Connector Reference Plane
Why the Mating Face Defines the Electrical Boundary
Every two-port measurement assumes a known starting point. For a coaxial connector that point is the outer-conductor butt face: the flat shoulder where the male and female bodies seat against each other at the rated torque. Precision interfaces (SMA, 2.92 mm, 2.4 mm, and 1.85 mm) are built so the center conductor makes contact at exactly the same axial position as the outer shoulder, leaving no air gap or pin overlap. That coincidence is what allows a calibration to place the reference plane at a single, repeatable location and to treat everything beyond it as removable.
When a vector network analyzer is calibrated with a short-open-load-thru or thru-reflect-line standard set, the procedure mathematically moves the measurement plane from the raw test-port connectors out to the mating faces of the calibration standards. From that moment, the displayed magnitude and phase are referenced to those planes. A device connected there is measured cleanly; a device connected one adapter away is measured with the adapter included unless the plane is moved by port extension or de-embedding.
The sensitivity is geometric. A reflection seen at the plane travels out and back, so a small axial offset doubles into the reflection-coefficient phase. This is why a recession of only a few thousandths of a millimeter in a connector center conductor, well within normal mechanical tolerance at low frequencies, becomes a measurable error in the millimeter-wave bands where RF Essentials components operate.
Transferring and Moving the Plane
Two tools shift the plane after calibration. Port extension applies a frequency-linear phase term to rotate the plane along a low-loss, well-matched line; it is fast and adequate for trimming out a known launch but cannot cancel real reflections. De-embedding instead subtracts a measured or modeled S-parameter network of the intervening fixture, removing both its phase and its mismatch. For the highest accuracy at a non-standard plane, recalibrating directly at that plane beats either correction.
Reference-Plane Equations
Δφ = 360° × (d × √εr) / λ0
Reflection-coefficient phase (round trip doubles the path):
ΔφS11 ≈ 2 × 360° × f × (d × √εr) / c
One-way delay of an air line:
τ = d / c ≈ 3.34 ps/mm
Where d = axial plane offset, εr = relative permittivity (≈ 1 in air), λ0 = free-space wavelength, f = frequency, c = 2.998 × 108 m/s. Example: d = 0.1 mm in air at 40 GHz gives a one-way phase of ≈ 4.8°, so the round-trip reflection term is ΔφS11 ≈ 9.6°.
Plane Definition Across Connector Interfaces
| Interface | Plane Location | Max Freq | Mating Repeatability | One-Way Phase @ 0.1 mm Offset |
|---|---|---|---|---|
| SMA | Outer-conductor face | 18 GHz | ± 0.05 mm | ≈ 2.2° @ 18 GHz |
| 2.92 mm (K) | Outer-conductor face | 40 GHz | ± 0.02 mm | ≈ 4.8° @ 40 GHz |
| 2.4 mm | Outer-conductor face | 50 GHz | ± 0.013 mm | ≈ 6.0° @ 50 GHz |
| 1.85 mm (V) | Outer-conductor face | 67 GHz | ± 0.010 mm | ≈ 8.0° @ 67 GHz |
| 1.0 mm (W) | Outer-conductor face | 110 GHz | ± 0.005 mm | ≈ 13.2° @ 110 GHz |
Frequently Asked Questions
Where exactly is the reference plane located on an SMA or 2.92 mm connector?
It sits at the outer-conductor mating face, the flat shoulder where the two bodies butt together at the rated torque (about 0.56 N-m for SMA). Precision interfaces are designed so the center pin and socket contact at that same axial position, leaving no air gap. Any recession or protrusion of the center conductor relative to that face shows up directly as phase and impedance error referenced to the plane.
How do I move the reference plane after calibrating a VNA?
Use port extension or electrical delay to rotate the plane along a known line: it adds a phase term of 360 × f × τ degrees, so a 10 mm air line at 40 GHz rotates phase by roughly 480°. Port extension fixes phase and loss but not real reflections, so for adapters or fixtures with mismatch, de-embed a measured S-parameter model or recalibrate directly at the target plane.
What measurement error results from a 0.1 mm reference-plane offset at 40 GHz?
In air, 0.1 mm is about 0.33 ps of one-way delay, which is a one-way phase of roughly 4.8° at 40 GHz. A reflection (S11) travels out and back, so it picks up about 9.6° at the same 0.1 mm offset. The one-way figure drops to about 1.2° at 10 GHz, which is why plane control dominates millimeter-wave accuracy. Magnitude barely changes, but the phase rotation corrupts time-domain gating, impedance peeling, and any de-embedding that assumes a fixed plane.