VNA Insertion Loss Test | Measurement to 40 GHz

What Is a VNA Insertion Loss Test?

A VNA insertion loss test uses a vector network analyzer to measure the S21 (forward transmission) parameter of a device under test—typically a PCB trace, connector, filter, or cable—and express the magnitude of power lost as a function of frequency. The result is reported in negative decibels: a 50-ohm microstrip on Rogers RO4350B might show −0.8 dB at 10 GHz over a 100 mm (3.94 in) length, while the same geometry on standard FR-4 could exceed −3 dB. Because the VNA captures both magnitude and phase, the same sweep also reveals return loss (S11), letting you separate mismatch loss from dielectric and conductor loss in a single acquisition.

Why Insertion Loss Matters for RF PCBs

Every decibel lost in a trace is a decibel your power amplifier must make up—or your receiver sensitivity gives away. At millimeter-wave frequencies used in Ka-band radar and satcom boards, conductor loss and dielectric loss (Df) climb steeply, making accurate measurement essential for design validation. A VNA insertion loss test answers three critical questions:

• Does the fabricated board meet the loss budget? A channel that was simulated at −4.5 dB but measures −6.2 dB will fail an eye-diagram mask.
• Is the laminate performing to its datasheet Dk/Df? Lot-to-lot variation in Df can shift insertion loss by 0.3–0.5 dB over a long trace.
• Are process defects present? Excessive roughness from oxide treatment, under-etched traces, or void-filled vias all raise measured loss above prediction.

Equipment and Test Setup

VNA Selection

For measurements up to 40 GHz, a two-port or four-port VNA with a calibrated dynamic range of at least 80 dB is standard. Instruments from Keysight (PNA/PNA-X), Rohde & Schwarz (ZNA/ZNB), and Copper Mountain (Planar series) are common in PCB labs. The key specification is trace noise at the frequency of interest: at 30 GHz, you need the noise floor well below the expected S21 so that a −0.05 dB/inch difference between two laminate lots is not buried in measurement uncertainty.

Connectors and Launch Structures

The physical interface between the VNA cable and the PCB trace is the largest single source of error. End-launch SMA connectors work reliably to about 18 GHz; 2.92 mm (K) connectors extend to 40 GHz; 1.85 mm (V) connectors reach 67 GHz. Solder-less, compression-mount connectors (such as Southwest Microwave end-launch) are preferred because they are repeatable and removable. Ground-signal-ground (GSG) probes on a probe station eliminate the connector entirely and are the gold standard for coupon-level loss measurement above 20 GHz.

Calibration Standards

A full two-port SOLT (Short-Open-Load-Thru) or TRL (Thru-Reflect-Line) calibration must be performed before every measurement session. TRL calibration is strongly preferred for PCB work because it sets the reference plane on the board itself, removing connector and launch effects. The IEEE 370 standard defines a consistent methodology for de-embedding fixtures from high-frequency S-parameter data, and following it makes your results comparable across labs.

Step-by-Step VNA Insertion Loss Test Procedure

1. Design the test coupon. Include a thru line of known length (typically 150 mm / 6 in for sub-6 GHz, or 50 mm / 2 in for mmWave) plus TRL calibration structures on the same panel. A well-designed RF loss test coupon is the foundation of a trustworthy measurement.
2. Verify connector torque. Torque SMA/K connectors to 8 in-lb (0.9 N·m) with a calibrated torque wrench. Over- or under-torquing shifts the reference plane and can introduce 0.1–0.3 dB of phantom loss.
3. Perform calibration. Execute a full two-port TRL cal using on-board standards, or SOLT cal using the connector manufacturer’s cal kit. Verify by measuring the thru standard after cal—S21 should read 0.00 ± 0.02 dB across the band.
4. Measure the DUT. Connect the trace under test and sweep. Record S21 (insertion loss) and S11 (return loss) simultaneously. If S11 is worse than −10 dB at any frequency, mismatch loss is contaminating the insertion loss reading and the launch must be improved.
5. De-embed fixtures. If TRL cal was not used, apply time-domain gating or AFR (automatic fixture removal) to strip launch parasitics from the S21 data.
6. Normalize to length. Express loss as dB per unit length (dB/inch or dB/cm) so results can be compared across coupon lengths and different boards.

Understanding the S21 Result

The raw S21 curve from a VNA combines three loss mechanisms:

The differential-length (Delta-L) approach uses two traces of different lengths on the same coupon to cancel launch effects and isolate per-unit-length loss, which is especially useful for separating dielectric from conductor contributions.

Insertion Loss Budget Calculator

Before you measure, you should know what to expect. Enter your trace length, substrate Df, frequency, and conductor roughness below to estimate the total insertion loss budget for your channel. The tool returns a per-unit-length loss and a total channel loss in dB.

For example, a 75 mm (2.95 in) microstrip on RO4350B (Df ≈ 0.0037) at 28 GHz with standard rolled-copper roughness (Rz ≈ 3 µm) typically yields around −0.22 dB/cm, giving roughly −1.65 dB total. If your measured VNA result is significantly worse—say −2.4 dB—the discrepancy points to a process issue: excessive etch undercut, rough oxide treatment, or a Df that does not match the datasheet. Building a solid insertion loss budget before fabrication lets you set a clear pass/fail criterion for the VNA test.

Common Error Sources and How to Avoid Them

Even experienced labs make mistakes that corrupt VNA insertion loss data. The most frequent pitfalls—and how to guard against them—are covered in depth in our guide to insertion loss measurement pitfalls. Here are the top offenders:

• Cable movement after calibration. Flexing a test cable shifts its phase response and can add ±0.1 dB ripple. Use phase-stable cables (e.g., Sucoflex or Gore) and secure them after cal.
• Insufficient warm-up. Most VNAs need 30–60 minutes to stabilize their internal reference oscillator. Measuring immediately after power-on introduces drift.
• Ignoring connector wear. After ~500 mate cycles, 3.5 mm and 2.92 mm connectors degrade. Gauge connectors with a pin-depth gauge per the IEC 61169 series connector standards before critical measurements.
• Aliased resonances. A step size that is too coarse can miss narrow resonant nulls caused by stub effects or via transitions. Use at least 1601 points across the band, or set the frequency step to less than 1/(2 × electrical delay).
• Ambient temperature shift. Dk and Df are temperature-dependent; a 20 °C swing in lab temperature can shift measured loss on PTFE laminates by several percent. Record ambient temperature with every dataset.

Verifying Laminate Dk and Df from VNA Data

Once you have a clean S21 dataset, you can back-calculate effective Dk and Df. Effective Dk is extracted from the phase of S21 (group delay), and Df from the slope of the magnitude. This provides an independent check against supplier datasheet values and against resonator-based Dk/Df measurements such as SPDR or split-post methods. Agreement within ±3 % on Dk and ±10 % on Df between VNA trace measurement and resonator measurement is typical for well-controlled processes.

Shock and Environmental Pre-Conditioning

Insertion loss can degrade after environmental stress. Before declaring a board qualified, many programs require a VNA re-test after mechanical shock, thermal cycling, or humidity exposure. The shock profile depends on the end application—automotive radar modules follow a different severity level than airborne EW systems.

Use the tool above to identify the applicable shock test profile for your application class. After exposing the board to the specified shock or thermal shock test, re-measure S21 on the same coupon. A shift greater than 0.1 dB at the design frequency usually indicates micro-cracking in vias, delamination at the copper-dielectric interface, or connector damage. Correlating pre- and post-stress VNA data is the most direct way to catch latent reliability failures before they reach the field.

Reporting and Correlation

A professional VNA insertion loss test report should include:

• Calibration method, cal-kit part number, and verification thru result.
• Ambient temperature and humidity at time of measurement.
• S-parameter data in Touchstone (.s2p) format for import into simulation tools.
• Loss normalized to dB/inch or dB/cm, with mismatch loss separated if S11 was worse than −15 dB.
• Comparison against the design’s loss budget, with a clear pass/fail verdict.

Providing raw Touchstone files lets the design team overlay measured data on their EM simulation (in tools like Ansys HFSS) and close the loop between prediction and reality. If you need third-party measurement support, a VNA test service with calibrated fixtures to 40 GHz can supply traceable data without the capital cost of owning the equipment.

Frequently Asked Questions

What is an acceptable insertion loss for a 50-ohm PCB trace?

It depends on frequency and material. On a low-loss laminate like RO3003 (Df ≈ 0.0010), expect roughly −0.08 dB/cm at 10 GHz. On mid-tier Megtron 6 (Df ≈ 0.004), expect about −0.18 dB/cm at 10 GHz. The acceptable total loss is set by your link budget, not by a universal threshold.

Can I measure insertion loss with a two-port VNA on a differential pair?

You can measure single-ended S21 on each line individually, but true mixed-mode (Sdd21) requires a four-port VNA or a two-port VNA with a balun fixture. Four-port measurement is strongly preferred because it captures mode conversion (Scd21) that a two-port setup misses entirely.

How often should I recalibrate during a measurement session?

Recalibrate whenever you change cables, connectors, or adapters, or if the lab temperature drifts more than 2 °C. For long sessions, verify the cal every two hours by re-measuring the thru standard.

Does surface finish affect VNA insertion loss results?

Yes. ENIG adds roughly 0.02–0.05 dB/cm more loss than bare copper at 10 GHz due to the nickel barrier layer’s high resistivity. Immersion silver and OSP preserve copper conductivity and show the lowest additional loss among common finishes.

What is the difference between insertion loss and attenuation?

Insertion loss (measured as |S21|) includes mismatch loss plus dissipative (attenuation) loss. If S11 is better than −20 dB, mismatch loss is under 0.04 dB, and insertion loss and attenuation are practically equal. When return loss is poor, you must separate the two to avoid blaming the material for what is actually a launch problem.