Microstrip-to-stripline transitions: return vias, keep-out, and the stub you did not mean to build

Started by LunaWei · 2026-10-25 23:24 UTC · 5 views · 1 replies
L Admin
[Technical article, rfforum Editorial Team]

Moving an RF line from outer microstrip into an inner stripline layer is a via transition problem, not a "draw a via and hope" problem. A bad transition shows up as a notch, excess IL, or a mode you only see above a few GHz.

1. Return path first
The signal via needs nearby ground vias so return current can change layers with the signal. One lonely signal via with grounds millimetres away is a loop inductance and a slot radiator. Place at least one (often a small fence) of ground vias within a small fraction of a wavelength at your highest frequency.

2. Antipad and clearance
Oversized antipads reduce capacitance but can create a cavity; undersized ones load the via. Match the antipad to the stackup and target Z0, then check with a 3D via model, not only a 2D field calculator.

3. Unwanted stubs
If the signal via continues past the exit layer, the unused barrel is a stub — the same physics as a digital via stub. Blind/buried vias or back-drill when the stub electrical length approaches your band.

4. Launch geometry
Taper or pad size at the microstrip entry sets the discontinuity. Keep reference copper continuous under the launch; a ground void under the entry pad is a classic S11 spike.

5. Verification
Simulate the transition alone (ports on microstrip and stripline). On the bench, a gated TDR or a TRL kit that includes the same via style beats blaming the IC package first.

Discussion: for your last board above 6 GHz, how many return vias did you put around each signal via, and did you back-drill or use blind vias?

— rfforum 编辑部 / Editorial Team
L Admin
Return-via count and stub length are the two knobs that fix most microstrip–stripline S11 spikes above 6 GHz. Gating the transition alone on the bench is the right blame-assignment tool before chasing the IC.

Do you standardise a via footprint in the corporate PDK, or re-tune per stackup?

— rfforum 编辑部 / Editorial Team
广告 AdNXP Airfast 5G 基础设施射频功率多芯片模块示意图NXP Semiconductors射频功率、汽车雷达与连接芯片 · RF power, radar and connectivity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.