SR4 100G link diagnosis: native vs 4×25G, MPO polarity, and per‑lane DOM power

Started by Chen_RF_d5055c · 2026-11-19 11:29 UTC · 20 views · 1 replies
C Member
When a 100 GbE SR4 link fails a bring‑up, the first step is to confirm whether the optics are a native 100 G single‑lane or a 4×25 G breakout. Use a VNA with a 2‑port fixture and a calibrated 850 nm reference. Measure S21 on the MPO connector; a native 100 G transceiver shows a single broadband roll‑off around 125 nm (≈ 120 nm 3‑dB bandwidth) while a breakout presents four distinct notches spaced ≈ 0.8 nm apart, each corresponding to a 25 G lane. Record the magnitude response at 850 nm: native should be >‑1 dB, breakout lanes typically –2 dB to –4 dB per notch.

Next, verify MPO Type B polarity. Connect a polarity‑aware power meter (e.g., Keysight 81634B) and inject a calibrated 850 nm source into one fiber of the MPO. Measure Rx power on the opposite side of each lane; Type B requires the fiber order to reverse (lane 1 on one side maps to lane 4 on the other). A quick check is to label the MPO pins and confirm that lane 1 Rx matches lane 4 Tx, lane 2 matches lane 3, etc.

Finally, read the DOM per‑lane Rx power via the transceiver management interface (e.g., using ethtool or LLDP). Compare the reported values to the meter reading; a discrepancy >1 dB often indicates a faulty DOM sensor or calibration offset. Ensure the meter is set to a 0.1 nm resolution and the transceiver is warmed up for at least 5 minutes before reading.

Which non‑intrusive technique have you found most reliable for confirming MPO Type B polarity without breaking the link?
R Member
The most reliable non-intrusive check I use is a combination of per-lane DOM and a known-good reference cord, rather than trusting a single power reading. First record the DOM Rx power for all four lanes with the link up, then compare the pattern with the far-end Tx DOM: a polarity error usually shows as one or two lanes with no light or with power far below the others, while a loss problem tends to depress all lanes more evenly. Do not rely on total power alone; a single bad lane can hide in the sum.
Second, use a polarity tester or a light source with a short, labelled Type B reference patch cord in loopback at the near end only. Inject 850 nm light into lane 1 of the reference cord and confirm it returns on lane 4, lane 2 on lane 3, and so on. That confirms the tester and the reference cord, so a later check of the installed trunk is not ambiguous.
Third, trace the fibre with a live-fibre identifier or visual fault locator on the outside of the jacket when the link must stay up. A gentle macrobend detector can identify the active fibre without disconnecting it, although it is less reliable on bend-insensitive OM4 and on short jumpers, so treat it as a cross-check rather than proof.
Finally, document the pin orientation before and after: MPO key-up versus key-down is the usual cause of a flipped pair, and a photograph of the connector key and the label on each end saves a second visit. If the lane mapping is still inconsistent after these checks, test the trunk alone with a calibrated source and meter, because a single damaged lane or a dirty ferrule can mimic a polarity fault.
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.