Neal Wilson ✅ @neal@social.typica.us

If you're interested in more on how I figure out how to roast different coffees, I have a little series of case studies on this video playlist: https://www.youtube.com/playlist?list=PLPTTFznZ65M-lSNokmiP2uByEoTZAg4oI

After a production test batch, that's brewed as a customer might enjoy it as a final check. This final check almost always passes, but if it doesn't it's important to figure out why and try again.

If I don't find what I want, I'll at least have a good idea of the approximate roast level that I want and enough information to decide how I want to approach a 2nd batch roasted with different time to milestone parameters (that is, taking some portion of the roast and going through that faster or slower), pulling samples from that batch in a narrower range, repeating as needed until I'm satisfied with the results.

After cooling, those samples are labeled. Typica recorded where I pulled each of these samples so if I find what I want I can look up exactly how I roasted that and attempt to replicate that with a larger batch on the production roaster.

In this case I expect that I'll want something that is on the lighter side of what I'd consider a medium roast or lighter, so most of these samples are pre-2nd crack, but taken above the temperature where coffee is invariably grassy. That consideration is going to be different for every coffee based on how pre-purchase samples performed, the intended use of that coffee, and how I envision fitting this into a product line.

The thing that's going to make the single largest flavor difference is roasting lighter or darker, so on this first batch I pull samples at various roast levels. Ideally this covers a range such that on tasting it's obvious that it runs from too light to too dark.

One of these days I'll upgrade the data logging on the lab roaster. The hardware is only a 10 bit ADC with suboptimal full scale range getting fed from a signal conditioner with too broad a range. Channel isolation is bad and signals are noisy. I needed to bust out the math to get this working acceptably, but it would be better to just swap in better data acquisition hardware.

As usual, I start with 2 pounds at the lab roaster.

It's not the nicest looking green coffee, but it can taste great.

Time to figure out how I want to roast this #coffee.

Coffee arrived. Unfortunately I wasn't here to tell the driver to not block the door. That makes it harder to bring in and it's a potential safety issue.

running cat is loud

daily cat

Google has been pretty messed up the past couple days. YouTube says that on the 9th my videos only had 21 views (that's missing somewhere between 100-200) but AdSense is claiming that on those 21 views they served 253 ad impressions (for an average of just over 12 ads per view, which I'm pretty sure didn't really happen). The view count/watch time for that date used to be higher (in line with what's normal for my channel).

I'll probably ditch that platform eventually.

cat

A web server that only returns status codes 400 and 404, but does it very fast.

boost.beast is actually pretty nice.