Define the role before comparing gain.
An erbium-doped fiber amplifier increases optical signal power in its specified wavelength band. It does not retime data, remove accumulated noise or regenerate the waveform. C-band and L-band devices must be matched to the intended channel plan. Optical amplifier fundamentals ↗
| Role | Position | First engineering question |
|---|---|---|
| Booster | Transmit terminal, typically after multiplexing | Can it produce the required per-channel launch power at the planned loading? |
| Preamplifier | Receive terminal, typically ahead of demultiplexing | What noise and receiver operating point result from the weakest valid input? |
| Inline amplifier | Between fiber spans | What happens to power, noise and spectral tilt across the complete cascade? |
Those roles can share a chassis without sharing an operating range. Cisco’s NCS 1001 documentation, for example, specifies separate preamplifier and booster paths with different gain ranges. Terminal amplifier example ↗
Read the envelope, not the largest number.
For each candidate, collect wavelength coverage, input-power range, usable gain range, output-power ceiling, noise figure versus gain, gain flatness or tilt, and the measurement conditions. Identify whether power limits refer to a channel, the signal sum, or a monitored total that includes amplified spontaneous emission (ASE).
Maximum gain may not include the same spectral control as the normal range. The NCS 1014 EDFA2 specification distinguishes controlled-tilt and uncontrolled-tilt gain limits. That distinction matters when edge channels already have less margin than center channels. Current line-system specification example ↗
Build a loss record at defined reference planes. “Fiber loss” should not silently include a mux at one end but exclude it at the other. Record connectors, splices, mux/demux paths, filters, monitor taps and planned allowances separately. Determine the minimum and maximum input power that the amplifier will actually see.
Worked example: +20 dBm is not +20 dBm per channel.
For N equal-power channels, ignoring ASE in this first-pass calculation:
Ptotal = Pchannel + 10 log10(N)
| Signal loading | Aggregate signal power | Power per channel |
|---|---|---|
| 32 equal channels | +20 dBm | About +5.0 dBm |
| 64 equal channels | +20 dBm | About +1.9 dBm |
| 32 channels at 0 dBm each | About +15.1 dBm | 0 dBm by assumption |
These are accounting examples, not recommended launch powers. A +20 dBm ceiling is capacity, not a command to operate there. Allocate margin for real loading, gain tilt, noise and the manufacturer’s specified operating conditions. Unequal channel powers must be summed in linear units, not averaged in dBm.
Likewise, an illustrative −10 dBm total input with 25 dB gain implies +15 dBm total signal output before considering saturation and other limits. A headline “25 dB amplifier” is unsuitable if the required input or resulting output is outside its documented envelope.
Gain restores power; it does not restore OSNR.
EDFAs add ASE. Cascading stages therefore requires a noise assessment as well as a power budget. Check the receiver’s required optical signal-to-noise ratio for the exact coherent mode and normalize measurement bandwidth before comparing values. Adequate received dBm alone does not establish usable signal quality. Noise and cascade behavior ↗
Evaluate the control mode with a changing channel count. Holding aggregate power constant while dropping half of a set of equal channels would, absent other controls, allocate about 3 dB more power to each survivor. Holding gain constant has a different objective. Actual devices may combine gain control, equalization, transient suppression and protection; record what the implementation does rather than inferring it from an AGC or APC label.
Check commissioning at sparse loading, planned full loading and after channel add/drop. Include input loss, restoration and restart behavior. Cisco documents gain flattening and transient behavior as explicit amplifier characteristics; they are part of selection, not an optional operational afterthought. Control and amplifier characteristics ↗
Make the selection reviewable.
- Optical path: exact endpoint modes, channel frequencies, passbands, measured losses and reference planes.
- Operating points: minimum and maximum loading, gain targets, per-channel launch and receiver limits.
- Quality margin: noise contribution, OSNR requirements, spectral tilt and other modeled line impairments.
- Operations: control modes, channel monitoring, alarms, restoration and approved laser-safety procedures.
- Validation: line-system assessment followed by measured channel power and receiver performance under the intended loading.
Choose the amplifier that meets these cases with an understood margin. A larger gain or output-power number is not a substitute for that comparison.
Sources and applicability
Original WBN selection framework. Manufacturer sources checked 2 November 2026. Worked power values are illustrative arithmetic, not amplifier settings or a qualified line design. Named line systems demonstrate specification fields, not a WBN offering.
-
Cisco DWDM Engineering and Planning Guide — optical amplifiers
Release 7.x, updated 20 March 2015. Used for stable EDFA principles, not its historical reach or product recommendations.
-
Cisco NCS 1001 — Configure Optical Modules
IOS XR 6.3.x. Separate booster/preamplifier paths and specified operating ranges.
-
Cisco NCS 1014 EDFA2 Line Card Data Sheet
13 July 2026. Controlled versus uncontrolled tilt limits, input/output ranges and monitoring.
-
Cisco ONS 15216 EDFA3 — Introduction
Release 1.0. Reference for gain flattening, noise and amplifier control characteristics; no current product recommendation.
