EddyQ wrote: ...the IR energy the sensor collects is also includes the surrounding background causing temperature shifts. In the case of the Bullet, their sensor must be collecting IR from the hot drum along with the beans. Maybe they are doing a correction, but it isn't simple to do.

FYI, Aillio have put up a very high level description and examples of the raw IR data and how they filter it here:
https://medium.com/@aillio/the-start-of ... aa01d08fa9

I'm not a bullet owner, but at face value, IR bean temp appears a step forward to me (assuming it works as marketed). The simple reason is it should improve repeatability as more timely data should allow more timely changes to react to what's going on.

I think that in general there are many further gains available in roasting (on any machine) by (a) reducing lag for both inputs and outputs, and (b) using machine learning to predict the future where lag which can't be removed. Both should help roaster operators to be able to more repeatably produce the specific roast they want to.

Coffee roasting is not something that benefits from automation; at least as far as optimum flavor goes.

For the photography buffs among us, it's akin to batch processing your image files with an auto enhance feature vs doing custom adjustments to each image. The former gets you in the ballpark. The latter gets you a better result...assuming you know what you are doing. A really good photographer switches the camera to manual, measures the light and tries to capture the best light, composition and moment "in-camera". They do not like spending a lot of time in front of a computer trying to "fix it in post".

As far as I know, IR sensors for coffee require algorithms to process the data. So it really doesn't matter how fast the response. That is secondary to what the algorithms are doing.

The Zone system is an algorithm Ansel Adams masterfully used to creative effect. You are using algorithms of roast charting and BT curve design and don't have an automatic way to get there. IF you had an automatic way to plug in EOD, RoR at start of 1C, dev time, with an algorithm that could hold that roaster off of crash and flick, wouldn't pursuit of optimum flavor quite benefit from automation?

It hasn't yet. But I'm not inclined to care. I like roasting too much.

Calling the Zone system an algorithm is more than a stretch.

Obsession wrote:FYI, Aillio have put up a very high level description and examples of the raw IR data and how they filter it here:
https://medium.com/@aillio/the-start-of ... aa01d08fa9

I have read that article and it is not correct. Knowing the emissivity will not guarantee accuracy. You also need to know background temperatures. And I don't think they are measuring the drum temperature or drum end temperatures.

Almico wrote:It hasn't yet. But I'm not inclined to care. I like roasting too much.

Calling the Zone system an algorithm is more than a stretch.

It is not a simple algorithm, however a system of rules it is; it still uses the data of light metering, film sensitivity, minutiae of the development process in a systematic way in pursuit of a photographers artistic vision. I did not say visualization was an algorithm. The devil's advocate in me came out when I read your statement.

Coffee roasting is not something that benefits from automation; at least as far as optimum flavor goes.

To pursue optimum curves using automatic data acquisition and recording is taking advantage of automation to optimize flavor.

Are you saying that (a roaster like) the Bullet takes away a roaster's creative choices in roasting relative to less automated roasters like the Quest?

EddyQ wrote:I've lurked over at the Bullet FB site for a while. There sure seems to be issues. Firmware updates, panel errors and other things going wrong. Now, maybe it is just a spot all dump their issues. But I haven't had any reasons to fuss for over a year or more (just sayn). That said, replacement parts for the Bullet do seem well supported and owners don't seem upset with reliability.

Beyond price, this is why I am hesitant about the Bullet. When you buy hardware that is tightly integrated with software, you become reliant on the company that sells the product in a new way. Software is like a garden-- you need people actively working on it to keep the weeds at bay. If they go out of business or lose interest in supporting your hardware, you're often out of luck. You need to factor this in when you consider the likely lifetime of the product. It does seem like the folks behind the Bullet are well incentivized to keep the lights on, though.

I don't think the coffee community has a lot of exposure to this problem as the vast majority of coffee hardware for sale does not need software to work. I do wonder if/how attitudes may change over coming years with more and more products like this entering the market. Eventually, somebody is going to go out of business and leave their customers in the lurch. By comparison, a roaster made mostly from sheet metal and off-the-shelf components seems like a safer investment, even if it is a little less cool.

If you learn how to drive them properly, you should be able to make excellent coffee on either machine. And I haven't used either one, so take everything else with a huge grain of salt.

With the disclaimer out of the way, I think the benefits of IR bean mass temperature measurement remain unproven. Bean mass temp is an average of a lot of noisy data, and the more precisely you measure it moment to moment, the more you risk confusing the noise of individual bean temp variation for the signal of overall changes in bean mass temp. Thus, I think there is a happy medium of smoothing + probe thickness that works best for managing roasts in real time, and I'm skeptical of the benefits of getting a measurement so noisy that you are consciously second-guessing whether an apparent crash or flick is real.

Also, the more responsive your roaster is, the more you risk becoming a reactive roastmaster, constantly making little adjustments in order to get pretty looking curves. In my experience you get better results by steering a roaster proactively, and living with the results, rather than trying to overcompensate for what has already happened as a roast is in progress. That isn't to say that you cannot roast proactively on the R1, just that it might encourage bad technique in those still learning the craft.

And on a more general note, if you look at the comments from experienced roasters, you will likely see less and less emphasis on comparisons of the taste profile that different roasters produce, and far more attention given to usability issues. I could roast pretty awesome coffee on my old HG/BM setup, but it was an ergonomic nightmare compared with my USRC sample roaster. The best way to make your roasts taste better is to get lots of practice roasting, cup the results, and learn over time. The more you enjoy working with your equipment, the more likely you are to do extra roasts for practice rather than the bare minimum you need to stay caffeinated.

Chert wrote:To pursue optimum curves using automatic data acquisition and recording is taking advantage of automation to optimize flavor.

Are you saying that (a roaster like) the Bullet takes away a roaster's creative choices in roasting relative to less automated roasters like the Quest?

I think the issue is that, for best results, coffee roasting heat adjustments are as a rule made lower not hotter. Typical PIDs programed to follow a curve will increase and decrease heat depending on current conditions. I learned on my nimble air roaster that coffee does not like that.

The algorithm needed to roast coffee will have to take into account the coffee moisture, density, ambient conditions and age of the bean in order to accurately predict...in advance...the minute heat reduction adjustments that need to me made to keep and RoR curve spinning gently and smoothly downward, while the coffee is fighting that every step of the way. When that happens, I'll give it a try. But from my experience, and talking with several coffee professionals, we are not nearly there yet. Even the venerable Loring leaves a lot to be desired when switched to auto mode.

And it certainly removes the process from the purview of craft roasting. Like throwing a hunk of wood into a 5 axis CNC carving machine compared to a few well-sharpened chisels. One is a craft, the other is production. Neither is right or wrong, just different. And if you are auto roasting, you really shouldn't be calling yourself a small-batch, craft roaster.