Adapter Characterization
Understanding Adapter Characterization
When measuring a high-frequency RF component, the Vector Network Analyzer (VNA) must be calibrated to a specific "Reference Plane" (the exact physical tip of the test cable). If you calibrate the VNA using 3.5mm test cables, your reference plane is perfect. But what if the Device Under Test (DUT) has SMA connectors? You must screw a 3.5mm-to-SMA adapter onto the cable. Instantly, your calibration is ruined. The adapter introduces phase delay, insertion loss, and VSWR reflections. To fix this, you must perform Adapter Characterization.
Adapter characterization is the metrological process of mathematically figuring out the exact S-parameters (S11, S21, S12, S22) of the physical adapter itself. Once the VNA knows the exact electrical "fingerprint" of the adapter, it can mathematically subtract (de-embed) those effects from the final measurement. The VNA essentially moves the reference plane straight through the metal adapter, landing perfectly on the threads of the DUT.
The Dilemma of the Un-measurable Adapter
You cannot simply plug an adapter into the VNA to measure it. If Port 1 of the VNA has a 3.5mm cable, and Port 2 has a 3.5mm cable, you cannot plug a 3.5mm-to-SMA adapter into Port 2! It won't physically fit. Therefore, engineers must use advanced mathematical tricks like the Unknown Thru method, or they must characterize two identical adapters connected back-to-back, and then mathematically slice the resulting matrix exactly in half to find the parameters of a single adapter.
The raw VNA measurement includes both the Adapter and the DUT:
[Tmeasured] = [TAdapter] × [TDUT]
The VNA software instantly isolates the true DUT performance by multiplying by the inverse of the characterized adapter:
[TDUT] = [TAdapter]-1 × [Tmeasured]
Comparison
| Characterization Method | How it works | Accuracy |
|---|---|---|
| Unknown Thru Calibration | Uses the adapter itself during the 8-Term calibration process. | Extremely High (Industry Standard) |
| Adapter Removal (2 Cals) | Requires two full 12-term calibrations with and without the adapter. | Very High (Time consuming) |
| Back-to-Back Slicing | Connect two identical adapters, measure them, divide matrix by two. | Moderate (Assumes both adapters are exactly identical) |
| Phase-Delay Approximation | Just typing the adapter's physical length into the VNA Port Extension. | Terrible (Ignores VSWR and loss entirely) |
Frequently Asked Questions
Why can't I just use the VNA's 'Port Extension' feature?
Port Extension is a poor man's adapter characterization. It simply adds a mathematical phase shift to compensate for the electrical length of the adapter. It assumes the adapter is perfectly 50 ohms and has zero insertion loss. At 2 GHz, this is a decent assumption. At 40 GHz, the microscopic impedance mismatch inside the adapter causes massive VSWR reflections. Port Extension ignores these reflections, completely corrupting your S11 measurement. You must fully characterize the adapter.
What is an 'Insertable' vs 'Non-Insertable' device?
An 'Insertable' device can be plugged directly between Port 1 and Port 2 of your test cables without any adapters. For example, if Port 1 is Male and Port 2 is Female, a device that is Female on port 1 and Male on port 2 is perfectly Insertable. A 'Non-Insertable' device has the same sex on both sides (e.g., Male on both sides). You are physically forced to use an adapter to connect it to the VNA, requiring adapter characterization.
How does 'Unknown Thru' characterize the adapter automatically?
In modern VNAs, if you calibrate Port 1 (Male) and Port 2 (Male), you cannot connect them together for the 'Thru' step. You must use a Female-to-Female adapter. Instead of needing to know the adapter's specs, the VNA software relies on the principle of Reciprocity (assuming the adapter's forward loss S21 is exactly equal to its reverse loss S12). Using this mathematical constraint, the VNA solves the entire 8-Term error model and characterizes the adapter simultaneously in the background.