Coffee Data Science

DiFluid R2 Coffee Refractometer Device Variation

Better quantizing device differences

Robert McKeon Aloe
4 min readJan 17, 2023

By Robert, Joe, and Jeremy

DiFluid has come out with two refractometers in 2022, and the original one had some issues with device variation. For their second device, the R2, we were curious to look at similar experiments. How do the devices vary compared to groundtruth, a VST refractometer, and an Atago refractometer?

All images by authors

The Data

The data was collected using 5 DiFluid R2 devices, 1 VST, 1 Atago Coffee, and 1 Atago RX-5000i. Several solutions were used, each providing different insight:

  1. Sucrose solution (the basis for Brix measurement; well-established normative data; a “clean” assessment of hardware)
  2. Instant Coffee at Filter Strength (low coffee soluble concentration with minimal interference from non-solubles; reduced signal strength compared to instant coffee espresso but relatively low noise compared to real-world solutions as instant coffee is almost entirely coffee solubles — 99.9%).
  3. Espresso (real-world application at high coffee soluble concentration; a difficult testing solution with increased noise but strong signal)
  4. Filter Coffee (real-world application at low coffee soluble concentration; the most difficult testing solution with decreased signal and increased noise, testing robustness of both hardware and software)

All refractometers were zeroed before each experiment. All data was collected at room temperature. A precision scale measuring to 0.001g was also used.


All the R2 devices were close in performance with a similar variation across solution samples.


Espresso samples did not have a groundtruth, but the measured samples had less variation than DiFluid.

Instant Coffee at Filter Strength

Filter strength performance was the main issue with the DiFluid, and the performance was similar to the other tests in terms of device variation and groundtruth.

After the first five samples, the R2 values appeared closer to the groundtruth than VST and Atago. So more samples were taken across the solution made a few more times.

We can remove VST and Atago to look at just the R2 unit variability.

We can sort these series of samples, to give another view.

We can also look at the absolute error given the groundtruth, and the R2 units have some variation but are close to one another unlike the original DiFluid.

Filter Strength

We can look at multiple filter samples across four brews. These were sorted by VST.

We can sort each series of samples, and only one unit gives a much different reading. However, there seems to be a clear offset between R2–1 and the rest.

These results showed less device variation than DiFluid. One caveat is that we don’t know if these devices were manufactured sequentially or randomly selected post-production because there may be a dependency on the production line for device to device variation.

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 Future Book

My Links

Collection of Espresso Articles

A Collection of Work and School Stories



Robert McKeon Aloe

I’m in love with my Wife, my Kids, Espresso, Data Science, tomatoes, cooking, engineering, talking, family, Paris, and Italy, not necessarily in that order.