As I’ve been improving my espresso shot, I came across the Kruve coffee sifter. The central aim is to keep only a band of particle sizes for which ever method one would choose to make coffee. A good grinder should grind within a tight tolerance, and Kruve aims to help people calibrate their grinder or make a good grind out of a terrible grinder. These are all admirable aims, so I decided, I must try one.
After publishing this article, I bought the 200um and 500um filters, so I have added them below.
Espresso is the ultimate embodiment of coffee.
First, I had to buy one. I didn’t want to commit to getting all the screens in case I was disappointed. Even getting the two screen option was $50, and that was unappealing. Luckily, craigslist exists, and I negotiated the price to $33 for 2 screens (400 um and 800 um). Kruve recommends the 250 um and 500 um screen for espresso. I figured, if I liked it, I could buy more screens.
I tried it out, but because I don’t have the right screens, I can’t take full advantage of what they intended. So I’m still messing around to see what I can get out of it as is and determine if it is worth more money. Let’s see how well the screens do their stated job.
As a data scientist, I decided to take a deeper look at the sieves. I decided to use the same hole diameter analysis I did for espresso filters. Keep in mind, Kruve claims their spec to be within +/-12 um. VST claimed their espresso filters to be within +/- 20um, and my measurements showed they were within spec.
I initially took the images of the filters after using them, but they were too clogged with coffee grounds. Then I cleaned them using water. I also have three takes of the 200um filter because I was concerned I wasn’t getting the accuracy desired.
If we center the distributions by the median, we can see the tolerance on the holes.
Even is you take the tolerance as between the 25% and the 75% percentiles, the 800 um would be within tolerance, but the 400 um is not within their said tolerance. Tolerance is usually the minimum and maximum variance, so both filters are quite a bit off.
The boxplot is interesting, but what about the actual distribution?
The effectiveness of the sift is completely dependent on the tolerance of the screens, but what if we could visualize these hole sizes so we could have a better understand if the differences in hole sizes are systematic or random? Random would be better than having a local region of holes that are too big.
You’re in luck! There is a way! Here are the two filters cleaned, and here is how to understand these beautiful pictures:
Left: Hole diameter (mm)
Middle: Hole diameter (mm) enlarged for easier viewing
Right: Distance (mm) to the closest hole for each point on the filter.
These images show quite nicely that the red areas have holes larger than tolerance, which are the greater concern. For holes smaller than the tolerance, it might limit what goes through, but the tolerance for a sifter should be viewed as blocking larger particles from getting through.
Unfortunately, the distribution of the larger holes is not random, being localized in the middle. Even this placement is the worse area assuming coffee grounds would be cover the center of the sieve more than any other part. This issue is particularly true for the 200um filter.
Yes, this sifter may not be to the exact tolerances they claim, and I’d be interested in doing this same analysis on the full set of the screens. But I still love the idea. Nobody else is making a sifter as affordable as Kruve and definitely not for coffee. I did a lot of research to make sure there wasn’t a cheaper option.
Further readings of mine: