【OEForum_Ed】Bias-T at 112G: when the inductor's SRF shows up as a hole in S21, do you first change the inductor, the decoupling, or the return path?

Started by OEForum_Ed · 2026-11-18 01:30 UTC · 10 views · 2 replies
O Member
【Cross-site note from oeforum 编辑部 / OEForum_Ed】

Hi rfforum — an analog / interconnect question from the 112G optical-module bench that we think belongs on an RF forum.

On a 112G PAM4 TOSAs we keep seeing a dip in S21 that looks like a driver or laser bandwidth limit, but sometimes it is the Bias-T: the DC bias and the RF path share one line, and the inductor's self-resonant frequency (SRF) lands inside the working band. We wrote a short note on oeforum: https://www.oeforum.com/t/1518

The split we would like to check with RF benches:

1) The inductor has to look like a high impedance at the modulation frequency and still pass the bias current cleanly. If its SRF sits near the operating band, S21 shows a hole. Changing drive amplitude will not move that hole.

2) Decoupling on the DC side and the RF return path can create a second resonance even after the inductor itself is “wide enough on the datasheet”. One change at a time — inductor, decoupling, or return/shielding — and keep the trace, otherwise you cannot tell which cut worked.

3) A module-level eye or TDECQ failure is a late symptom. If you can see the same notch on a fixture S21 or a Bias-T breakout, fix the bias network before you retune the TOSA.

Questions for the RF / components / test corner:
- When you catch a Bias-T self-resonance at 56G/112G, do you first swap the inductor, retune decoupling, or redo the return path?
- How do you put SRF and bias-network resonance into the acceptance table — a fixture S21 mask, or only a module-level test?

Technical only; soft promo = 0.

— oeforum 编辑部 / contact@oeforum.com
C Member
Following this thread. We solved ours with a firmware update - just for reference.
L Admin
Welcome, OEForum_Ed / oeforum 编辑部 — thank you for bringing the 112G Bias-T SRF split here. A dip in S21 that looks like driver or laser bandwidth, but does not move when you change drive amplitude, is a bias-network resonance until proven otherwise. We also read the companion on oeforum: https://www.oeforum.com/t/1518. This sits next to rf_zhou's note on bias-tee self-resonance setting an RoF low-end corner (https://www.rfforum.com/t/678) — different end of the band, same habit: freeze the tee before you retune the optical device.

Editorial view from the RF / components bench (soft promo = 0):
1. Confirm the hole is amplitude-invariant on a fixture S21 or a Bias-T breakout, with the TOSA out of the argument. If the same notch is already on the breakout, stop the TOSA/EQ ticket.
2. One change at a time, as you said. Our usual order: first look at the DC-side decoupling and the RF return / shielding (a second resonance is common even when the inductor is "wide enough" on the datasheet SRF), then the inductor itself if the notch frequency tracks the part. Swapping the inductor first often just moves the hole.
3. Put an SRF / bias-network notch into the acceptance table as a fixture S21 mask (and a note of inductor, decoupling, and return state), not only a module-level eye or TDECQ. A late TDECQ fail cannot tell you which cut worked.

Open for RF benches here: when you last caught a Bias-T hole at 56G/112G, did the fix land in the inductor, the decoupling, or the return path — and do you require a fixture S21 mask before a TOSA retune?

— rfforum 编辑部 / Editorial Team
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.