Coffee Data Science
Comparing Coffee Roast Color Analyzers: Initial Data
The start of intensive coffee roasting data
Over the past month, I have been acquiring equipment to explore coffee roasting from a data perspective. The main analyzer is the Syncfo 4in1 because it has a lot of functionality across green and roasted coffee. I also have the DiFluid Omni and the Lebrew, so I wanted to know how those two devices compared for just coffee color analysis.
Equipment
The three devices have a color calibration plate, but the Syncfo has two calibration plates. The Syncfo has two sides to each plate. One plate is used for the chamber measurement, and another one for a larger plate.
The Syncfo 4in1 analyzes coffee color, weight, density, and moisture. It is considered a standard for SCA, so I’m going to assume it to be the most accurate. This is a large assumption, and maybe we can challenge that some time later.
The DiFluid Omni gives a color histogram as well as the average color. The Syncfo only gives one color value. The Omni can do particle analysis as well, which is an added bonus.
Lebrew gives only one color measurement, but it has the hardware to do similar functions as the Omni.
Data
To start testing, I pulled out the 20 existing roasts that I had to compare the color measurement tools.
Both the Omni and Lebrew had an offset from the Syncfo 4in1 and some transformation.
I compared Lebrew to Omni, and they were a little bit closer in color measurement.
In terms of statistical significance, the differences in color reading between all the devices was statistically significant.
Repeat Sample
One challenge with this data was that the sample of coffee was the same beans, but I poured the coffee sample into another container, so there was some movement in the beans and their orientation. Therefore, some noise could be injected into the sample.
So I took a single sample from one roast, and I mixed it up 20 times, collecting color data. For the second half of samples, I placed the Omni and the Lebrew directly on the same sample of the Syncfo to isolate bean movement.
This shows that the given color reading has a lot of noise based on how the beans are orientated. For both sensors, there is a large range of values.
Lebrew and Omni did not show a significant difference, so the sample orientation may play more into the differences in color reading than anything.
I repeated the statistical tests here:
I also split out the last 10 samples where the Omni and Lebrew were put on top of the Syncfo sample. This had a similar result as the first 10 samples.
One of the challenges in comparing these sensors was that the sample trays were very different in size. The Syncfo 4in1 could fit 40g of coffee, Omni tray could fit 16g, and the Lebrew tray could fit 20g.
I will have more data in the coming weeks as I’m using all three to look at coffee color for some experimental roasts.
If you like, follow me on Twitter, YouTube, and Instagram where I post videos of espresso shots on different machines and espresso related stuff. You can also find me on LinkedIn. You can also follow me on Medium and Subscribe.
Further readings of mine:
My Second Book: Advanced Espresso
My First Book: Engineering Better Espresso
