A disc coupling with a single flex plane (one disc pack) does not accept parallel shaft offset at all, per SKF catalog guidance: parallel offset is not permissible in single-configuration disc couplings, and a second disc pack is required to convert offset into opposing angular flexes [S4].
The arithmetic behind the limit is small-angle geometry. With spacer length L between two disc packs and per-pack angular limit θ_limit, the allowable parallel offset is Δ_allow = L · tan(θ_limit); the actual pack angle is θ = atan(Δ / L). For a Firgelli-worked Thomas example at L = 114.6 mm and θ_limit = 0.25°, Δ_allow = 0.5 mm, which is the geometric ceiling before the pack runs out of flex margin [S1].
Why single disc packs reject parallel offset
A single disc pack is a torsionally stiff membrane loaded in bending only along the shaft axis. Pushing one hub sideways relative to the other loads the disc as a beam-column, not a flexure, and pushes that force straight into the bearing housings. That is why the SKF W4 single-flex disc coupling is specified for angular and axial misalignment only, with the parallel-offset envelope left for the W4 double-flex build [S4].
Operationally, parallel offset on a single pack shows up as a steady radial reaction force. Regal Rexnord's Kop-Flex failure-analysis note ties the symptom to a per-pack angular cap of 0.10° at recommended operating conditions, with combination of angular and parallel offset the most common misuse pattern seen on returned units [S3]. The 0.10° per-pack number is a sustained-duty cap, distinct from the 0.25° per-pack short-stroke limit used in sizing examples [S1][S3].
Double disc packs, Thomas geometry, and offset scaling
Adding a second disc pack, separated by a spacer, lets parallel offset be absorbed as equal-and-opposite angular flex at each pack. R+W's coupling technology blog makes the rule explicit: the greater the distance between disc packs, the larger the parallel misalignment which can be compensated for at a given bending angle, which is why long-spacer floating-shaft designs are used on turbine and refinery trains [S7].
For Thomas / Rexnord double-flex couplings, published per-pack angular limit is on the order of 0.25° with a full coupling envelope of roughly 0.5 mm parallel offset across the assembly at short spacers, scaling linearly with L [S1]. Lovejoy's flexible-coupling technical note describes an alternative highly specialized design that accepts up to 3° angular misalignment but only up to 0.5 inches (12.7 mm) of parallel offset, illustrating the angular-versus-parallel tradeoff that defines every disc-coupling selection (2018) [S6].
Oldham sliding disc: a different mechanism, a different limit
An Oldham coupling is not a disc-pack coupling. It uses a center sliding disc keyed 90° offset between two hubs, so the hubs can slide radially relative to each other while the disc transmits torque through its key slots. Ondrives.US markets the geometry specifically for parallel (radial) shaft offset with zero backlash, accepting more parallel offset than typical beam or bellows couplings because the motion is sliding rather than flexural [S2].
The practical envelope is large. The trade is wear: the Oldham center disc is a sacrificial part, deliberately designed as a mechanical fuse that breaks before downstream gearboxes, lead screws, or bearings are damaged, so the parallel-offset envelope is paid for in periodic disc replacement rather than in bearing reaction [S2].
How to read the offset number on a dial indicator
Field measurement has a footgun. Rexnord's CP3-008 installation note states: if parallel offset is measured by rotating the hubs with a dial indicator on the outside diameter, the total indicated reading should be divided by 2 to get the true offset, because the indicator rides twice the actual shaft-to-shaft displacement as the hubs are rotated [S8].
That factor-of-two matters when the spec is tight. A 0.5 mm allowable offset [S1] corresponds to a 1.0 mm full-indicator reading; missing the divide step will pass a coupling that is already at its geometric limit and will fail prematurely, which is the failure pattern Regal Rexnord's 0.10° per-pack cap is meant to prevent [S3][S8].
Decision criteria: which coupling for which job
Three couplings, three jobs, with parallel offset as the differentiator. A disc coupling double-flex (Thomas / Rexnord / Kop-Flex W4 double) is the right pick when the driver is high-speed, high-torque, and lubrication-free, accepting ~0.5 mm offset at 114.6 mm spacer with a 0.25° per-pack limit and zero backlash [S1][S3][S4]. An Oldham sliding disc is the right pick when the priority is large parallel offset (up to ~10% of OD), zero backlash, and a deliberately replaceable wear part, accepting speed and torque limits well below disc-pack designs [S2][S9]. A single-flex disc pack is the wrong pick the moment any sustained parallel offset is in the alignment budget, because SKF catalogs flatly disallow it [S4].
Selection also has to clear the rest of the gate. Turn2Engineering's 2026 coupling design guide restates the rule: torque, peak, bore, speed, combined misalignment, stiffness, backlash, environment, and serviceability all have to pass for the same candidate, and misalignment must be checked at the actual load and speed, not at no-load [S5]. Parallel offset capacity in the catalog is meaningless if the rest of the drivetrain gates fail on speed or torque. For related reading on a high-flex variant, see Sliding Disc Couplings for Turbine Shaft Axial Float; for a rebar-coupler case where slip tolerance is similarly geometry-bound, see Type 2 Rebar Coupler Slip Tolerance.
Limits, failure modes, and the right way to spec
Parallel offset on a disc coupling is not a soft tolerance. Going past Δ_allow on a double-flex design does not give more misalignment, it converts the disc packs from pure bending into a beam-column load, raising bearing reaction forces and shortening disc-pack fatigue life. The visible symptom is fretting on the disc-pack bolt holes and a hot bearing housing downstream of the coupling [S3][S7].
The trackable signal to watch on the next design review is the ratio L · tan(θ_limit) to measured indicator reading (÷2): the coupling is in spec only when that ratio is above 1.0 with margin for thermal growth. The Oldham alternative, sized at ≤ 10% of OD, trades that margin for a wear part that has to be stocked, with a center disc SKU on the shelf before commissioning [S2][S8][S9].
The underlying component specifications are covered under limit switch, and limit switch box.