A seal that works by being squashed
An O-ring seals because it is deliberately installed somewhere it does not quite fit. The groove is machined shallower than the ring is thick, so closing the assembly squashes the rubber cross-section a controlled amount, and that compression presses the ring against both metal surfaces hard enough to block the fluid. Everything in O-ring design is bookkeeping around that squash: how thick the ring is, how deep and wide the groove is, and what percentage of compression, volume fill, and stretch the combination produces.
The system runs on three percentages — squeeze, gland fill, and stretch — and each has a published working range from the seal-industry design charts. This guide walks all three for one running example: an AS568 -214 O-ring sealing a 1-inch static piston, dimensioned from dash number to finished groove. The scope is the static radial gland — the common piston-and-bore case where nothing slides in service — which is also the case the verified design chart covers.
The dash number: one digit of series, the rest is size
AS568 is the U.S. dash-number catalog for O-rings, and the three-digit code splits neatly: the first digit picks the cross-section series, and the remaining digits step along an inside-diameter ladder, mostly in nominal 1/16-inch increments. Five cross-sections cover the standard range — the -0xx series is 0.070 in thick, -1xx is 0.103 in, -2xx is 0.139 in, -3xx is 0.210 in, and -4xx is 0.275 in. So the example -214 decodes on sight: 0.139 in cross-section, with the 14 marking its catalog position along the ID ladder, landing at 0.984 in inside diameter.
Two conventions in the system trip up newcomers. First, an O-ring is specified by ID and cross-section, never by OD — the outside diameter is just ID + 2 × CS, an arithmetic consequence (1.262 in for the -214) that suppliers print for convenience but no standard fixes. Second, the actual ID deliberately runs a few thousandths under its fractional nominal, so a "1-inch" ring measures 0.984 in and arrives slightly tight on its groove. Both conventions exist because groove design works from ID and cross-section: the ID sets the stretch onto the gland, and the cross-section sets the squeeze.
Squeeze: the working compression
Squeeze is the fraction of the cross-section the closed gland compresses: (W − L)/W, where W is the ring's cross-section diameter and L the gland depth. The design chart builds the right squeeze into its depth numbers — for static radial seals the bands run from roughly 22–32% on the skinny 0.070 in sections down to 15–20% on the fat 0.275 in sections. The trend is not arbitrary: small cross-sections need proportionally more squeeze because manufacturing tolerances consume a bigger share of a thin section than a thick one.
Squeeze is also why tolerances must be checked at both extremes rather than at nominal. A -214 carries a cross-section tolerance of ±0.004 in, and the groove depth has its own machining band, so the honest question is whether a skinny ring in the deepest groove still compresses enough, and whether a fat ring in the shallowest groove compresses too much. For the example gland, that worst-case analysis spans 16.3% to 22.4% — and the chart's printed band for the 0.139 in family is 16–23%, so every legal combination of ring and groove lands inside it.
Gland fill: why the groove must stay under 90% full
The second percentage compares volumes instead of thicknesses. Gland fill is the ring's cross-sectional area, π/4 × W², divided by the groove's area, depth times width. The handbook design range is 60–85% fill with 75% as the optimum — and a hard ceiling of 90%, because at least a 10% void in the gland is essential. The reason is a property of the material class itself: rubber is incompressible. It changes shape freely but not volume, so when temperature rises or the fluid swells the compound, the displaced volume must have somewhere to go.
A gland filled past the ceiling becomes hydraulically locked — the trapped, expanding rubber generates forces that blow the ring into the clearance gap or split the groove corners. This is the least intuitive failure in O-ring work precisely because it comes from too much sealing material, not too little. It is also why the groove is always wider than the ring is thick: the extra width is not sloppiness, it is the expansion room the incompressible cross-section will eventually claim.
Stretch: the 5% installation limit
The third percentage is set at assembly, before any pressure arrives. Installing the ring on a male gland stretches its ID onto the groove root diameter, and the handbook caps that installed stretch at 5%. Beyond the cap two things go wrong at once: the stretched ring's cross-section thins measurably, quietly eating the squeeze the groove depth was designed to deliver, and most compounds age faster while held in tension.
The other direction has its own failure. If the groove root is smaller than the ring's ID, the stretch goes negative — the ring is compressed circumferentially, and instead of seating it bunches and waves in the groove. The fix is not force but catalog position: drop to the next smaller dash number and let the slight designed-in tightness of the AS568 ID ladder do its job of keeping the ring seated.
Groove numbers come from a chart, not a guess
None of the groove dimensions in this guide were derived from first principles, and that is the point: static gland design is a lookup, standardized in the Parker O-Ring Handbook (ORD 5700) Design Chart 4-1 for industrial static radial seals. For each cross-section family the chart fixes the gland depth, the groove width (sized for use without back-up rings), the maximum bottom radius, and the diametral clearance range the mating parts must hold. The clearance number matters more than it looks: it is the extrusion gap, the annular space pressure will try to force rubber into, and the chart keeps it small — thousandths, not hundredths.
Machining to the chart is what makes the three percentages come out right simultaneously — depth delivers the squeeze band, the depth-times-width area holds fill in the 60–85% range, and the sealing surfaces get finished to the handbook's 32 µin RMS. The O-ring groove calculator runs this whole procedure from a dash number: it returns the verified Chart 4-1 gland, then recomputes squeeze across the tolerance extremes, worst-case gland fill against the 90% gate, and installed stretch against the 5% limit, each with a pass/fail verdict.
One -214 in a static gland, dimensioned end to end
Assemble the running example. The ring: AS568 -214, cross-section 0.139 in, ID 0.984 in, computed OD 1.262 in. The chart gland for the 0.139 in family: depth 0.111–0.113 in, groove width 0.187–0.192 in, bottom radius 0.010–0.025 in, diametral clearance held to 0.003–0.006 in. Squeeze across the tolerance extremes: 16.3% to 22.4%, inside the printed 16–23% band. Fill: 0.01517 in² of rubber in a nominal 0.02122 in² groove is 71.5%, rising to 77.4% in the worst-case fat-ring, tight-groove combination — comfortably under the 90% ceiling.
Last, stretch: on a groove root diameter of 1.004 in, the 0.984 in ID installs at (1.004 − 0.984)/0.984 = 2.0%, well under the 5% limit and positive enough to keep the ring seated during assembly. Every number in the paragraph above came from two documents — the AS568 size table and Design Chart 4-1 — plus three lines of arithmetic. That is the entire discipline of static O-ring design: no tuning, no iteration, just the right lookups checked at the right tolerance extremes.
The gland in one pass
Recap the -214 build: dash decodes to 0.139 × 0.984 in; the chart gland is 0.111–0.113 deep and 0.187–0.192 wide with a 0.003–0.006 in clearance; squeeze lands at 16.3–22.4% against a 16–23% band; fill runs 71.5% nominal and 77.4% worst case against a 90% ceiling; stretch installs at 2.0% against a 5% cap. Three percentages, each with a printed range, each checked at the unfavorable end of tolerance — that is what separates a gland that seals for years from one that leaks on the first hot day. And when the application turns dynamic or the seal becomes a face seal, the dash dimensions stay the same but the chart changes; the lookup discipline carries over even though these particular numbers do not.