Coffee Data Science
Particle Analysis by DiFluid Omni
Revolutionary for the price?
DiFluid has come out with the Omni which analyzes particle distribution and coffee color. They sent me one to analyze, and I have done a few tests (see the very end for more comment). The first two sets of tests are just to understand the repeatability of the device for particle analysis and coffee color. I have some other experiments later. This article focuses on just particle analysis.
Overall, the device is a good user experience, which can be expected based on their refractometer (R2), and they are very interested in understanding how the Omni can be used regularly by baristas.
With respect to particle analysis, I am waiting for the ability to export raw data before comparing it to my image based particle analyzer. However, what I can say is that the number of particles in the given area is small.
Typically, I measure between 1,000 and 10,000 particles when I generate a particle distribution on a piece of paper, while with the Omni, usually I get 400 to 800 particles for an espresso grind. For the 80 particle distributions I have measured, I can see some particles stuck together. This isn’t their fault; measuring particle sizes with good separation is quite difficult. However, it may require multiple readings averaged to better understand the particle distribution of a grind. This is an inherent limitation due to the size of the device.
Repeat Testing
I started testing by doing 10 samples of the same coffee. I looked at the average and standard deviation. There seemed to be a small amount of deviation in the D50 reading which is the middle of the distribution (50% of the particles are greater than D50, and the rest are less than D50).
This was also true for the individual bins.
Buzz
The Omni detects if the particles are too clumped together, and it audibly buzzes to separate the particles. I looked at what would happen if you forced it to buzz a few times. While the buzzing helps, it can also push particles to the edge of the measurement surface.
This had some impact for these measurements, but it is unclear if this difference would greatly impact the use of the Omni.
The cumulative shows the shift. It is a little weird, but the number of particles goes from 604 to 648 by the end.
The STD stays about the same, but the particle count and D50 change quite a bit especially on the 3rd and 4th buzz.
Buzzed vs Dusted
For my own data collection, I use a brush to tap the coffee grounds to separate out the fines that are stuck on larger particles. So I compared using the buzz feature vs the dusted technique. Visually, the buzzed has some grounds clumped still.
I look data across 23 shots, and I found the distributions to have a variation.
This difference was clearer in the measurement of the 300um and 400um bins.
The difference was more pronounced in the cumulative distribution at these bins.
From a statistical level, the difference between buzzed and dusted was statistically significant.
I don’t know if dusted is better. I have some visual evidence to indicate that it is, but I need the raw data from the Omni to better confirm and compare to my technique.
Overall, the Omni is a very interesting tool, and it could be revolutionary because it is at a very accessible price point. I haven’t seen anything in this analysis that gives me cause for concern except for the particle count. This might be more of a problem for coarser grinds. However, you can compensate by collecting more samples.
For regular uses, I think it could be helpful in dialing in coffee. My dirty little secret is that I don’t dial in my coffee; I use the same grind setting for all.
The Omni could be helpful in analyzing grinds or burr sets quickly, but I think it needs more shape analysis to give a better understanding of how the particle shapes differ between grinders. The images are there to do this type of analysis.
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
Omni didn’t offer payment or even ask explicitly for a review. We have had a relationship before their two refractometers came out, and they also sent me one to test. I’m very interested in these tools for what they could contribute for coffee science, and I believe data on their performance will speak to that end. They also showed me a demo before their release of the Omni and asked for feedback. I get no financial contribution from them and no profits from their devices sold.
