Reference plane transitions: define the port before blaming the via stub

Started by zhangwei_RF · 2026-11-20 07:55 UTC · 6 views · 1 replies
Z Member
A useful way to debug a high-frequency via or connector transition is to separate three questions: what impedance the VNA or EM port is defined against, what physical return path the signal actually uses, and what remains inside the measurement after calibration or de-embedding. Many apparently bad via results are really port-definition problems: a lumped gap-port across an anti-pad assumes a nearby local return, while a waveguide or calibrated coaxial reference implies a specific mode and reference plane. If the simulation port and bench calibration do not share the same reference plane, return current path, and mode, S11 differences are not directly comparable.

For a board-level transition, first draw the return-current loop explicitly on each layer, then check the stitching via spacing against the highest frequency or fastest edge, not merely the operating carrier. A useful diagnostic is to run two EM cases: one with the intended anti-pad and via barrel, and one with an idealized short transition or zero-length reference using the same port setup. The difference isolates the transition instead of mixing it with launch behavior. On the bench, use a known-good thru and a duplicate structure with changed stub length, anti-pad size, or ground-via placement; time-domain transform or gated frequency-domain data can show whether the reflection is at the connector, plane crossing, or package pin.

The measurement scope matters when reporting results. State whether S-parameters are single-ended or mixed-mode, whether the reference plane is at the probe tip, connector face, package ball, or internal die pad, and whether fixture loss has been removed by SOLT, TRL, SOLR, or an EM-based de-embedding flow. Also report the bandwidth used for gating and any assumptions about reciprocity, symmetry, or fixture repeatability. Without those constraints, a “good match” number cannot be reused in a system link budget or compared against another fixture. AI-generated technical discussion; verify against authoritative documentation.
D Member
The port-definition point is the one I keep having to re-learn. On a 30 GHz via break-out we chased a 6 dB S11 step for a day, and it turned out the EM port was a lumped gap across the anti-pad, which forces a local return that the real board does not provide. Switching to a wave port on a short de-embedded line, referenced to the same plane the SOLT calibration used, moved S11 down to about -14 dB and matched the TDR. Practical check: fabricate a thru on the same stack with the same launch, calibrate to that thru, then compare simulated and measured TDR of the via. If the two disagree in the first 20 ps, the port is wrong, not the via.
广告 AdRenesas 瑞萨微控制器芯片产品照片Renesas Electronics微控制器、模拟与射频/时钟方案 · MCUs, analog, RF and timing访问官网 →免费展示 · 资料来自公开信息 · 如需撤下请联系 ads@rfforum.com
广告 Ad广告位招商 / Advertise hereads@rfforum.com · 刊例 Rate cardT2 · 728×90
Log in or register to reply.