A couple of weeks ago, I spent a day grinding 115 doses of donation coffee from Level Ground, measuring the density of the tamped puck using a caliper, then measuring true density using a gas pycnometer. I’ve uploaded the raw results for this experiment to GitHub.
I did a similar experiment previously, but identified several sources of error in that experiment, and wanted to repeat it with a more rigorous measurement protocol. A few changes were made for this experiment:
- My doses were 15.5 g, 14.5 g, 13.5 g, and 12.5 g. This kept the measurements out of the range where I know my scale is not accurate.
- Beans used in the experiment were mixed by hand, so that we were using more or less the same mixture of beans for each measurement.
- A spring loaded tamper was used to provide a consistent tamping force (although this turns out not to be quite as true as one might hope).
- To keep the surface of the puck near the middle of the tamper’s range, a smaller portafilter basket was used and dose was periodically adjusted.
The following plot compares the results of the two experiments.

You may notice that the values for the previous experiment are a little different than what I previously reported. In order to get an accurate comparison of the two data sets, I recalculated density using the measured volume of the portafilter basket used in each experiment. I’ll go into more detail about this procedure in a future post.
There are a couple of notable features in this plot.
Puck composition
First, there is a kind of “pinch” at a grind setting of 4.50, where all the measurements are in close agreement. On either side of that grind setting, we see more variation in the measurement, as shown in the following plot:

This suggests to me that the composition of the puck changes with grind setting, and that some compositions are more susceptible to changes in tamping force than others. For example, the puck may become more stable when the fines exactly fill the gaps between the nominal particles, but when there is an excess of either fines or nominal particles, the puck becomes less stable.

From a soil engineering perspective, what I’m suggesting is that the compression index of ground coffee depends on the grind setting. I am planning to test this in a future experiment using a homemade oedometer.
Tamping force
The second notable feature is the small drop at a grind setting of 6.00. I believe this is an artifact resulting from a change in dose from one grind setting to the next. To see why I performed the experiment in this way, we’ll need to look a little more closely at spring loaded tampers.
In a previous experiment, we showed that tamping force has a significant effect on the density of the tamped puck. In the current experiment, we used a spring loaded tamper to control that effect. However, the spring loaded tamper also has some limitations.

The Normcore Spring Loaded Tamper V4 contains two springs. The outer spring (not shown) holds the tamper level against the rim of the basket, but doesn’t affect tamping force at all. Tamping force is determined by the inner spring. This spring is almost completely extended at rest, and compresses when the tamper reaches the surface of the puck.

The tamper has a limited range. Above the maximum depth, about 18.25 mm, the tamper doesn’t reach the surface of the puck at all, so no force is applied. Below the minimum depth, about 7.25 mm, the self-leveling mechanism doesn’t make contact with the rim of the portafilter basket, so there is no way for the tamper to limit the force applied to the puck.
Within these limits, tamping force varies as shown in the following plot:

At the maximum depth, the spring doesn’t compress at all, so we apply no force to the puck. As the spring compresses, the force increases linearly, as described by Hooke’s Law:
The tamping force continues to increase until we reach the minimum depth. At less than the minimum depth, the spring loaded tamper does not limit the force applied to the puck.
Ideally, when we’re making depth measurements, we want to apply the same force to the puck every time. However, with a spring tamper, this only happens if depth is constant. When the depth changes, so does the tamping force, as we saw in the plot above, and when the tamping force changes, so does the density.
To limit this effect, I tried to adjust the dose in order to keep the depth measurements near the middle of the tamper’s range.

I was able to keep the depth between about 11.00 and 13.25 mm, but in the top plot we can see that there are “steps” in puck density which correspond with changes in dose.
To confirm this, I repeated measurements from a grind setting of 5.5 to 6.25, using a dose of 13.5 g, with the following result:

Notice that the repeated measurements, shown in orange, do not show the same step that the original measurements do.
Gas pycnometry
I also measured true density using gas pycnometry. The results of those measurements are shown below.

Note that the true density and puck density are not plotted on the same scale here. The scale for true density (orange) is shown on the left, and for puck density (blue) on the right.
There is more noise in the measurements than I would like. The syringe I’m using is labeled as “single use”, but in this experiment alone I made 52 measurements with it. The plunger started to stick, which made it difficult to set an accurate volume.
I later learned that the plunger on these syringes is made with a relatively cheap rubber, with a coating of silicone applied to prevent sticking. This coating wears off over time, which is part of the reason they aren’t meant to be reused. Samo Smrke suggested that a food grade dry silicone spray could be used to replace the original coating, and I was able to find this in a local sporting goods store:

After applying a light coating of the dry silicone spray, the plunger moves freely again. In the future, this is something I will do as a part of regular maintenance.
Conclusion
I originally set out to measure puck density in order to validate a proposed model of grinding. At this point, I think it’s likely that the pronounced “knee” I saw in those measurements was due to variations in tamping force, rather than the transition from a puck dominated by nominal particles to one dominated by fines.
However, the “pinch” noted at the start of this post suggests that there is some truth to the underlying model, although I would stop short of saying the results of this experiment validate the model. Because of the sensitivity of puck density to tamping force, and because it takes many puck density measurements to get a good fit, it seems likely that gas pycnometry will be a better way to explore the model further.
Research by others, including Socratic Coffee and Tomonori & Smrke, suggests that variations in tamping force have little to no effect on extraction—at least in terms of total dissolved solids and extraction kinetics. This isn’t to say there is no effect at all—it’s possible there are important effects which lie outside those measurements.
I suspect what’s happening here is that the puck wetting process at least partly undoes what’s done by tamping. In soil engineering, consolidation is often described as the process of squeezing fluids out of soil, so it’s possible that replacing those fluids can reverse the effects of consolidation.
There’s a lot more research to be done here, but in the next few posts I’m going to turn my attention to extraction. First we’re going to measure extraction kinetics by splitting shots, then we’re going to compare this with an analytic solution to the differential equations presented by Moroney et al. in their 2019 paper.