Adapter Removal (Technique)
Understanding the Adapter Removal Technique
In RF metrology, the most dreaded scenario is testing a "Non-Insertable" device. Imagine your VNA test cables end in two 3.5mm Male connectors, but the filter you need to test has 3.5mm Male connectors on both sides. You are physically forced to use a Female-to-Female adapter to connect the circuit. If you just screw it on, the adapter's electrical phase delay and VSWR reflections will completely corrupt the filter's data. Historically, the only way to solve this was the brutal, time-consuming Adapter Removal Technique.
Developed before modern algorithmic solvers like the "Unknown Thru" method existed, Adapter Removal requires the engineer to perform two completely separate, exhaustive 12-Term calibrations. It is an exercise in extreme patience and matrix algebra.
The Two-Calibration Process
First, the engineer attaches the adapter to Port 1. They treat the adapter as a permanent part of the cable, and perform a full SOLT calibration at the end of the adapter and the end of Port 2. Second, they take the adapter off Port 1, screw it onto Port 2, and perform an entirely new SOLT calibration. The VNA's internal computer now has two massive error-correction matrices. Because the adapter was on the left side during Cal 1, and the right side during Cal 2, the software can use matrix division to completely isolate and mathematically annihilate the adapter from the final dataset, resulting in a flawless measurement of the non-insertable device.
Let [Cal2] be the error matrix with the adapter on Port 2.
The VNA uses cascaded T-parameters to solve for the adapter's true matrix:
[TAdapter] = [TCal1] × [TCal2]-1
Once the adapter is characterized, it is completely de-embedded from the final measurement, moving the reference plane exactly to the device's connectors.
Comparison
| Metrology Method | Cals Required | Standard Kit Wear & Tear | Accuracy |
|---|---|---|---|
| Adapter Removal | Two Full Cals (Lots of screwing) | High (Standards used twice) | Excellent |
| Unknown Thru (Modern) | One Cal (Using adapter as Thru) | Low | Excellent (Industry Preferred) |
| Electronic Cal (ECal) | One Cal (Automated) | None (Internal solid-state switching) | Excellent (If ECal matches the connector sex) |
Frequently Asked Questions
Why does nobody use the Adapter Removal technique anymore?
Because it is infuriatingly tedious and prone to human error. You have to connect and disconnect delicate calibration standards (Short, Open, Load) dozens of times. Every time you torque a connector, you introduce microscopic stress and phase variances. The modern 'Unknown Thru' algorithm accomplishes the exact same mathematical de-embedding using only one single calibration step, saving 20 minutes of labor and significantly reducing wear on the $5,000 calibration kit.
Can Adapter Removal be used if Port 1 is Waveguide and Port 2 is Coaxial?
Yes! This is actually one of the few areas where classic Adapter Removal is still highly relevant. If you are testing a transition component (like a Coax-to-Waveguide adapter), Port 1 is a coax cable and Port 2 is a waveguide horn. You cannot use 'Unknown Thru' because you cannot physically plug a waveguide into a coax cable. You must perform a Coaxial calibration on Port 1, a Waveguide calibration on Port 2, and use Adapter Removal math to bridge the two distinct physical mediums.
What happens if the adapter is physically bumped between the two calibrations?
The entire process is ruined. The math relies on the adapter's S-parameters being perfectly identical in Calibration 1 and Calibration 2. If the cable is bumped, the phase of the cable changes by a few degrees. The matrix division will fail to perfectly cancel out the adapter, resulting in massive, artificial ripples in the final S11 and S22 measurements of your device.