IMS Superfine vs VST: A Small Sample Espresso Filter Comparison

Using data to find a way to compare and contrast

Published in

Towards Data Science

6 min readApr 17, 2020

When I first got a VST filter basket, it changed my espresso dramatically. Once I got into making staccato espresso, the thought of filters intensified. I did an analysis on my VST filters and other filter baskets, and at some point, I ran across the IMS Superfine filter. Someone online was asking about the filters, and as I mentioned I had used them, I thought about the data.

Did I have enough data to better compare the IMS Superfine to the VST? Did I have enough data to compare my VST 7g filter to my VST 20g filter? So I dug into 6 months of data to see if I could control enough variables to find something interesting.

Keep in mind, this is a super small sampling in terms of the number of filters used. This analysis should be seen in the vein of “which of these specific filters is better”, not whether IMS or VST is better in general. For that test, a much larger study is required.

Top images: 7g VST, 20g VST, IMS Superfine. Bottom Left: VST 20g, IMS Superfine. Bottom Right: IMS Superfine

An Extremely Brief History

VST filters were made after some extensive studies which determined the hole tolerance required to improve espresso. They provide a spec-sheet with purchase, and I have independently verified their claims.

The IMS Superfine filter uses a mesh screen to filter out particles above 170um instead of the typical 250um and above. The filter is also considered a precision filter.

Previously, I have used image processing to analyze the hole size variations across filters. I present that analysis for these filters to compare. Each one has certain hot spots where the holes are larger and impact flow. I’ve noticed these overall flow pattern while pulling shots. It’s harder to see the specific hole values having a big impact, but it is something to be aware of. They all have some bias, but there is not a major bias that would indicate their performance would be terrible. For the Superfine filter, the analysis includes the metal screen on top making each hole have 4 to 10 smaller holes. However, I’m unsure if this is how the filter actually acts because the metal screen is only attached around the edge, not welded to the filter.

Metrics of Performance

I used two metrics for evaluating the differences between filters: Final Score and Coffee Extraction.

Final score is the average of a scorecard of 7 metrics (Sharp, Rich, Syrup, Sweet, Sour, Bitter, and Aftertaste). These scores were subjective, of course, but they were calibrated to my tastes and helped me improve my shots. There is some variation in the scores. My aim was to be consistent for each metric, but some times the granularity was difficult.

Total Dissolved Solids (TDS) is measured using a refractometer, and this number combined with the output weight of the shot and the input weight of the coffee is used to determine the percentage of coffee extracted into the cup.

Espresso Data Collected Slowly over Time

I took six months of shots across my VST single (7g), VST double (20g), and IMS Superfine (18g), and I started looking for patterns. Ultimately, I don’t have a fair comparison because I don’t have one with the same weight, but I have a lot of data, and maybe we can see a story.

First, all the data across Final Score and Extraction compared to output to input ratio. A pattern doesn’t quite emerge in this data aside from that it seems they have similar performance.

Then let’s compare with and without a Paper Filter in the Filter (PFF) basket. Still, a pattern isn’t quite clear except the extraction for Superfine seems less correlated to taste (Final Score).

To improve this comparison, let’s make some paired data. We will pair the best taste for each roast and compare the best of VST against IMS. Again, there is a lack of data, but there isn’t some clear convincing pattern even in that small data.

So let’s take all the shots for each roast, sort them by taste, and make pairs of VST and IMS shots to compare the best shot from VST to the best shot from IMS, the second best to the second best, etc. This gives more data, and from taste, they don’t seem too different. From extraction, VST is either the same or much better than IMS.

VST 7g vs VST 20g

Since we’re comparing data, let’s do the same thing to compare VST 7g vs VST 20g. Let’s start go straight to pairing the best shots of each roast together. Generally the issue is that a roast has 20 shots or so, and I wasn’t being deliberate on using all three filters for each roast. So it was a bit short on data.

Let’s expand the data by including data from the past year even if I didn’t have extraction measurements. Some of the shots were experimental, and some were used to dial in a grind. Overall, I haven’t seen much of a preference in performance. I prefer more frequent shots when I have time, so a smaller basket, but I have had less time these days of sheltering in place.

The two VST filters have differences, but ultimately over the past year, their best shots against each other are nearly equal.

Precision Filters at the End

There is some debate on which is better. Ultimately, I only have 2 VST filters and 1 IMS filter in this write-up. To have a better population sample of just the filter performance would require around 30 filters each with multiple shots across each filter. If I only had the time and money, I would try the experiment.

I am disappointed because I thought the IMS Superfine would be the next incremental improvement, but it is the same as another VST filter. However, it is comforting to know the IMS filter was worse than the VST filter.

If you like, follow me on Twitter and YouTube where I post videos of espresso shots on different machines and espresso related stuff. You can also find me on LinkedIn.