Noise needs a bandwidth definition
Section 9.5.1 of the March 2025 ITU-T G Supplement 39 defines OSNR using channel signal power and amplified spontaneous emission noise. It distinguishes the noise-equivalent measurement bandwidth from the reference bandwidth, commonly 0.1 nm. Bandwidth units used for normalization must be consistent. 1
A report stating only “OSNR 25 dB” therefore lacks comparison conditions. Request the reference bandwidth, measurement location, channel identification and algorithm. Preserve the optical spectrum analyzer's resolution setting separately. Ciena's published interview with EXFO also explains why measurement methods must suit the signal type. A displayed 0.1 nm value alone does not explain the complete normalization. 3
This guide addresses amplified, WDM and coherent-link documentation. A short passive MPO cord does not need an undefined OSNR result added to its ordinary loss report simply to increase the number of test fields.
Separate four quantities
| Result | Typical unit | Object described | Does not directly establish |
|---|---|---|---|
| Received optical power | dBm | Absolute power at a location | Signal quality relative to noise |
| Insertion loss | dB | Power change under a defined reference | An active system's OSNR limit |
| OSNR | dB with reference bandwidth | Signal relative to defined noise power | Universal BER or FEC acceptance |
| OSA resolution bandwidth | nm or frequency units | Spectrum-analyzer measurement setting | Identity with OSNR reference bandwidth |
In a hypothetical calculation, assume channel signal power is minus 3 dBm and the applicable noise, already normalized to 0.1 nm, is minus 28 dBm. Their difference is 25 dB. This is arithmetic only, not a measured result, an ARRANEX specification or an acceptance limit for any transmission rate.
When reference bandwidths differ, check the normalization convention before comparing results. Do not recalculate acceptance without knowing the noise-spectrum assumptions and instrument definitions.
Why between-channel noise may not represent in-band noise
Interpolation estimates noise under a signal from neighboring spectral positions. Filtering can make those positions unrepresentative of noise within the channel. The ITU document warns of this limitation, while EXFO AN350 discusses in-service OSNR measurement for polarization-multiplexed signals. 12
An instrument's OSNR feature is not evidence that one method applies to every modulation, spacing or filtered route. If a method or instrument changes, obtain approval for the comparison plan and retain its applicability statement. Otherwise a method difference can be mistaken for a network change.
Define an acceptance record
| Field | Information to preserve | Review purpose |
|---|---|---|
| Measurement location | Port position relative to amplifiers, filters and receiver | Compare the same optical location |
| Channel | Wavelength or frequency, width, neighbors and signal type | Avoid a wrong peak or adjacent signal |
| Noise method | Interpolation or applicable in-band algorithm | Explain result applicability |
| Instrument | Model, software, resolution and reference bandwidth | Enable repeatable review |
| System acceptance | Equipment requirement, service state and separate link tests | Avoid an invented universal OSNR limit |
These are engineering handover recommendations. Agree whether service interruption is permitted before testing. Any method requiring a change to transmitted light or channel state needs an approved procedure for an operational network.
Retain spectra and settings, not just a pass screenshot. Passive-component loss and connection records cover their agreed scope; they do not become a coherent-system OSNR guarantee.
Procurement FAQ
- Does normal received power guarantee normal OSNR? No. Review absolute power and signal-to-noise quality against their separate requirements.
- Must every OSA use 0.1 nm resolution? No. OSNR reference bandwidth and measurement resolution are separate fields governed by the method and instrument.
- Can interpolation be used after a ROADM? Check filtering and method applicability rather than assuming neighboring noise represents in-band noise.
- What OSNR does 800G require? A rate label is insufficient. Obtain the modulation, module and complete system acceptance conditions.
- Is one screenshot enough? Preserve the channel, location, algorithm, bandwidth settings and original spectrum so the result can be reviewed.
Related products and reading
- Test reference assemblies
- Fiber connectivity products
- WDM total, peak and channel power
- Insertion and return loss
- 400ZR and DR4 cabling boundaries
These product pages describe existing assembly scopes. Confirm interfaces, applicability and validation against the actual engineering requirements.
Contact an engineer
Prepare interface drawings, optical-path conditions or existing test records and contact an engineer about assembly configuration and validation scope.
Sources and review date
- 1. ITU-T G Supplement 39, March 2025, section 9.5
- 2. EXFO in-service polarization-multiplexed OSNR note AN350
- 3. Ciena and EXFO OSNR technical interview, 2018
Sources checked on 2026-10-26. Third-party references explain general technology or the named manufacturer's configuration, not ARRANEX performance, certification or supply. Procurement checklists are editorial recommendations; hypothetical examples are not test results.
Sources checked: 2026-10-26