Rx power and the link budget

Receive power is the one DDM value that is almost always real, and it is the end of a simple equation: what the far end launched minus what the path lost. Read both ends and you have the measured loss of the link; compare it with what the plant should lose and you know whether the fibre, the connectors or the module is the problem — and how much margin is left before the link fails.

The equation

Rx(here) = Tx(far end) − loss(path)
loss(path) = fibre attenuation × length + Σ connectors + Σ splices + Σ passive components
margin     = Rx(here) − sensitivity(this receiver)

Both Tx and Rx come from DDM (Parameters); the loss elements come from the plant record or from these typical figures:

Detailed budget method for CWDM: Mux/demux & link budget; unit conversions: Units & conversions.

Worked example: 10GBASE-LR, 12 km

• Far-end Tx (DDM): −2.5 dBm
• Expected loss: 12 km × 0.35 = 4.2 dB + 4 connector pairs × 0.4 = 1.6 dB + 6 splices × 0.1 = 0.6 dB → 6.4 dB
• Expected Rx: −2.5 − 6.4 = −8.9 dBm
• Sensitivity (10GBASE-LR, average): about −14.4 dBm → expected margin ≈ 5.5 dB

Both directions

Measure A→B and B→A. Fibre loss is symmetric to within a few tenths of a dB; a difference larger than the DDM accuracy (±3 dB, in practice > 2 dB) means the asymmetry is at an end, not in the path:

Per-lane on parallel optics: compare the four or eight Rx values with each other before comparing with a specification (Per-lane diagnostics).

Overload: too much light

Every receiver has a maximum input; above it the photodiode or TIA saturates and errors appear even though "the signal is strong". Typical overload points: −1 … 0 dBm for 10G SR/LR PIN receivers, −3 dBm for 1G, −7 … −9 dBm for APD receivers in ZR/DWDM modules, +2 … +4 dBm for PAM4. A long-reach module on a short link needs an attenuator: size it so that Rx lands in the middle of the window (Typical values, Receivers).

Power is not everything: dispersion and penalties

IEEE budgets reserve several dB for penalties — dispersion, reflections, jitter — that do not show up as lost power. For 10GBASE-LR the budget is 9.4 dB of which only 6.2 dB is channel insertion loss; the rest is penalty allocation. So a link can show "3 dB of margin" by Rx power and still be at its limit on a long span at 1550 nm, on CWDM 1610 nm, or on OM1 fibre at 10G. Symptoms are errors and flaps with good levels (Link flapping); the cure is a dispersion-tolerant module (EML, shorter reach, lower rate) rather than more power.

Margin policy

Design new links for at least 3 dB of margin after ageing (0.5–1 dB for laser ageing over life, 0.5 dB per future repair splice).

Recording

Record Tx and Rx at both ends at installation — the link passport (Monitoring). Later readings are then deltas, which are immune to the ±3 dB absolute accuracy of DDM (Accuracy & limits).

In CodingBox

The DDM screen shows Tx and Rx with the module's own thresholds and logs them over time; reading both modules of a link on the bench before installation gives the launch powers for the budget above. For a module pulled from a failed link, the logged Rx history shows whether the loss grew slowly (plant) or the module's Tx fell (Tx bias & laser ageing).

Budgets with splitters and class windows: ODN classes & PON budget. Amplified links where noise, not power, is the limit: Amplification & OSNR.

Design-side detail — allowances by standard, penalties, margin policy and four worked budgets: Link budget engineering; measuring the plant loss with LSPM and OTDR instead of inferring it from DDM: Testing and measurement.