Ah, I didn't realize you had used the starter fluid to spread it on the scanner bed. I'll see if a 1:1 macro lens can work as well.

another_jim wrote:I'll see if a 1:1 macro lens can work as well.

That crop I posted earlier on is from a photo taken with a 1:1 macro.

I still cannot see how you plan to create a small sample which is a true representation of the whole grind portion (say 18g).

Robert do you know how much coffee in grams you normally spread on the scanner? is that a random batch of a bigger amount of ground coffee? are we talking espresso settings or brewed coffee size?

Regards,
dsc.

I have been using espresso grinds. I have used samples that were between 50 and 100 milligrams. The best technique that I have developed to date is as follows:

Put the 50 to 100 mg of ground coffee in a small vial. (The ones I have been using are 6cm tall and 1.5cm in diameter) It needs to have a lid. From a tin of starting fluid, add about 3 mL of it. Shake. Let stand at least 15 minutes with cover on. With an eye dropper carefully remove the ether from above the settled grounds. Leave a bit of solvent so as to not remove the fines. Add about 1 mL more ether and dump contents onto a sheet of white writing paper. The paper will absorb most of the ether and take the oils in it mostly into the paper. Let evaporate. This will not take long with the ether. Remember it is used as starting fluid, so is highly flammable! Take an artists brush and sweep the powder into a clean dry container like a ramekin or cup.

Be sure the top of the scanner is clean. Take a kitchen strainer or other screen and carefully distribute the grounds as they pass through the screen onto the scanner surface. Put a clean sheet of white paper over the particles and scan at highest resolution. DO NOT SAVE AS JPEG. PNM, Targa, pnm, any lossless format is fine. You can use this image directly into imageJ, or you can pre edit it with an image editor like gimp to convert it into a strictly black and white image. If you fail to do this, imageJ will tell you to use its built in threshold converter. (However, I have always cropped the image before giving it to imageJ, looking for a place on the image where there is the most even distribution of particles.) I try to get a 2000 or 2500 square image. I never have scaled the image. That really makes no sense.

Imagej must be told how many pixels there are per unit. You can use absolutely any measurement you like, you could use microns, millimetres, even cubits! The first time you use imagej you must also tell it what to calculate, there is a menu to check your choices. <analylise><set measurements>

It turns out its data files by default as text files. I wanted that, so I left that alone.

It would work just as well with coarser grinds, but you would need larger samples. I thought it best to strive to have from 500 to 1500 particles in the 2500 square image. You would want to be toward the lower range for coarse grinds. I suspect you will still have to clean the particles with solvent to prevent their sticking together.

Although most people like plotting with spreadsheets, GNUplot is simpler. There are many other ways to plot the data. I plotted generally feret diameter vs circularity. imageJ produces lots of information, and other plots would be just as easy.

This would certainly also work with Turkish grinds, but would require a microscope. If you use a microscope you need suspend the ground coffee in a solvent to dissolve oils to prevent clumping. I have examined specimens suspended in propanol, and that worked well, I did not attach the camera and make images, but that would work just as well as a scanner.

I think this neat!

to prevent the loss of small particles, I think that you could try to adopt a soxhleth-type configuration. I mean that you could try to put the coffee grounds in a porous paper thimble. Then you could add solvent from the top letting it go out from the bottom, through the paper. Or you could submerge the thimble in a small flask full of solvent removing it after some minutes.

Thanks for the description Robert. How do you pick a sample? do you simply grind a bigger amount of coffee into a container, mix and then randomly pick your 50 - 100mg?

Regards,
dsc.

I have just randomly scooped some from the grinds when I was making espresso. If you dump the ether solution as suggested previously onto a sheet of white paper, it first looks like the grounds stick to the paper, but when it becomes dry they will sweep of easily with a brush. I swept them into a ramekin, and then poured them through the kitchen strainer onto the bed of the scanner. That worked better than any other thing I did.

dsc wrote:I still cannot see how you plan to create a small sample which is a true representation of the whole grind portion (say 18g).

Wrong problem -- it assumes the grinder output is nonuniform -- in which case the grinder is no good. To prove this for yourself, just dose by weight. Do you get incosnistent behaviour when adjusting the final 1/2 gram. If not, the final 1/2 gram is a good sample.

The first order of business is to find out how to describe a grind distribution. Here's one simple test at which the laser and sieve sizers failed: distinguish the grind distributions of the same espresso grinder set for a 12 gram dose and a 18 grams dose The taste and flow differences from these two grind setting is obvious to everyone; so the failure of single variable distribtuons of size to account for the difference is damning.

Robert using ImageJ to get both size and sphericity of the particles is really exiting; these bivariate disributions may pass this apparently simple grind adjustment test.

An interesting thread, I hope it bears fruit. I've always watched from the sidelines, but even spectators want to know why one team always wins.

distinguish the grind distributions of the same espresso grinder set for a 12 gram dose and a 18 grams dose

To complicate the matter, the taste difference between two such samples would be due to absolute particle size as well as the distribution of sizes wouldn't it? I.e., in the 12g dose each particle would be more completely extracted (having a higher surface area:volume ratio).

Wouldn't it be better to distinguish differences between the same dose on two different grinders with different flavor profiles?

another_jim wrote:Wrong problem -- it assumes the grinder output is nonuniform -- in which case the grinder is no good. To prove this for yourself, just dose by weight. Do you get incosnistent behaviour when adjusting the final 1/2 gram. If not, the final 1/2 gram is a good sample.

Why wouldn't you assume the grinder to be an unknown? I thought the whole idea of this testing is to decide whether the grinder is good or not. This sample selection is something to be aware of especially when single dosing, as we have no idea what that does to a grind and it's consistency, even if you pull shots properly later on. I'd say you need to grind either in batches say 2-3g at a time and sample each portion to check for distribution, or grind the whole lot (typical espresso dose), mix and then sample a few times to compare them against each other. This would answer the question why on some grinders, dosing straight into a basket doesn't work and mixing in a container and transferring it over to the basket works.

Regards,
dsc.