Angular Contact Bearing Pairing: Choosing DB, DF or DT and a Preload Route Before the RFQ

Rule out arrangements that cannot carry your axial load direction

Start with the direction of the thrust, because it eliminates options before any stiffness argument begins. A single-row angular contact bearing has its raceways offset along the axis, and that offset is what lets it take axial load in one direction only⁹.

If your duty reverses — a spindle that cuts in both directions, a shaft that must be located axially both ways — a single bearing cannot do the job. The arrangement has to be a pair⁹.

Tandem (DT) is not the answer to a reversing load either. In DT the two bearings share the load in the same direction, which raises one-directional axial capacity, but it does not add capacity in the opposite direction⁸⁹.

Where a tandem pair is used, the opposite thrust still has to be taken by something else on the shaft³. So the first cut is simple: bidirectional axial load points to DB or DF, and DT is reserved for a duty that is heavy in one direction only.

DB or DF: moment rigidity against misalignment tolerance

Both DB and DF carry bidirectional axial load, so the choice between them is about moment and alignment, not thrust capacity. In back-to-back (DB) the contact-angle apexes point outward, which gives the pair a large effective span.

That span is what produces high moment rigidity and resistance to overturning¹². DF inverts the geometry, shortening the effective span and lowering moment capacity.

But the shorter span is more tolerant of angular misalignment between shaft and housing⁸¹². Mounting practice adds a second condition.

Where you need interference between the inner ring and the shaft and mounting error is possible, face-to-face simplifies assembly¹³. Where the requirement is high rigidity under preload, back-to-back is the recommended arrangement¹³.

That is why DB is described as the most common and versatile arrangement for machine spindles¹². What you cannot get from this evidence is a number.

No source here gives a stiffness or moment-load threshold that says where DB stops being adequate and DF becomes necessary⁸¹²¹³. Treat the choice as a qualitative one and put the moment load and alignment budget in front of the supplier.

Matched duplex set or non-matched pair with ground spacers

Preload in a paired set is most often achieved by ordering a matched duplex pair¹². In a factory-matched set the preload or axial internal clearance is already defined by the spacer rings supplied with the pair.

So the setting work is done before the bearings reach you⁴. Pairing non-matched bearings moves that work in-house.

It requires preliminary measurement of the relevant axial dimensions of the bearings and, where applicable, the housing; a decision on the target axial clearance. And, for an O (DB) arrangement, the manufacture of an inner spacer ring in addition to an outer one, ground to the accuracy the service conditions demand⁴.

If a preloaded pair is to be supported rigidly in a back-to-back arrangement through the housing, the housing width also has to be determined to work out the effective spacer length⁴. Spacer width is not a neutral detail once the pair is assembled.

If the shaft spacer is narrower than the housing spacer, preload increases in the O (DB) arrangement and decreases in the X (DF) arrangement²². That asymmetry is the lever you use to trim preload on a non-matched pair.

And it is the reason a matched set, where the supplier has already fixed the geometry, removes a variable from your assembly line.

Choosing a preload class or a discrete force value

Preload is quoted either as a class or as a discrete force, depending on which specification system your drawing follows. The ISO route uses GA, GB and GC for light, medium and heavy preload; the US route uses suffix codes tied to a stated force in newtons³.

Preload itself is what removes internal clearance, raises stiffness and prevents ball skidding, so the class you name sets the behaviour of the pair¹⁸. What the table does not give you is the value for your bearing size.

The force behind a GA, GB or GC class, and the axial displacement or contact-angle change any of these produces, is not published in this evidence³¹⁸. Quote the class or the discrete force, then ask the manufacturer for the per-size figure and the resulting displacement.

Contact angle and speed ceiling before you lock the arrangement

The contact angle is the other half of the load-versus-speed decision, and it is fixed at order, not at assembly. A steeper angle raises axial capacity and axial rigidity; a shallower one supports higher speed and more radial load⁸⁹¹⁸.

High-precision, high-speed bearings typically use 15°, while 40° is where axial capability is maximised⁸⁹. Speed then constrains the cage and the lubrication.

Where a bearing is selected to run within 80% of its limiting speed, the manufacturer's engineering department should review the application and recommend the lubrication method and cage¹⁰. Where operating speed is much greater than the limiting speed, a stronger cage construction is necessary because of the centrifugal stresses involved¹⁰.

One supplier states that small-angle 15° bearings can exceed 30,000 RPM under oil lubrication for certain sizes. But that is a supplier-specific figure for particular sizes rather than a limit you can apply to your own selection¹⁸.

So before you commit the arrangement, check the operating speed against the limiting speed for the size you intend to buy, and route the lubrication and cage decision to the supplier rather than assuming it.

Whether the published equivalent-load factors apply to your preloaded pair

If you plan to compute L10 life yourself, the arrangement you chose changes the arithmetic. The dynamic equivalent load factors X and Y differ between a single-row or tandem arrangement and a DB or DF pair.

And they also depend on the contact angle and on the ratio of axial to radial load²⁵. At a 15° contact angle with Fa/Fr above e, for example, Y is 1.47 for a single-row or tandem arrangement against 2.39 for a back-to-back or face-to-face pair²⁵.

Using the wrong column moves the equivalent load, and therefore the computed life. There is a further condition to check.

The paired-arrangement equivalent-load equations are stated to apply when internal clearance and preload during operation are zero¹⁵. A preloaded pair does not meet that assumption, so the published factors are a starting point rather than the final answer for a preloaded set.

Ask the supplier which equivalent-load basis they used for the life figure they quote, and whether it accounts for the preload you specified.

What to obtain from the supplier before the RFQ goes out

The published evidence names the preload and mounting requirements but stops short of the numbers you need to buy against. These are the items to request rather than assume.

• ✓Per-size preload value for the class or discrete force you specified, plus the resulting axial displacement and contact-angle change
• ✓Precision tolerance class the supplier proposes for the arrangement and preload, and the standard it is quoted against
• ✓Shoulder and fillet dimensions, and locknut thread data, for the shaft and housing you intend to use
• ✓Whether the pair is supplied as a matched duplex set with predefined spacer rings, or as individually packaged bearings you must pair and shim
• ✓Whether the bearings are universal-matchable, so any two from stock can be mounted together
• ✓Thermal preload change over your operating temperature range — some suppliers offer a thermal range calculator for this
• ✓L10 life for your actual load spectrum and the arrangement/preload combination, not a catalogue rating

Where the sources disagree or leave a figure open

A few quantities appear in more than one form across the sources, and the differences matter when you write the specification.

What the sources do not establish

• No per-size preload values, and no resulting axial displacement or contact-angle change for any preload class or discrete force.
• No precision tolerance class (ABEC/ISO P4, P2) tied to a particular preload or arrangement.
• No quantitative stiffness or moment-load threshold for deciding DB against DF.
• No thermal preload change quantification and no limiting-speed tables for specific arrangements.
• No shoulder or fillet dimensions, locknut thread data, or universal-matchability information.
• No L10 life results for specific arrangement and preload combinations or load spectra.
• No angular-contact-specific unit cost, MOQ or matched-pair availability data.
• Spring preload and shim or locknut preload methods are named in the requirement but not substantively described in the evidence, so they cannot be compared with matched-set preload here.