Someone hands you a lab report or a supplier spec sheet with three numbers on it: D10, D50, and D90. Your job is to turn that into a screen for a round vibratory separator. Those numbers won't name your mesh. What they will do is show how much of your material sits close to any opening you're considering, and that has a lot to do with whether a screen runs well or chokes.
What D10, D50, and D90 actually mean
Each number is a point on the cumulative undersize curve of a particle size distribution, which tracks how much of the sample is smaller than each size. Horiba's specification note lays it out plainly:
- D10: 10% of the distribution is smaller than this size, and 90% is larger.
- D50: the median. Half is smaller and half is larger. It is not the average, and mean values are reported separately.
- D90: 90% is smaller and 10% is larger.
10% of the distribution sits above D90, 50% sits above D50, and 80% sits between D10 and D90. This article uses one set of example numbers throughout: D10 of 20 microns, D50 of 80 microns, and D90 of 200 microns. Those are the round figures from Horiba's note, used here for teaching, not a real product.
D90 is not your largest particle, since 10% is bigger by definition. And D100 won't rescue you, because Horiba calls it the least robust value an instrument calculates. Something as small as an air bubble can throw it off.
You may also see span, which is (D90 minus D10) divided by D50. For the example, that's (200 minus 20) divided by 80, or 2.25. We didn't find a published general cutoff for "narrow" versus "broad," so treat span as a way to compare lots, not a grade.
Before you trust the numbers, check the method and the basis
The same powder can read very differently by method. Check these two things first.
| Laser diffraction | Sieve analysis | |
|---|---|---|
| Reports | Spheres whose predicted scattering pattern matches the measured one | Weight retained on or passing a set of openings |
| Usual basis | Volume, so a bare D50 usually means Dv50 | Mass |
| Particle dimension that counts | An equivalent sphere | Mainly width, not length |
Shape is where this bites. Malvern's guide gives the example of a rod 360 microns long and 120 microns wide, which has the same volume as a 198 micron sphere. A screen opening would mostly respond to that 120 micron width. ISO 13320 says outright that laser results on non-spherical particles differ from sieving results. We also didn't find a reliable general factor for converting one to the other, so don't assume how far apart they'll be or in which direction.
Basis matters too. In Horiba's guidebook, one measured sample had a median of 11.58 microns by volume and 0.30 microns when converted to number. That's more than 38 times. On a Dv report, "90% below" means 90% of the volume, not 90% of the particles. Some reports are built from sieve data instead of laser data, so check which kind you have.
Then ask how the lab dispersed the sample. A laser run may be set up to break agglomerates down to primary particles. If those agglomerates survive on your separator deck, they can behave like larger particles than the report shows.
Mesh number doesn't fix the opening
Before matching a report to a screen, convert the screen to microns. The number on the screen is a count of openings per inch, so wire diameter changes the actual opening. Here are published figures from Midwestern's wire mesh comparison chart, set next to the ASTM E11 test sieve with the same number:
| Mesh | Tensile bolting | Mill grade | Market grade | US test sieve |
|---|---|---|---|---|
| 20 | 1,041 µm, 67.2% open | 914 µm, 51.8% open | 864 µm, 46.2% open | 850 µm |
| 60 | 310 µm, 53.3% open | 259 µm, 37.5% open | 234 µm, 30.5% open | 250 µm |
| 100 | not listed | not listed | 140 µm, 30.3% open | 150 µm |
| 200 | 86 µm, 46.2% open | not listed | 74 µm, 33.6% open | 75 µm |
"60 mesh" runs from 234 to 310 microns on that chart, depending on the cloth grade. That's a spread of about a third, and it's why the opening in microns is the number to compare to your report. For more on the conversion, see our guide to mesh count vs. micron rating and the effect of wire diameter on open area.
From report to shortlist: a worked example
1. Name the job
Kason defines scalping, or sifting, as removing a small percentage of oversize from a large percentage of on-size, and de-dusting as removing a small percentage of undersize from on-size. Grading or classifying splits the material into size fractions. You may hear oversize and undersize called overs and throughs on the shop floor.
2. Mark the near-size band around each candidate
Near-size particles are the ones that queue at the openings and lodge in them. Kason defines them as 1/2 to 1 1/2 times the opening. Definitions vary, and our own glossary uses a tighter working range of 75% to 125%. We'll use Kason's wider band here. It lands on the example powder like this:
| Candidate | Opening | Near-size band (1/2 to 1 1/2 times) | Where the report falls |
|---|---|---|---|
| 20 mesh tensile bolting | 1,041 µm | 521 to 1,562 µm | D90 is well below the band. Light near-size load, but the report can't tell you what's in the top 10%. |
| 60 mesh mill grade | 259 µm | 130 to 389 µm | D90 sits in the band and D50 sits below it. Expect a real near-size load. |
| 100 mesh market grade | 140 µm | 70 to 210 µm | D50 and D90 are both in the band, so at least 40% of the distribution is near-size. D90 is above the opening too, so this is more of a grading cut than a scalp. |
For a scalp on this powder, 20 mesh and 60 mesh stay on the list. 100 mesh comes off, because more than 10% of the distribution reads above its opening. Between the two that stay, 60 mesh catches smaller oversize but carries more near-size, so the size of your oversize decides it. If you want a grading cut near 140 microns, 100 mesh is the candidate, with a heavy near-size load to plan for. Remember that a laser number is an estimate of what the screen sees, not a measurement of it, so test before you commit.
3. Weigh near-size against the job
Particles smaller than an opening don't all pass on the first try. In a classic flat-screen probability model described in KONA Powder and Particle, at 25% open area a sphere 90% the size of the opening has only a 0.25% chance of passing per try, and about a 22% chance after 100 tries. The authors treat 100 tries as very good industrial screening. It's an idealized model built on spheres, and the same paper says first-principles prediction isn't really possible in practice, but the direction is the lesson: anything close to the opening is slow.
Sweco describes the same effect on round separators using granulated sugar. If 95% of it is slightly smaller than a 20 mesh screen, particles jam up at each opening like traffic, capacity falls, and you need far more screen area. A tighter mesh isn't automatically a sharper separation. It can just be a slower one, and near-size is also one of the main routes to blinding and pegging.
You may run across a rule elsewhere that says to size the opening some fixed percentage over your target. We looked for one from Sweco, Kason, Midwestern, and Russell Finex and didn't find one. Sweco's answer on capacity is extensive testing of your material through the equipment maker.
4. Adjust for what the report doesn't show
Percentiles don't capture stickiness, moisture, or static. Sweco notes that powdered sugar blinds mesh and screens slower than granulated, and that humidity and static shift results as well. The pharmacopeial test-sieving text says that below about 75 microns, particles have too little weight to overcome cohesion and adhesion, so particles you'd expect to pass get retained. That passage is about lab sieves, not production machines, but it's a fair warning. If your D50 is down in that range, read our guide on screening fine powders and consider what deblinding system belongs under the screen. Starting points by material: flour and grain, pharmaceutical powders, powder coatings, and battery materials.
Check the choice on your own material
- Sample in motion. Powders segregate in handling, so take the sample across a moving stream rather than scooping from one spot. Dividing method matters too. In NIST's sizing guide, spin riffling showed a relative standard deviation of 0.125%, compared with 5.14% for scooping and 6.81% for cone and quartering.
- Don't overload the sieve. An overloaded sieve blocks and reads too coarse. The harmonized pharmacopeial method (the 2009 ICH text, since revised by USP in 2023) uses 25 to 100 grams on a 200 mm sieve, and suggests comparing 25, 50, and 100 grams for the same time. If 100 grams shows less through the finest sieve, it's too much.
- Know your endpoint. In that method, sieving is complete when the weight on any sieve changes by no more than 5% or 0.1 gram of its own previous weight. That is not 5% of the whole sample. Other methods use other rules, so note which one you ran.
- Look at both streams. Lots of good product in the oversize can point to a screen that's too tight, blinded, or overloaded. Oversize showing up in the undersize calls for a look at the screen and the seals before you change the mesh.
One caution for regulated plants: that same pharmacopeial chapter says it isn't intended for finding the portion passing or retained on just one or two sieves. A quick overs-and-throughs check is a useful practical comparison, not a compendial result. Your QA group should set the method where it counts. Our mesh selection guide covers what to watch during a trial run.
FAQ
What does D50 mean on a particle size report?
D50 is the median particle size. Half of the sample is smaller than that size and half is larger. On a laser diffraction report, it normally means half by volume, so it is not the same thing as an average particle size.
Is D90 the largest particle in my material?
No. By definition, 10% of the measured distribution is larger than D90. D100 is not a dependable top size either, because it is the least robust number a laser instrument calculates.
Can I convert a laser D90 directly to a mesh size?
We haven't found a reliable general conversion. Laser diffraction reports an equivalent sphere, while a sieve responds mainly to particle width. The two can disagree, especially with needles and flakes, so use a laser D90 to build a shortlist and confirm the choice with a test on your material.
How close can the screen opening be to my target particle size?
No published percentage rule from the major round separator makers turned up in our research. The closer the material sits to the opening, the fewer particles pass per contact and the more capacity you lose, so a test on your own material is the reliable way to settle it.
Send us your report and we'll help narrow it down
If you have a D-value report and a screen decision to make, send it over with your machine make and diameter, the job you're trying to do, and what's going wrong today if anything. We'll help you compare openings and cloth grades and find the right fit in our replacement screens for round vibratory separators. No report? Tell us your machine and what the operator asked for, and we'll start there. Fit questions? See our Sweco, Kason, and Midwestern fit guides. Contact our team or call (866) 265-1575.
