Energy efficiency to ITU-T L.1310: why two bids still cannot be compared

Two datasheets for the same type of equipment are sitting on your desk. One declares efficiency in megabits per second per watt, the other in watts per gigabit per second. The specification you wrote asked for energy efficiency measured in line with ITU-T L.1310, and both datasheets comply. In the evaluation those two figures end up in the same column, and the column decides. Nobody cheated: it is the Recommendation itself that permits both directions.

Two directions, and a third metric of the supplier’s choosing

Recommendation ITU-T L.1310, Energy efficiency metrics and measurement methods for telecommunication equipment, is in its fifth edition, approved by Study Group 5 on 22 September 2024. Forty pages, listed as In force, with four superseded editions behind it. It covers telecommunication network equipment and, in addition, the small networking devices used in homes and small business premises.

Clause 6.1 defines the metric as the ratio between the functional unit and the energy needed to deliver it, then adds a single line: “The inverse metric, energy divided by the functional unit, could be used as an alternative.” This is not a textbook hypothesis: the Recommendation uses both directions itself, watts per port in clause 9 and megabits per second per watt in clause 11.

Clause 6.4, Metrics robustness, opens by asking for precisely what you would expect: values obtained in strict accordance with the ITU-T documentation and the referenced standardisation documents. Then it opens a door. Wherever full application is not possible for technological reasons, “manufacturers can use another metric that should be clearly documented”.

Look at the verbs: can, not shall; should, not shall. Writing nothing more than measured in line with L.1310 grants the supplier, in writing, the freedom to use a different metric, with an invitation — not an obligation — to declare which one it is. And clause 6.2 makes only system-level and equipment-level metrics mandatory: those at component level are offered as suggestions, not as obligations.

The ambient is advised, the voltage is prescribed

Clause 7 sets the general test conditions, and it needs reading with a finger on the verbs. For the environment they are all should: a temperature of 25 ± 3 °C, with the unit powered at that temperature for not less than three hours; humidity between 30% and 75%; pressure between 812 and 1,060 hPa.

For the power feed the verb changes: “The equipment powered by −48 Vdc source shall be evaluated at a DC voltage of −53 V ± 2 V”; for other nominal voltages, ±4% of the specified float voltage. The test is not run at the nominal voltage but five volts above it, so if your specification talks about −48 V power feeding in an exchange, the figure that belongs in the test certificate is the other one.

Clause 7.3 is the most demanding of the forty pages. The meter accuracy is only desired to be no worse than ±1%; then, in a single sentence: “All measurement instruments used should be calibrated according to [ISO/IEC 17025], be within the calibration due date, and the measurement tolerance must be within ±1%”. A should and the only prescriptive must in the body, on the same line. The rest reads like a laboratory: resolution of 0.5% or better, on AC a minimum sampling rate of 40 kHz and a bandwidth of 80 kHz, accurate readings up to a crest factor of at least 5, power factor correction, power sources rated at 1.5 times the equipment, one hundred evenly spaced readings per cycle. None of that is improvised with a clamp meter in the equipment room: it takes a bench and a calibration still within its due date.

The report has no format, and the configuration is chosen by whoever measures

Clause 8, first line: “This Recommendation does not preclude any specific reporting formats.” The minimum list that follows — software versions, hardware board revisions, configurations, traffic generator, actual feed voltage, ambient conditions — is introduced by should include. Two reports, both of them compliant, may have not a single item in common.

And for modular equipment, which is very nearly everything you buy as a shelf: “Metrics for modular telecommunication systems shall be obtained using the most common set-up and configurations.” Here the shall is present. Who gets to decide which set-up is the most common is written nowhere at all: in case of doubt, the seller decides.

Where the requirements sit, and where they send you

Counting across the text gives 41 shall, 17 should, 13 may and 3 must. One shall and one must sit in the cover note that explains what those words mean, and two shall sit in the appendices, which do not form an integral part: the requirements in the body and in the two normative annexes come to 38.

Then you look at where they fall. All seventeen should sit in clauses 6, 7 and 8 — the general part, the only one a specification ever cites; from clause 9 onwards there is not one left. And that part holds eight shall in total: two on modularity, three on voltage, three on instruments. Clause 8 holds zero.

Nine of the thirty-eight, finally, do not prescribe a measurement: they prescribe reading somebody else’s document. For DSLAM, MSAN and GPON OLT equipment the methodology sits in clause 5.2 of ETSI EN 303 215. For wireless, “the following standards shall apply”, subject only to the general conditions of clause 7: ETSI ES 202 706-1, ETSI TS 102 706-2, ETSI TS 103 786. For routers and Ethernet switches it is ATIS-0600015.03.2023, which also sets the a, b, c weighting multipliers in the formula; and for WDM/TDM/OTN transport, ATIS-0600015.02.2016 defines the metric itself. The ATIS documents have to be obtained separately.

The edition it cites moves on its own

That leaves a date. L.1310 references ETSI ES 202 706-1 V1.7.1 (2022). On the ETSI file server, however, V1.8.1 (2024-09) is recorded as deposited on 10 September 2024, twelve days before L.1310 was approved; and since 14 August 2026 the final draft of V1.9.0 has been sitting there too. The other two cited for wireless are still the latest published.

Clause 2 allows for exactly this: it warns that the references are subject to revision and invites users to check for the most recent edition. It is a dynamic reference, and a specification that writes nothing but L.1310 inherits that movement without saying so. The trap repeats one line further on: ITU-T L.1304 on sustainable procurement requires declared efficiency to be measured in line with L.1310 — and cites the 2020 edition.

What to add, line by line

For two figures to be comparable you need six items, and they fit on half a page: the metric and its direction, with the unit spelled out; the declared configuration, reported next to the figure; the actual test conditions, feed voltage included; the edition of each referenced document, with its year — L.1310 (09/2024) plus whichever ETSI or ATIS documents apply to your equipment; the report format; and the laboratory, with a calibration still in date. The first three can be copied straight from the text. The other three cannot.

The point

A consumption figure declared in line with L.1310 is not a commitment: it is a referral. In the specifications and compliance work we deliver, that line arrives complete — metric and direction, configuration, conditions, editions with their years, format and laboratory — and it becomes the running order of the check carried out at acceptance, with measurements and documentation handed over together with the work, not once the tender has closed. On the other axis, consumption is not an acceptance figure: it shifts with load, with the season and with the configuration a unit carries after three years of expansion. Meter readings, per-shelf current draw, traffic carried, temperatures and as-built records of the equipment in the field come together in a single map of the network, where an AI recognises the unit that has stopped being efficient before the electricity bill does, and a team of ours closes the job without subcontracting. That AI runs inside the client’s perimeter, on-premises or in a dedicated cloud with data centres in Italy, with CSIDIA, the other company in the group.

Are you comparing two bids on declared consumption, or writing that line right now? Talk it through with an engineer: the site visit is free of charge, and the metric is fixed before the tender, not after.