TGP II: Particle Distribution Analysis of Grinder Adjustments - Interim Results - Page 2

Grinders are one of the keys to exceptional espresso. Discuss them here.
User avatar
another_jim
Team HB

#11: Post by another_jim »

Tristan, the graphs are not of a total number or weight of anything, the numbers show the proportion of grinds and percentage weight. That is, 1 gram of the 18 gram sample should have more weight of coarse particles and large coarse particles than 1 gram of the 15 gram sample. The Y axis on the raw distributions is a local and arbitrary measure based on the number of bins used by the sizer. You can't even compare the number to other distributions from other sizers. You can compare all the posted graphs

I think we need to get over our tech idolatry here, and look at the facts:
  • The 15/18 gram adjustment is much larger than is usual in the day to day, or coffee to coffee, tuning of an espresso grinder. Changing baskets will not change this (the adjustment range is about the same, at least for my grinder, when using kinder and gentler baskets)
  • The laser sizer, when testing the same sample twice, does not show exactly same distribution. It has, like all measuring instruments, some built in variability. This variability is miniscule when measuring non-porous and nearly spherical particles. It is much worse as the particles become more porous and more irregular.
  • In 15 out of 20 cases, the two very different espresso samples did not show up as more different than the same espresso sample measured and remeasured,
  • Therefore, the laser sizer, when used for ground coffee, has a measurement error larger than the range used for almost all espresso adjustments.
  • This fact has never been mentioned in the literature. It has also never been denied, except by implication. Either coffee labs never tried this or failed to publish the negative result
Kendall has access to several instruments that can analyze the exact shapes of large samples of particles (like the coffee samples). These are considerably more difficult for him to get time on then the laser sizer, since they are more expensive and more powerful. This means he will only be able to run a few pairs. Given this, we will take two pairs from the same grinder where one shows up with a large difference, and other other with none. The K10 Lundgaard or the CJR Wookie samples have this property, and one or both of these will be used.

If the exact analysis shows that there is an equally large difference between the 15 and 18 gram hopper/single pairs, we will have confirmed the problem is with the laser sizer. If the SEM confirms the laser sizer was accurate, then we were massively incompetent in preparing the samples.

I'll gladly take bets the culprit is the sizer, not us; and am cheerfully willing to fleece anyone who backs the sizer. :D
Jim Schulman

User avatar
shadowfax

#12: Post by shadowfax »

Jim, thanks for your thoughts. For my part, I don't have to suspect myself. I logged and clearly labeled the grind setting on my samples, so for 2 samples at different grind settings to look the same there definitely has to be a problem with the instrument resolution/accuracy for coffee grinds. I had suspected some possible wonkiness with the Vario, as I had some trouble with one set and the prolonged use at short shot intervals made the grinds coming out quite hot indeed, much hotter than I'm used to, borderline uncomfortable on my hand—I have no idea if things like that could skew the way the particle distribution or the burr spacing (having the chamber be so darn hot). Still, the K10s (mine at least) didn't have that issue by any stretch of the imagination and shows a similar 'anomaly' to the Vario samples so I'll readily ignore it. I'm looking forward to seeing what we turn up with the SEM.
Nicholas Lundgaard

User avatar
TrlstanC

#13: Post by TrlstanC »

another_jim wrote:Tristan, the graphs are not of a total number or weight of anything, the numbers show the proportion of grinds and percentage weight. That is, 1 gram of the 18 gram sample should have more weight of coarse particles and large coarse particles than 1 gram of the 15 gram sample. The Y axis on the raw distributions is a local and arbitrary measure based on the number of bins used by the sizer.
Oh, I had assumed that the Y axis was proportion of total volume, so that if we integrated under the curve we'd end up with 1. I was trying to extrapolate from that data and the known dose sizes to the total volume of each kind of particle created, but that might not be possible.

Also, I was thinking in number of particles, but it's actually volume, which is interesting because the volume of the fines and coarse particles is fairly similar (at least within an order of magnitude) which is surprising because a coarse particle would have 1,000 times the volume of a fine (again, assuming spheres). So, that means that for every coarse particle there are 100s of fine particles.

User avatar
another_jim
Team HB

#14: Post by another_jim »

The biggest number of particles, millions times more than either fines or coarse, are in the 1/2 to 2 micron range. These come to about 2% to 4% of the total mass.

I don't show them on the raw graphs, which I start at 4 microns, for two reasons. First, they don't block liquid flow. They are in the colloidal range (i.e. they suspend in water and act pretty much like solubles). If you measure the solubles yield in coffee by weighing the loss of grinds or by evaporating, you get around 23 to 25%, if you measure using pure dissolved solids, you get around 19% to 21%. The difference is in these colloidal suspension sized particles. Second, I did a due diligence test to see if these "ultra-fines" explained the anomalies, but they did not. Their amount is nearly constant, and what variation there is strongly correlates to the average size of coarse particles and their proportions. This means they add no new information.

Kendall analogy may be the most accurate --- these distributions are like out of focus photographs, and the details just don't show.
Jim Schulman

kmills

#15: Post by kmills »

I would be very curious to know just how much solid material is extracted. This can be done through nano filtration or evaporation. I'l see what I can do about this. Im not convinced they play no roll in extraction. At work we separate 50nm particulate from a gas stream with microporous stainless foam. Attractive forces cause the pores to close quickly and no particulate passes through. I was surprised at first but its very effective.

Ken Fox

#16: Post by Ken Fox »

How about a less-cerebral explanation.

One starts with a simplistic and probably untrue theory attempting to explain why some grinders work "better" for espresso than others. This simplistic theory is then "tested" with an obtuse "methodology" that involves numerous variables, many of which can easily be misinterpreted by those participating. Test samples obtained with this obtuse methodology are submitted and there are no results of any consequence obtained. One is left with no results of consequence, and no understanding of why this is so.

Given the circumstances, the outcome could have been easily predicted before the "study" was done.

ken
What, me worry?

Alfred E. Neuman, 1955

User avatar
shadowfax

#17: Post by shadowfax »

Seems a mite early for the "told you so" gloating; maybe that's just me.
Nicholas Lundgaard

wookie

#18: Post by wookie »

another_jim wrote:This leaves just one hypothesis. The laser sizer isn't all it is cracked up to be. In fact, it usually misses the effect of grind adjustments in the espresso range. How can this be? Here's the theory. Several lasers are used so that each particle casts several shadows. From this, a particle size is computed that is the diameter of a sphere that would cast shadows of equal area. This conversion may not be right when investigating particle packing, since if the particles are not at all spherical, but say flakes, they would pack based on their thinnest dimension, not their average one.
I wonder if laser diffraction simply has inadequate resolution for measuring irregularly shaped particles (ground coffee) for the reasons stated above.

A few cursory Google searches about laser diffraction suggest that it would have more than enough resolution when working with spherical particles. But an equally cursory search of the coffee literature seems to find quite a few researchers relying upon automated image analysis particle sizers, which apparently have considerably better resolution for irregularly shaped particles. Kendall can probably correct me if I am wrong but perhaps laser diffraction sizer isn't accurately characterizing the grinder samples because the coffee particles are irregular and non-spherical.

It may be that automated image analysis is simply a superior tool for coffee grounds. Or that laser diffraction needs to be used in conjunction with microscopy & statistical modeling to properly characterize non-spherical particles.
.

User avatar
TrlstanC

#19: Post by TrlstanC »

Ken Fox wrote:Given the circumstances, the outcome could have been easily predicted before the "study" was done.
Ken, if you knew that the resolution of laser diffraction was going to be a problem with this study, or if you could have "easily predicted" that outcome, that's something you should have brought up ahead of time. That would have been a great example of a post you could've made that would've been a positive contribution to the discussion. However, if like the rest of us you didn't know about these limitations, I think we can all agree that we've learned something interesting and useful.

Ken Fox

#20: Post by Ken Fox » replying to TrlstanC »

Tristan,

Do you, or does anyone else here know whether the above "explanation" for the lack of results is due to this limitation of the equipment used and how it would relate to any "results" from this study? Useful results come out of studies using equipment with limitations all the time, in fact it is the norm since few tools are perfect.

I think it is just as likely, probably much more likely, that the theory about simple particle size distribution causing the observed differences among certain grinders, is flawed. That would be my interpretation at this point. This theory (about particle size distribution vs. "observed results") is truly very simple, and if true, it should not be all that difficult to do a study whose results could be interpreted to support it. That is what should have been done in this case and was not.

The simple "fact" is that if one does a simple and straightforward study with few variables, and one gets a positive result, or "disprove" a negative one, then one can posit whatever one wants to explain the results. Others will challenge those interpretations, and that will form the basis for the next experiment. If on the other hand one has no results, as in this situation, then it is hard to move forward with the next level. Rather, one is put in the unpleasant situation of choosing whether to cling to the original theory, using such explanations as are given above in this thread, or deciding that the theory is wrong. Neither choice is appealing and either could easily be wrong.

If, on the other hand, the idea is, that we "know" this theory to be true, and we are going to go out and find a way to prove it, that is simply "junk science," or whatever similar label one would care to apply to it.

ken
What, me worry?

Alfred E. Neuman, 1955