On a 25 mm OD precision coupling, the torsional-stiffness spread between the three common flexible designs is roughly two orders of magnitude: bellows around 10,000 to 30,000 Nm/rad, Oldham in the 500 to 2,000 Nm/rad band, and elastomer-beam at 200 to 2,000 Nm/rad depending on the spider durometer [S1][S8].
The same source documents a worked wind-up case at 0.5 Nm applied torque, 1,000 Nm/rad coupling stiffness, 1.5 deg angular and 0.2 mm parallel misalignment, 10,000 rpm: elastic wind-up lands at 1.72 arc-min, and the misalignment envelope sits at 80% of the stated 2 deg / 0.25 mm limit [S1]. For a 1 Nm-continuous indexing axis, that single number already disqualifies a 200 Nm/rad beam: at 5x the rated torque, wind-up scales linearly to 8.6 arc-min, more than the positioning budget of most CNC and pick-and-place axes.
Where each coupling class sits on the stiffness ladder
R+W's published primer states the same ranking from a different angle: bellows couplings typically offer the highest torsional stiffness of any commercially available flexible coupling, with their continuous symmetry avoiding the cyclic energy storage and release that produces measurable output speed variations in asymmetric designs [S2]. That statement of position, plus the absolute number band from the Firgelli reference, is the canonical decision anchor most servo OEMs publish against.
Beam couplings are the deliberate trade-off below bellows. A 6-beam helix, as catalogued in the Huco comparison document, raises torsional stiffness relative to a 3-beam unit, extends length, and increases parallel-misalignment capacity to roughly 3-7 deg, but still trails bellows by an order of magnitude in rotational stiffness [S7]. Ruland's servo-coupling note, in turn, describes disc couplings as "very stiff torsionally, with stiffness ratings slightly lower than that of bellows couplings" [S5], putting disc just below bellows in the same high-tier group. Motion Control Tips' design summary confirms bellows as the top stiffness option, with the caveat that high torsional stiffness costs you the soft, damped response a start-and-stop servo sometimes wants [S4].
Oldham couplings are the outlier. The Firgelli comparison matrix rates Oldham torsional stiffness as "medium" against "very high" for bellows, and lists an acetal-center wear life of 10,000 to roughly 100,000 cycles under typical duty before the sliding mid-member starts producing measurable backlash [S8]. The sliding element is also the reason Oldham units tolerate the largest parallel misalignment in this group, often 1-3 mm, but the same sliding interface caps the torsional stiffness and introduces wear that a metal bellows, a one-piece helical beam, or a stacked disc pack does not.
Selection criteria mapped to axis duty
For a high-torsional-stiffness axis where wind-up under load must stay below the positioning budget, bellows is the default pick, with disc as a second choice when the bore or torque rating requires it. The 10,000-30,000 Nm/rad band on a 25 mm OD unit maps to roughly 0.06-0.17 arc-min of wind-up per Nm of applied torque, well inside the sub-arc-minute envelope most precision indexing specs demand [S1][S5].
For low-to-moderate torque axes with shock, vibration, or frequent reversal, a 6-beam elastomer coupling is the conventional answer. Helical beam designs in this class handle 3-7 deg of parallel misalignment, provide elastomer damping that a metal bellows cannot, and accept modest misalignment-induced reaction loads without complaint, which is why Design World notes beam couplings "excel on axes transmitting moderate to light torque, as on leadscrew" drives [S6][S7]. PIB's coupling comparison adds that beam couplings are "moderately compliant" with "lower torsional stiffness compared to bellows or disc" couplings, framing the trade as deliberate rather than deficient [S3].
For moderate misalignment where backlash can be tolerated on a non-positioning axis, Oldham is the cost-effective pick. The medium-stiffness tier (roughly 500-2,000 Nm/rad) and acetal-center wear life of 10,000-100,000 cycles is a reasonable spec for indexing tables, encoder drives, and slow-speed packaging shafts, but the sliding element rules Oldham out of any loop where zero backlash is a contractual requirement [S8].
Comparison matrix: criteria vs coupling class
Lining the three types up against the four criteria that drive most spec decisions: torsional stiffness at 25 mm OD ranks bellows first (10,000-30,000 Nm/rad), disc second (slightly below bellows), Oldham third (500-2,000 Nm/rad), and beam last (200-2,000 Nm/rad elastomer-dependent) [S1][S5][S7][S8]. Parallel-misalignment tolerance reverses the order, with Oldham handling roughly 1-3 mm, beam reaching 3-7 deg (which translates to comparable offsets in mm depending on length), and bellows typically limited to about 0.2-0.25 mm on a 25 mm OD unit [S1][S7][S8].
Backlash and wear behavior: bellows is zero-backlash by construction, with no sliding interface to wear; beam is also zero-backlash in the one-piece helical sense but the elastomer spider has a finite fatigue life; Oldham is the only one in this group that develops backlash as the acetal center wears, with the catalogued 10,000-100,000-cycle envelope setting the maintenance interval [S2][S8]. Wind-up at 1 Nm on a typical 25 mm OD part: bellows lands at roughly 0.06-0.17 arc-min, Oldham at roughly 1.7-6.9 arc-min, and a 200 Nm/rad beam at 17.2 arc-min, which alone disqualifies the soft beam from any sub-arc-minute spec [S1].
What each coupling cannot do
Bellows couplings fail on large misalignment, not on stiffness. Going past about 0.2-0.25 mm of parallel offset or 2 deg of angular offset on a 25 mm OD unit pushes the convolutions into plastic deformation every revolution, work-hardens the stainless, and cracks the weld root within hours [S1]. They are also intolerant of side-load reaction forces from poor mounting geometry, and bonded bellows-to-hub assemblies can degrade in corrosive media or at temperature extremes where the bonding agent is the weak link, which is why welded assemblies are specified in those environments [S2].
Beam couplings fail on precision, not on misalignment. At 200-2,000 Nm/rad, the wind-up number under modest load is the deal-breaker for any sub-arc-minute axis, and the elastomer spider has a temperature and chemical envelope that metal bellows and disc packs do not share. The PIB comparison groups beam couplings with the "moderately compliant" tier explicitly to flag this [S3].
Oldham couplings fail on zero-backlash duty and on torsional stiffness. The sliding acetal or brass center is the part that wears, develops play, and ultimately sets the maintenance schedule; the medium-tier torsional stiffness is the reason they are typically used on encoder- and instrument-type shafts rather than on the drive side of a positioning loop [S8].
Standards, sourcing, and reference math
The referenced technical sources do not pin any of these stiffness ranges to a specific ISO or AGMA standard revision, so the spec language should reference manufacturer-published torsional-stiffness curves rather than a normative document. The underlying math, treated as a linear torsional spring, is theta_rad = T / Kt, and theta_arcmin = theta_rad x 180/pi x 60, which the Firgelli calculator makes explicit and which any servo loop analysis can reproduce [S1]. For Oldham and beam types, the published wind-up should be cross-checked against the manufacturer curve at the actual operating torque, not at the rated maximum, because the elastomer or acetal response is not strictly linear near the limits.
For a deeper treatment of flexible-coupling selection on leadscrew and servo axes, the motion-control tips technical summary and Ruland's servo-coupling article remain the most-cited reference pair; for comparison against disc and gear-pack options in the same high-stiffness tier, see the disc coupling and gear coupling encyclopedia entries, and for the elastomer-beam baseline reference the coupling and clutch page. Related coverage on ball spline deflection under radial load and on ball-screw spline combined shafts for SCARA Z-theta axes is worth pulling up when the same precision axis spec is being built out, since shaft stiffness between the coupling and the load often sets the real-world wind-up budget as much as the coupling itself does.
The next trackable signal is the disc-coupling curve relative to bellows: any new manufacturer datasheet that quantifies the bellows-vs-disc delta at matched bore size will refine the second-tier pick for high-torque precision axes. A second signal worth watching is acetal-center wear-life data on Oldham units under closed-loop servo duty, which the existing 10,000-100,000-cycle band does not currently distinguish by duty profile [S5][S8].