RF connector selection and signal integrity

A coaxial link is a transmission line, and every connector in it is a short section of transmission line with its own geometry. When that geometry is right, the connector is nearly invisible to the signal; when it is wrong — the wrong impedance, a worn interface, a poorly terminated cable, an under-torqued coupling nut — the link loses power, reflects energy back towards the source and, in multi-carrier systems, can generate interference of its own. This article covers the parameters that decide RF connector performance and how to use them when selecting a connector family and a cable assembly.

Characteristic impedance: 50 Ω and 75 Ω

The characteristic impedance of a coaxial line is set by the ratio of the outer conductor's inner diameter to the inner conductor's diameter, and by the permittivity of the dielectric between them. Two values dominate. 50 Ω is the standard for RF and microwave equipment, wireless infrastructure, test and measurement, and most antenna feeds; it is a practical compromise between power handling and conductor loss. 75 Ω is used in video, broadcast, cable television and many timing and telecom interfaces, where the receiving end is high-impedance and lowest attenuation per metre is the priority.

What matters for signal integrity is that source, cable, connectors and load share the same impedance. A connector does not have to be 'better' than the cable; it has to be the same impedance as the cable and have a controlled transition into it. That is why an RF connector is always specified for a particular cable group, with its own crimp or clamp parts and strip dimensions.

Connector families and typical frequency ranges

Each family has an upper frequency set mainly by its size. Above a certain frequency, the coaxial geometry can support higher-order propagation modes besides the wanted TEM mode, and performance becomes unpredictable. Smaller outer-conductor diameters push that limit up; larger ones carry more power and are mechanically more robust. The interface dimensions of most families are standardised in the IEC 61169 series of sectional specifications, but the usable range of a specific part also depends on its design, its cable and how it is terminated.

Two points follow from the table. First, mechanical compatibility is not the same as electrical equivalence: SMA mates mechanically with 3.5 mm and 2.92 mm precision connectors, but a worn or out-of-tolerance SMA can damage a precision test port, so test labs commonly protect expensive ports with connector savers. Second, reverse-polarity variants (for example RP-SMA, where the pin and socket are swapped relative to the coupling body) mate mechanically with the standard version but make no centre-conductor contact. Both are routine causes of 'mysterious' link failures.

Reflections: VSWR and return loss

Reflection coefficient (Γ)

The ratio of reflected to incident voltage at an impedance discontinuity: Γ = (Z_L − Z₀) / (Z_L + Z₀).

Return loss (RL)

How far the reflected power lies below the incident power, in dB: RL = −20 log₁₀|Γ|. Larger is better.

VSWR

Voltage standing wave ratio, (1 + |Γ|) / (1 − |Γ|). 1.0 is a perfect match; larger is worse.

Mismatch loss

The power that does not reach the load because it was reflected: −10 log₁₀(1 − |Γ|²) dB.

The table shows why connector VSWR matters more than the mismatch loss alone suggests. A single connector with a VSWR of 1.2 wastes very little power, but reflections from several discontinuities along a link add up — sometimes constructively at particular frequencies — producing ripple in the frequency response, degraded receiver sensitivity and, in transmitters, reflected power at the amplifier output. In wideband and digital modulation systems, ripple and group-delay variation are often more important than the average loss.

Datasheets usually state VSWR as a maximum over a frequency band, sometimes as a frequency-dependent expression of the form VSWR ≤ a + b·f. A connector's figure applies to the connector as tested, usually in a fixture or with a specified cable. In a finished cable assembly the dominant contributors are frequently the cable-to-connector transition and the quality of termination — strip lengths, braid dressing, centre-contact seating — which is why assembled RF cables should be specified and tested as assemblies.

Insertion loss and cable attenuation

Insertion loss is the reduction in transmitted power caused by inserting a component into the link. For a connector pair it is small — datasheets often give it as a coefficient multiplied by the square root of frequency — but it is not zero, and it is counted once for every mated pair and adapter in the path. For any link longer than a few tens of centimetres, the cable dominates.

Coaxial cable attenuation rises with frequency for two reasons. Conductor loss increases roughly with the square root of frequency because of the skin effect, which confines current to an ever thinner layer of the conductor surface. Dielectric loss increases roughly in proportion to frequency. At lower frequencies conductor loss dominates; at higher frequencies the dielectric term grows in importance. Larger cables have lower attenuation but are stiffer, heavier and require larger connectors, so cable size and connector family are chosen together. Attenuation is normally specified at 20 °C and increases with temperature.

Average power handling also falls with frequency, because the heat generated by loss rises while the cable's ability to dissipate it does not. It is further reduced at high ambient temperature and altitude. Peak power is limited by voltage breakdown in the connector interface. Both are datasheet values for the specific connector and cable.

Passive intermodulation

Passive intermodulation (PIM) is the generation of unwanted frequencies when two or more strong signals pass through a component that is slightly non-linear. In a connector the non-linearity comes from the contact interface: loose or low-pressure junctions, oxide films, contamination, metal particles, and ferromagnetic materials such as nickel plating in the current path. The third-order products (2f₁ − f₂ and 2f₂ − f₁) can fall inside a receive band, where a signal far below the transmit level is enough to desensitise the receiver.

PIM matters wherever several high-power carriers share a path with a sensitive receiver: cellular base stations, distributed antenna systems and multi-band antenna sites. It is measured with two test tones, and the result is expressed in dBc relative to the carrier level; the IEC 62037 series defines the measurement methods. The 7-16 and 4.3-10 families are designed for low PIM, and the 4.3-10 interface separates the electrical contact from the mechanical coupling so that its electrical performance depends less on coupling torque.

• Specify a PIM limit and test conditions only where the application needs it; PIM-tested parts and assemblies cost more to make and verify.
• Avoid nickel in the RF current path for low-PIM links; check the plating stack in the datasheet.
• PIM is sensitive to workmanship: clean interfaces, correct torque, no metal swarf from cable preparation and no movement of the cable at the connector.
• A PIM figure is valid for the configuration in which it was measured. A different cable or a re-terminated connector needs its own measurement.

Mating, torque and handling

Threaded RF connectors are designed for a specific coupling torque. Too little, and the outer conductors do not seat properly, increasing VSWR, insertion loss and PIM and making the link sensitive to vibration. Too much, and the interface deforms, the centre contact can be pushed back, and the mating cycle life drops. Published values depend on the family and on the body material: stainless-steel SMA, for example, is specified at a higher torque than brass SMA. Use the value in the manufacturer's datasheet and a calibrated torque wrench of the correct size.

1. 01Inspect both interfaces before mating: no debris, no bent or recessed centre contacts, no damaged threads.
2. 02Align the connectors axially and push them together straight before engaging the thread.
3. 03Turn only the coupling nut. Rotating the connector body grinds the contact surfaces and wears the interface.
4. 04Tighten to the specified torque, and support the cable so that its weight does not lever on the interface.
5. 05Clean interfaces with suitable solvent and lint-free swabs where needed; check that the solvent is compatible with the dielectric.

Cable routing also affects electrical performance. Bending a coaxial cable tighter than its minimum bend radius deforms the dielectric and changes the impedance locally; repeated flexing near the connector is a common cause of intermittent faults. Strain relief, cable clamps and correct bend radius belong to the RF design as much as the connector choice. The general practices are covered in cable assembly design considerations, and shielding effectiveness — relevant where RF cables run near sensitive circuits — in shielded vs unshielded connectors.

A selection checklist

1. 01Impedance: 50 Ω or 75 Ω, matched across equipment, cable, connectors and adapters.
2. 02Frequency: the highest operating frequency, including harmonics or wideband content, with margin below the family's specified limit.
3. 03Power: average and peak, at the highest operating frequency and ambient temperature.
4. 04Return loss and insertion loss: what the link budget allows, specified for the assembled cable, not only for the connector.
5. 05PIM: whether multiple carriers share the path with a receiver, and if so the limit and the test conditions.
6. 06Mechanics and environment: coupling type, vibration, mating cycles, panel space, sealing, outdoor exposure — the IP rating applies to the mated pair as specified.
7. 07Cable: size, flexibility, attenuation at the operating frequency and the connector's approved cable group.

Reading the RF section of a datasheet critically — which frequency each value applies to, whether it was measured on a connector or an assembly, and with which cable — is covered in more depth in understanding connector datasheets.