Coffee Data Science

Coffee Moisture Meters Encore: DiFluid Omix vs Syncfo 4in1

A follow up to my technical review of the DiFluid Omix

Robert McKeon Aloe
6 min readMay 31, 2024

Previously, I reviewed the DiFluid Omix for use as a multi-meter for coffee roasts. I compared it to to the Syncfo 4in1, and they performed pretty similar to each other. Now, I want to focus just on moisture.

Moisture is challenging to have a ground truth in a regular setting. One could completely dry out the coffee beans to get to 0% moisture, and this can be challenging if the relative humidity is not 0%. Similar to my previous exploration of the Syncfo 4in1 moisture measurement, I looked at moisture in 3 different ways:

  1. Adding 4% moisture after the roast, and measuring the change in moisture.
  2. Slowly adding and removing water, knowing the groundtruth.
  3. Looking at the moisture metric when the chamber is over-filled or under-filled.

While my focus was moisture, I included the other metrics since they were free.

Adding 4% Moisture

In my typical process, I add 4% moisture after the roast to improve flavor and degas faster. I then collected data so I could see how both sensors responded.

For moisture, they had an offset, but they had a similar movement. They didn’t move 4% like they should, but moisture absorption into coffee beans can be weird, so it isn’t too surprising based on prior data.

DiFluid measured a higher reading.

Both sensors had a similar density trend with a slight increase.

Color also increased slightly for both sensors.

In terms of water activity, the values go from very low to almost back to the level of the green beans.

Moisture Testing vs Groundtruth

To further understand the moisture accuracy, I repeated a test on the Syncfo by adding moisture to green coffee beans using a sous-vide. This provided a high humidity climate. I put 150g of green coffee in a flat dish, and I put that into an instant pot set on a sous-vide setting. Underneath the dish is one inch of water, and it is set to 80F. This allows the beans to slowly gain 1% to 2% in moisture each day.

I assumed the first moisture measurement on the Syncfo of 10.6% was accurate, and I calculated the groundtruth with respect to that starting value. Any difference in measurements should fit a trend if that assumption is false.

Then I can measure the moisture change as a groundtruth by measuring the weight of the beans. After hitting close to 18%, I then let the beans dry out. This dropped moisture to close to the original around 10.6%. Then I put it back into the Instant Pot on a dehydrate setting to get close to 0%.

Both devices trended similarly until 15.5% for the DiFluid. I mentioned this to the people at DiFluid, and they said that their bean model assumed beans don’t have moisture content about 15%, and they will look into it.

The lower bounds for both sensors was about 9%. This seemed weird, and I suppose I would expect lower readings to be possible. So I took some samples using the roasted bean settings on the Syncfo. Whatever the calibration change, it was able to get closer to the groundtruth in those samples. I suspect this issue could be fixed with calibration modifications. I’m not sure if accuracy below 9% is necessary in every day roasting, and it is a problem shared by Syncfo.

I looked at comparing readings. They seem nearly identical from 8% until 14%, but then they start to deviate. If Syncfo is very accurate to groundtruth, then DiFluid can fit the data to a curve or improve their model.

I looked at water activity over these samples. The water activity got the lowest during dehydration, but not below 0.3.

Over/Under Filled Chamber

If a user added too many or too few of coffee beans, how does that affect moisture? I took 12 measurements to get a sense of how much error there could be. More data would be nice too, but this is what I have.

For Syncfo, Overfilled and the regular fill performed similar, but in some cases, underfilled was closer to groundtruth.

For DiFluid, overfilled was closer to groundtruth until 14% moisture, while underfilled and baseline were very close.

I looked at water activity for fun, and it seemed to match close for overfilled vs baseline, which indicates how to get better performance.

Side Note: Color

Color changed during the test, but it never went back to the baseline. I’m not sure what that says about color measurement.

Overall, the DiFluid Omix does well for the range it was designed for. Luckily, it has the potential for software updates to make improvements, and I think that has great value considering they have made software updates to the Omni after launch as well.

The moisture measure appears accurate in these tests compared to groundtruth, but there could be more tests. The level of accuracy for moisture measurement or water activity has not been clearly established for coffee roasting. Is 1% accuracy required or 0.1% accuracy? I’m curious to see how others use this data because I have not seen a lot of this type of data shared online.

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

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.