The Noise Bandwidth Behind an Optical OSNR Number

Identify the signal and the noise being compared

For the basic signal/ASE comparison used here, call the isolated signal power for one optical channel Psignal and the noise power referred to the declared bandwidth Pnoise,ref. With both powers in the same linear unit, OSNR in dB = 10 log10(Psignal / Pnoise,ref). The denominator's bandwidth is part of the quantity, even when a summary omits it.

The signal power and the reference noise power have different spectral scopes. The numerator represents the defined channel signal, rather than only the signal within a reference-width slice. The denominator represents noise on the stated reference basis. Neither is automatically the total light arriving on a fiber carrying several wavelengths. Retain the channel's wavelength or frequency and the observation location when interpreting the ratio.

References: [IT1]

Why raising optical power does not remove amplifier noise

EXFO's August 22, 2012 explanation describes an erbium-doped fiber amplifier (EDFA) amplifying both signal and existing noise while adding its own amplified spontaneous emission. ASE originates in spontaneous emission that is subsequently amplified; it adds optical power without supplying the intended data signal.

In a simplified comparison where signal and existing noise receive equal gain, multiplying both powers by the same factor leaves their ratio unchanged. Adding noise then increases the denominator and lowers that ratio. This explains how an amplifier can raise signal power while degrading OSNR. A proposal's promise of more received power therefore leaves a separate noise-quality question to answer.

References: [OS1]

One milliwatt of signal can accompany two different ratios

Consider two invented, isolated component-power cases at a defined observation point. Each has Psignal = 1 mW, equivalent to 1,000 µW. Stipulate that the ASE noise powers have already been determined and normalized to the same chosen 0.1 nm reference band under the same definition. Case A has 10 µW of reference-normalized noise; Case B has 100 µW.

For A, the linear ratio is 1,000 / 10 = 100:1. Its OSNR is 10 log10(100) = 20 dB. For B, the ratio is 1,000 / 100 = 10:1, giving 10 log10(10) = 10 dB.

The signal powers are identical, but B's denominator is ten times larger, leaving its OSNR 10 dB lower. This is the information a signal-power entry alone cannot supply.

These are stipulated component powers, not a measured spectrum, typical link values or receiver acceptance limits. In particular, the 1 mW numerator is isolated signal power, not total received power from digital optical monitoring (DOM). The reference-normalized noise entry also does not describe all noise across an unspecified receiver passband. No total-power subtraction or method for extracting either component is established by this arithmetic.

References: [IT1]

The reference band describes how much noise the ratio represents

A bandwidth stated in nm is a wavelength interval. It is not the data rate or the spacing between optical channels. EXFO's Application Note 098 separates the measurement bandwidth from the reference bandwidth used to report OSNR, commonly 0.1 nm. A result can be normalized to that reference even when the measurement uses a different bandwidth. The reference label alone supplies no instrument setting.

An original qualitative comparison shows why the distinction matters. Hold the isolated signal power and spectral location fixed. Assume strictly positive, uniform noise power per nm across two nested reference intervals, with no filtering or other spectral shaping. The wider interval then represents more noise, so the corresponding signal-to-noise ratio is lower. The narrower interval represents less noise and gives a higher ratio. Changing the reference has changed the reported comparison without improving the optical signal.

That reasoning depends on the stated spectral assumptions. It supplies no numerical correction for an actual shaped spectrum and no universal conversion between nm and GHz. Compare declared reference bands and applicable normalization methods before ranking two quoted dB values.

References: [RB1]

Noise beneath a coherent signal needs an applicable method

VIAVI's June 10, 2020 discussion explains that a modulated coherent carrier can fill the available DWDM filter channel, leaving no exposed noise-only region within it on a traditional optical spectrum analyzer (OSA). Recovering a noise estimate is consequently a method question. ITU-T's September 2012 explanatory supplement also illustrates how filtering can make noise between channels unrepresentative of noise at the signal wavelength.

A reading beside a channel therefore does not automatically establish its in-band noise. Nor does a single total-power value separate overlapping signal and noise. If an offer relies on a module's estimated OSNR, the equipment owner needs its documented definition, reference basis and applicable conditions. A field labelled SNR or OSNR does not establish that it matches a separately specified optical measurement.

References: [OS2][IT1]

Use OSNR to compare the defined optical requirement

Separate optical noise from receiver performance. IEC's public abstract for IEC 61280-2-9:2009 states that its OSA-based noise measurement excludes source relative intensity noise and receiver noise. An optical OSNR is therefore not interchangeable with an electrical SNR after detection. Pre-FEC BER describes errors before forward error correction; it is a different quantity again.

The March 10, 2020 OIF 400ZR agreement illustrates a scoped receiver requirement: section 13.3.1 defines OSNR tolerance against the application's maximum BER under stated conditions and treats additional optical-path penalties separately. That historical definition is useful context, rather than a generic threshold for another interface. An OSNR figure cannot alone establish a universal BER curve or guaranteed reach.

For an actual proposal, the useful comparison pairs the identified channel and observation location with its signal/noise definition, reference band and applicable measurement or estimation method. The appointed optical and equipment owners then evaluate that evidence against the selected interface's requirements. If those definitions are missing, a higher quoted dB value cannot settle the choice.

The related power-versus-loss, average-power-versus-OMA and coherent-package guides answer the adjacent questions. Fiber attenuation, waveform amplitude and the supplied transport equipment remain distinct from the optical noise ratio.

References: [E1][O1][OS2]

Bring the remaining physical question to the right inquiry

When the optical review leaves an accessible installation question, Leviathan's Physical-Layer Testing & Installation Verification service is relevant to agreed cable identification, labels, physical-connectivity checks, defect records and handoff documentation. Use the service and project-inquiry links below with a nonconfidential outline of the path and the physical evidence needed.

Specialist OSNR estimation and optical-system qualification remain with the appointed optical and equipment owners. Any instrument work requires confirmation of equipment, qualified personnel, method, coverage, criteria and reporting for the engagement. This article explains proposal interpretation; it provides no field procedure or approval to alter a link.