Clarifying misperceptions about Power over Ethernet and cable losses
Device manufacturers should not be discouraged from tapping into the PoE market based on inaccurate assumptions about power loss over twisted-pair cables.
By David Tremblay and Lennart Yseboodt, Ethernet Alliance
It’s easy to see why so many application spaces are taking a look at Power over Ethernet (PoE)—by reducing requirements for separate electrical power cabling, simplifying network installation and enhancing energy management, PoE promises tremendous cost, efficiency and flexibility benefits.
But aren’t the conductors in network cable much thinner than those in well-known appliance cords or AC mains cable? And isn’t the amount of copper in network cable much less than in AC mains cable? It follows, then, that power losses in PoE must be high, right?
Wrong. Misperceptions about the cable losses in PoE are rampant, and the often-misinterpreted maximum-allowed loss numbers from the IEEE 802.3-2015 standard contribute to the misunderstanding. Correctly assessing cable losses in PoE demands a better understanding of the mechanics of PoE and the language of the IEEE 802.3 standard. Power losses in PoE are significantly lower than perceived.
How much copper?
How much copper is there really in network cables, and how much of it conducts electricity? These are relevant questions because the amount of copper in a cable is related to the resistance of the cable and, hence, the power that will be lost in the cable due to Ohm’s Law.
While the individual conductors in a network cable give the impression of very little copper, there is actually a substantial amount in network cables.
A network cable has eight individual conductors, forming four twisted pairs. IEEE 802.3 PoE proposes the transfer of power using two of the four available pairs in a network cable and allows up to 22.5W to be delivered to a load. A new version of the specification is being developed (IEEE P802.3bt), with an expected release date early 2018, and this new version will define operation over all four pairs of the network cable. This will mean that 100 percent of the copper in the network cable is used for power transfer in PoE. In the AC mains cable, 66 percent of the copper is used for power transfer.
The image above compares the cross-sectional area of typical mains wiring and typical network wires at various gauges. To be able to accurately compare the equivalent amount of copper in a network cable, the area of four conductors would need to be combined and displayed as if they were a single conductor at the same scale. Four conductors of a 24-AWG network cable are equivalent to a 1mm2 copper conductor; 22 AWG is equivalent to 1.3mm2.
Potential vs. actual losses
At the interoperability boundary conditions supported by the IEEE 802.3 standard, relative cable losses of 15 percent seem to be the norm. Operation at 90W even sees a potential cable loss of 20 percent. However, these numbers represent only the extreme conditions in which interoperability and operation are guaranteed by the standard. The cable standards on which IEEE 802.3 builds specify a maximum DC resistance of 12.5 Ω loop resistance for any cable type.
Actual cable resistance is substantially less than this, resulting in much lower actual losses than the worst-case possible. Actual losses in cables are influenced by the DC resistance of the cable, the length of the cable, the voltage of the power sourcing equipment (PSE) and the required power of the powered device (PD). The majority of PDs draw a constant amount of power to be consumed. If the source-side voltage of the PSE is higher, the required current is lower, which, in turn, affects cable power losses.
The performance of complete PoE systems, which consist of many PSEs and PDs and the cables between them, is determined by the total cable losses. Total cable losses are the sum of the power dissipated in each cable, relative to the total amount of power that is being sourced. In systems where there are many different cable lengths, therefore, the performance of the system is much better than the performance of the longest cable in such a system. For example, in a PoE-powered LED lighting system with short cables, an aggregated cable loss of 0.5 percent was calculated in a study recorded in a June 2017 Ethernet Alliance white paper. Even in a very large-scale PoE LED lighting system—with 650 high-power PDs connected to a single location—aggregated cable losses of only about 2 percent werecalculated (in comparison to a 7-percent loss for the worst-case cable within such a system).
PoE is perceived as a system, which comes with unavoidably high cable losses. This misperception is rooted in a misassumption that network cables have an insignificant amount of copper and a correspondingly high DC resistance, and compounded by a common misinterpretation of the corner-case operating points supported by IEEE 802.3. While the standard guarantees operation even with high-resistance cable, this should not be assumed to reflect typical performance.
The truth is that there is substantial copper in network cables, due to the multiple conductors, resulting in PoE power losses lower than one would expect. Additionally, the power lost in cables of a PoE system is far lower than the loss in the worst case in a system. The maximum potential loss numbers allowed by the IEEE 802.3 standard do not reflect the actual system performance, which is much better.
PoE’s application horizon is expanding with the proliferation of home automation, LED lighting and the Internet of Things, and innovation in the IEEE 802.3 Ethernet standards family. Device manufacturers mustn’t let misperceptions about cable losses discourage them from tapping into the breakthrough cost, efficiency, and flexibility benefits that PoE promises for their interconnected devices.
David Tremblay is technical chair of the Ethernet Alliance’s PoE Subcommittee. Lennart Yseboodt is an Ethernet Alliance member. Tremblay is a system architect at Hewlett Packard Enterprise. Yseboodt is a senior scientist at Philips Lighting.