ton of roasts and cuppings.
The best results almost always came after a blind cupping against a well-dialed manual roast from either a sample or production machine. It’s always illuminating to compare machine types and roasting styles, and even more so when the bias is taken out by random placement and coded samples.
I think there’s almost always an aesthetic element to creating roast curves as well… a desire to round off the edges, make something look more like what we believe to be true about stalling, exotherming, or heat deltas (ever tried to roast with a constantly declining rate of rise? — It’s a terribly boring roast to execute, with mixed sensory results, that makes a gorgeous chart in Cropster.)
I think that the Ikawa can be a double edged sword in this regard, offering the user an opportunity to create interesting visual puzzles that end up divorced from practical roasting application (ever try Rob Hoos’ 10 minute profile? It suffers from adherence to theory, with terrible-tasting results!).
I know my profiles look like they stall at the end. But after Color Tracking and cupping and going back to the drawing board time after time to get the right color and development time with a declining fan speed, I found that the results in the cup tend to override my aesthetic sensibilities.
Lastly, getting outside opinions can never hurt. Reaching out and sharing with your colleagues is the best way to bounce ideas and escape myopia. It can be a refreshing and enlightening way to experience other perspectives.