The question that starts most O-ring orders is whether a particular inside diameter exists. On an AS568 ring the answer is not "roughly" or "the next size up": the diameters are points on a grid, and the grid is a ladder whose steps double as it climbs.
The dash-number guide says the inside diameters are not evenly spaced. That is the negative version of the fact. The positive one is this: between one size and the next the gap is always a power of two of 1/64 in — 1/64 in being the smallest step in the whole standard — and every time the ladder goes up a level, that gap doubles and the number of sizes you can buy in an inch halves.
The step is counted in sixty-fourths
Two things have to be true before a claim like that can be checked. The step has to be measurable, and the measurement has to survive the way the data is printed.
The diameters here are given to three decimal places in inches. That is enough precision to see a step but not enough to name it cleanly: a 1/16 in step printed as 0.062 in one place and 0.063 in the next is one step, not two. Read literally, the table looks like it alternates between two numbers. Rounded to the nearest 1/64 in, it does not.
So every diameter is converted to whole cells of 1/64 in and each gap is rounded to the nearest one. The rounding is safe because the printing error is at most 0.001 in — one fifteenth of a cell — while the smallest step in the standard is two cells. Then the only question left is which integer each gap lands on.
Five cords, one ladder
Each of the 5 series that is cut by cord diameter has one cord that carries the bulk of its sizes, and those are the ladders. The table gives the first and last step of each, the number of sizes in that cord, and what the same range would need if the first step had been used throughout:
| Series | Cord (mm) | Inside diameter (in) | First step | Last step | Sizes | At the first step throughout | Kept |
|---|---|---|---|---|---|---|---|
| 000 Series | 1.78 | 0.070–5.239 | 1/32 in | 1/4 in | 47 | 166 | 28.3% |
| 100 Series | 2.62 | 0.049–9.737 | 1/32 in | 1/4 in | 77 | 311 | 24.8% |
| 200 Series | 3.53 | 0.171–17.955 | 1/16 in | 1 in | 84 | 286 | 29.4% |
| 300 Series | 5.33 | 0.412–25.940 | 1/16 in | 1 in | 87 | 409 | 21.3% |
| 400 Series | 6.99 | 4.475–25.940 | 1/8 in | 1 in | 51 | 173 | 29.5% |
Read the middle columns together and there is exactly one rule. The 000 Series and the 100 Series start at 1/32 in and climb to 1/4 in; the 200 Series and the 300 Series start one level higher at 1/16 in and climb all the way to 1 in; the 400 Series starts at 1/8 in. In every one of them the next level is exactly twice the last — never one and a half times, never three times. That is not a rounding of some smooth curve. A smooth curve would give a different multiplier at every level.
The last two columns are the price of the ladder. If each cord kept its own first step all the way along, the 5 ladders would need 1345 sizes; they hold 346. Between 21.3% and 29.5% of that grid survives — the ladder throws away roughly three sizes in four.
Each doubling halves the density
A step is easier to feel as a density than as a fraction. One inch divided by the step is how many different rings you can buy in that inch, and because the step doubles, so does the density halve — exactly, at every level:
| Series | Inside diameter (in) | Step | Sizes in the band | Sizes per inch |
|---|---|---|---|---|
| 000 | 0.070–0.239 | 1/32 in † | 7 | 32 |
| 000 | 0.239–1.364 | 1/16 in | 19 | 16 |
| 000 | 1.364–2.989 | 1/8 in | 14 | 8 |
| 000 | 2.989–5.239 | 1/4 in | 10 | 4 |
| 100 | 0.049–0.237 | 1/32 in | 7 | 32 |
| 100 | 0.237–2.862 | 1/16 in | 43 | 16 |
| 100 | 2.862–2.987 | 1/8 in | 2 | 8 |
| 100 | 2.987–9.737 | 1/4 in | 28 | 4 |
| 200 | 0.171–1.484 | 1/16 in | 22 | 16 |
| 200 | 1.484–5.984 | 1/8 in | 37 | 8 |
| 200 | 5.984–9.984 | 1/4 in | 17 | 4 |
| 200 | 9.984–11.984 | 1/2 in | 5 | 2 |
| 200 | 11.984–17.955 | 1 in † | 7 | 1 |
| 300 | 0.412–1.350 | 1/16 in | 16 | 16 |
| 300 | 1.350–5.975 | 1/8 in | 38 | 8 |
| 300 | 5.975–9.975 | 1/4 in | 17 | 4 |
| 300 | 9.975–11.975 | 1/2 in | 5 | 2 |
| 300 | 11.975–25.940 | 1 in † | 15 | 1 |
| 400 | 4.475–5.975 | 1/8 in | 13 | 8 |
| 400 | 5.975–7.975 | 1/4 in | 9 | 4 |
| 400 | 7.975–19.955 | 1/2 in † | 25 | 2 |
| 400 | 19.955–25.940 | 1 in † | 7 | 1 |
A † marks a band that contains a step which is not on the ladder — 6 steps in all, and they are the subject of the section after next.
The 200 Series shows the whole descent in one column: 16, 8, 4, 2, 1. Not approximately halving — halving. At the bottom of that series you can choose a ring every 1/16 in; at the top, once per 1 in.
That is the number to carry to a drawing. If your groove needs an inside diameter within a millimetre of a target, the lower third of a series can give it to you and the upper third cannot, and no amount of searching will produce a size that the ladder does not have.
At one diameter, the thick cord is the fine one
The density column invites a wrong inference: that a thin cord is fine and a thick cord is coarse. At any single inside diameter the opposite is often true.
Every cord series covers a range of diameters, and the ranges overlap. At 5 in every one of them is in range, and this is the step each one offers there:
| Inside diameter (in) | Cord 1.78 mm | Cord 2.62 mm | Cord 3.53 mm | Cord 5.33 mm | Cord 6.99 mm |
|---|---|---|---|---|---|
| 1.0 | 1/16 in | 1/16 in | 1/16 in | 1/16 in | — |
| 2.0 | 1/8 in | 1/16 in | 1/8 in | 1/8 in | — |
| 3.0 | 1/4 in | 1/4 in | 1/8 in | 1/8 in | — |
| 5.0 | 1/4 in | 1/4 in | 1/8 in | 1/8 in | 1/8 in |
The finest step at 5 in belongs to the 3.53 mm cord — 1/8 in — and the coarsest to the 1.78 mm cord, at 1/4 in. The cord that is 2.0 times thicker gives a step half as coarse, which is backwards from what the section suggests.
The reason is in the first column of the previous table. A step is not set by the cord; it is set by how many times the series has doubled. The 3.53 mm series starts at 1/16 in and has already climbed to 1/8 in by this diameter. The 1.78 mm series starts finer at 1/32 in but has climbed further, to 1/4 in. Start position and number of levels decide the step, and the cord decides neither.
That is why the step-to-cord ratio is not a constant. At the bottom of each ladder the step is between 0.298 and 0.455 of the cord — roughly a third; at the top it runs from 2.43 to 7.19 times the cord. The cord is a continuous choice; the step can only double. Where the two meet, the step has to round.
The 6 steps that are not powers of two
341 steps make up the 5 ladders. 335 of them land on a power of two of 1/64 in. The other 6 do not, and they are worth naming because they are the only place where the rule bends:
- One step at the bottom of a series. In the 000 Series, -005 to -006 runs 0.013 in — 0.83 of a cell. That is the part of the series where the smaller sizes carry a thinner cord of their own, and the inside diameters in that zone are not on the main ladder.
- 5 steps that end on a shared diameter. 15.955 in and 22.940 in are each a size that appears in more than one series — the same inside diameter offered with two or three different cords, so one shaft can carry either seal.
The second case is the interesting one, because it is not an error in either table. 43 of the 48 shared inside diameters here are above 100 mm, so sharing is almost entirely a large-size arrangement; 3 of those appear in three series at once. A shared diameter is a fixed number that both ladders must pass through. Two ladders with different step sizes cannot both put a rung on an arbitrary point, and two shared points are not always an exact number of inches apart — so whichever step runs from one shared point to the next has to give way, by between 0.015 and 0.029 in.
That is the whole explanation for the exceptions. They are the seam where two grids meet, not a defect in one of them. It is also why they should not be "fixed": snapping them to a power of two would break the shared diameter, and the shared diameter is the more useful fact.
The series cut by purpose has no ladder
One series here is not cut by cord at all. The 900 Series holds 20 rings in 9 different cords, from 1.42 to 3 mm, and the cord changes with the size — each ring is there for what it seals, not to fill a grid.
Its inside diameters reflect that. Adjacent steps run from 3.1 to 23.7 cells, which is not a grid at all, and 8 of the 19 steps do not land on a power of two. There is no ladder because there is no fixed cord to climb.
That is the test for whether the rule applies to a series you are looking at: does it hold one cord across a range of diameters? If it does, expect the doubling. If the cord moves with the size, expect nothing.
What to do with a ladder
- Fix the series before the size. The same inside diameter appears in more than one series with different cords, and the cord decides the groove. Choosing the ring first and the groove second is how a seal ends up with the wrong squeeze.
- Do not expect a fine adjustment high up a series. At the top of every ladder here the step is a quarter inch or more. If the target is between two rungs, the answer is a different series — or a different seal type.
- Read the step off the band, not off the cord. A thick cord does not mean a coarse ladder. At 5 in the finer step is on the thicker of the two cords.
- When two sizes look "wrong", check whether the diameter is shared. A step that is not a power of two usually means the size appears in another series too.
Where this stops
- Groove dimensions. The step tells you which rings exist, not which one your groove takes. Depth and width come from the AS568 groove tables or ISO 3601-2, neither of which is here.
- Tolerances. A step is the distance between nominal sizes. How far a real ring may sit from that nominal is a separate table — 1/64 in of ladder says nothing about it, and the tolerance guide is where that lives.
- Metric sizes. This is AS568 only. A metric groove needs ISO 3601-1, whose sizes are a different grid and are not in this database.
- Three decimal places. The data is printed to 0.001 in, which is enough to identify every rung but not enough to measure a ring. For that you need the tolerance, not the nominal.
- Why the standard chose these numbers. The doubling is visible in the data. The committee reasoning behind it is not, and this page does not invent one.
Start from the O-ring index for the six series side by side, or open any series to read its ladder rung by rung.