Comparison of TRL Calibration vs. 2x Thru De- embedding Methods Se-Jung Moon #1 , Xiaoning Ye #2 , Rex Smith #3 1 se-jung.moon@intel.com 2 xiaoning.ye@intel.com 3 rex.smith@intel.com Abstract— In this paper, TRL (Through-Reflect-Multiple Lines) and 2x thru de-embedding schemes of AFR (Automatic Fixture Removal) and SFD (Smart Fixture De-embedding) are compared from various perspectives: calibration fixture design, calibration / de-embedding procedure, and the measurement accuracy. Especially for the accuracy comparison, one of our test boards which support TRL calibration up to 40 GHz was used. In order to derive a valid conclusion, the calibration / de- embedding standards were thoroughly examined and calibration stability and de-embedding accuracy were tested. The comparison showed excellent correlation in IL, RL, NEXT and FEXT over the full frequency range of measurement, while the 2x thru de-embedding method significantly simplifies the de- embedding procedures. I. INTRODUCTION Accurate electrical characterization of high speed interconnects, such as connectors, vias, cables, and traces on a package or on a PCB (Printed Circuit Board), over the wide region of frequency is critical in successful high speed bus design. As the data rates increase, the characterization gets more difficult and challenging. Electrical characterization of high-speed interconnects is usually done in the frequency domain using a VNA (Vector Network Analyser). A DUT (Device Under Test) is typically mounted on a PCB (Printed Circuit Board), which provides connection of the DUT to the VNA, the complete removal of test fixture artifacts is very important as the test fixture manifests its own characteristics over high frequency region. The removal is done by calibration or a de-embedding process and the quality determines the accuracy of measurement [1]. Although TRL (Through-Reflect-Multiple Lines) has been known to be one of most accurate calibration methods, the complexity and difficulties in design and measurement limit its usage. For this reason, various simplified de-embedding techniques have been reported. Among those various de- embedding techniques, 2x thru de-embedding scheme such as AFR (Automatic Fixture Removal) [2, 3], SFD (Smart Fixture De-embedding) [4], ISD (In-Situ De-embedding) [5, 6], etc. stands out due to its simplicity. Even AFR claimed its accuracy was comparable or equivalent to the one of TRL [7]. In this paper, we validate the accuracy of AFR and SFD compared to TRL. For proper evaluation, we used our test board which was designed to support TRL calibration up to 40 GHz with high measurement stability and accuracy [8]. As 2x thru de-embedding needs only one structure of Primary thru from TRL standards, the test board was reused for the 2x thru de-embedding methods. This paper is organized as follows: First, the comparison between TRL and 2x thru de-embedding methods is summarized in terms of complexity in a test fixture design and calibration / de-embedding procedure. Second, fundamentals of TRL calibration / de-embedding structure design and the evaluation procedures are presented. Third, measurement data of TRL, AFR and SFD are compared. II. TRL VS. 2X THRU DE-EMBEDDING, COMPARISON OF TEST FIXTURE DESIGN AND MEASUREMENT PROCEDURE The differences between TRL versus 2x thru de-embedding methods are summarized in Table 1 considering typical design and measurement procedure in our lab. Table 1. TRL vs. 2x Thru De-embedding TRL calibration kit has four fundamental elements: Primary thru (2x thru), Reflect, Multiple lines and Load. On the other hand, 2x thru de-embedding needs only one element, a Primary thru for de-embedding. Hence a test board area covering de-embedding structures can be reduced by as much as 80 %, when 2x thru de-embedding is adopted. Also each TRL calibration standard needs to be designed in a way to follow its own design rule. However, designing the TRL standards properly satisfying its own design rule is not a trivial task. So the 2x thru de-embedding methods can save time and effort in the test board design. Once a test board is designed and the board is fabricated, all the components of TRL calibration standards should be evaluated before making measurements. Typical TRL calibration starts with setting up a calibration Kit for each element, by providing various customized information and 978-1-4799-1993-2/15/$31.00 ©2015 IEEE 176