A third-party transponder on someone else’s DWDM line: what gets measured, and where?

A customer has bought a transponder from one supplier and wants to plug it into an amplified DWDM backbone run by another operator. The question the specification has to settle before the order is placed is not whether they are compatible: it is where you measure, what you measure, and who answers for it when the number does not add up. Three ITU-T Recommendations have answered with the same method for years — the black link approach — and the newest one dates from this May.

The black link approach: the line becomes a black box between two thresholds.

ITU-T G.698.1, Multichannel DWDM applications with single-channel optical interfaces, edition 06/2023, in force, states: “This Recommendation defines and provides values for single-channel optical interface parameters of physical point-to-point and ring DWDM applications […] on single-mode optical fibres through the use of the ‘black link’ approach.” It does not matter how the line itself is built: only the values at the two ends count. But G.698.1 draws a hard line: “Applications containing amplifiers within the black link are outside of the scope of this Recommendation.” For amplified backbones — most metro links today — you need ITU-T G.698.2 (11/2018, in force), which flips that constraint: “The black links covered by this Recommendation may contain optical amplifiers.”

Two reference points, not a declaration of trust.

The mechanism is identical in both Recommendations. SS is the point where the transponder launches the signal into the line; RS is the point where it is received at the far end, and the recommendations define both for point-to-point and ring topologies alike, unidirectional or bidirectional. Everything in between — multiplexer and demultiplexer, OADMs and, with G.698.2, amplifiers — is the black link. G.698.1 spells it out: “This Recommendation specifies parameters in order to enable transverse (i.e., multivendor) compatibility at single-channel reference points SS and RS of the ‘black link’ approach DWDM network elements (NEs)”: two transponders that have never met, at opposite ends of the same black link, exchanging the same application code.

And when the codes do not match? G.698.2 neither promises compatibility nor rules it out: “Coexistence of tributary interfaces with different application codes over the same black link is a matter of joint engineering.” Outside the standard, in other words — something the two parties have to settle before the order and put in writing, because no recommendation will settle it for them afterwards. The same clause isolates one exception, and it runs one way only: on spectral excursion, a transmitter carrying the W indicator on a link marked N is incompatible, while “All other combinations are transversely compatible.”

Why it exists: it replaces the adapter, not your transponder.

Before these interfaces existed, connecting equipment from a supplier other than the line’s meant an extra adapter transponder for every channel, whose only job was to convert the signal into the line’s own format — whether the far end was a cross-connect with several optical interfaces from a different vendor, or several client devices, each from its own vendor, feeding one channel apiece. G.698.2, in its appendix, describes what the new approach buys back: “this interconnection can also be achieved while removing the need for one short reach transmitter and receiver pair per channel (eliminating the transponders) with obvious associated cost savings.” The customer’s transponder does not disappear: what disappears is the box that used to exist only to let it talk to the line.

What gets measured, point by point — and whose number it is.

G.698.2’s parameters fall into three groups, and the split is not just editorial. At point SS: output power, central frequency, side-mode suppression, extinction ratio — properties of the transmitter. Along the path from SS to RS: ripple, residual chromatic dispersion, PMD, reflectance, crosstalk — properties of the black link itself, regardless of which transponder sits at either end. At point RS: input power, minimum OSNR, receiver OSNR tolerance. On OSNR, G.698.2 is explicit about who has to deliver what: “This parameter places a requirement on the characteristics of the black link that the OSNR at any reference point RS must be greater than the minimum OSNR.” The Recommendation assigns no contractual liability — that is for the parties’ own contract — but it fixes a number and a point: if the OSNR measured at RS falls short while the transmitter meets its parameters at SS, the shortfall belongs to the black link, not the transponder. And the other way round. The same logic covers crosstalk: the black link’s isolation has to keep inter-channel and interferometric crosstalk at RS below threshold under the worst-case operating conditions, whichever transponders happen to be plugged in at either end.

Whether there is an amplifier inside is already spelled out in the application code.

Every application code carries a letter dedicated to exactly this. In G.698.1: “D indicating that the black link does not contain any optical amplifiers.” In G.698.2, the same position, a different letter: “A indicating that the black link may contain optical amplifiers.” Before you even open the parameter tables, the code itself tells you what to expect inside the box.

The newest of the three, ITU-T G.698.4 (05/2025, in force), adds a further layer: a bidirectional line over a single fibre, with the tail-end equipment automatically adapting its own frequency to whichever port it is plugged into, over a dedicated message channel back to the head-end. Alongside SS/RS sit two multichannel reference points on the head-end side, MPI-SM and MPI-RM. The Recommendation draws the explicit conclusion: “Thus, TEE, black link and head-end equipment (HEE) suppliers are not necessarily the same.” Three suppliers, four reference points, and the same logic throughout: each one answers for what is measured at its own boundary.

How it is actually measured: two points, one reference receiver.

G.698.2 does not stop at defining the parameters: an appendix describes how to measure them too. “Since the applications in this Recommendation use the ‘black link’ approach, single-channel reference points (SS and RS) can be used to access the individual signals for measuring OSNR penalty.” Three set-ups — added noise with no dispersion, added noise with worst-case residual dispersion, and after actually passing through the black link — with the bandpass filter and reference receiver fixed by ITU-T G.959.1 — clauses B.2 and B.3 respectively — not by whatever instrument happens to be in the lab. The same three set-ups separate two different numbers: how much OSNR the transmitter’s own residual dispersion costs, and how much the actual path through the black link costs on top of it — again, two figures with two different owners. One practical warning: where the noise is not flat ahead of the measurement point — typically at RS, from OADM filtering — an OSNR monitoring technique that gives an accurate reading is required, not the figure read off the nearest spectrum analyser.

What to write into the specification.

The full application code, letter by letter, declared by both parties — not a generic DWDM-compatible.
A measurement report at SS and at RS attached to the acceptance test, not an email confirming compatibility.
Minimum OSNR at RS as the black-link operator’s obligation, with a monitoring technique suited to any OADMs on the path.
The transponder’s OSNR tolerance declared by its supplier and measured against the ITU-T G.959.1 reference receiver, not estimated.
On a bidirectional single-fibre run with self-tuning tail-end equipment, include MPI-SM and MPI-RM on the head-end side too (G.698.4).

The point.

The black link does not solve a trust problem between suppliers: it removes it, replacing it with a number and a point where you read it. That is where we come in when an optical network design has to hold together equipment from different manufacturers on the same backbone: the application code becomes a verifiable clause, the SS-and-RS measurement report a record to produce, not a promise. And as runs multiply across an operator’s network, a data centre campus, an industrial plant or a public body, the same measurements — one per supplier, one per boundary — feed into the same network map, where an AI flags which black link is drifting towards its margin before an event pushes it out of spec. Wherever AI runs on network data, it runs inside the customer’s own perimeter — on-premise or on dedicated cloud with data centres in Italy — alongside CSIDIA, the group’s other company.

Do you need to plug one supplier’s equipment into a DWDM line run by another? Talk to an engineer: the application code and the measurement points get written down before the order, not discovered at commissioning.