Table of Contents
Power-over-Ethernet design errors often begin with one ambiguous number: “available watts.” Power supplied by the source is not the power available to the final load after the cable, input stage, and converter.
For a powered device (PD), select the controller around those boundaries and the system’s behavior when the source grants less power than desired. Classification, startup, and low-load operation are part of the requirement, not details to leave until the first network test.
Power boundaries and the load budget
Power sourcing equipment (PSE) supplies the cable. The PD receives power at its input interface, and the downstream converter supplies the application load.
The TPS2373 datasheet, reviewed September 15, 2026, describes IEEE 802.3bt PD behavior and distinguishes source-side and PD-side power allowances. For example, its table lists 60 W at the PSE and 51 W at the PD for a relevant four-pair Type 3 case, and 90 W versus 71.3 W for Type 4. These are defined interface figures, not promised application output power.
Use the power class and operating conditions actually supported by the selected design. A product-category label such as “PoE++” is insufficient to define the budget.
A downstream power calculation
Assume a hypothetical load needs 40 W, the downstream conversion efficiency at that operating point is 90%, and 1 W is consumed by circuitry outside that conversion path.
The required power at the chosen input boundary is:
40 W / 0.90 + 1 W = 45.44 W.
If 51 W is available at the PD interface under the applicable conditions, the remaining allowance is about 5.56 W. Assign that allowance to input losses, operating variation, and design margin that have not already been included.
Do not subtract cable loss again if starting from a PD-side power figure that already accounts for the standardized cable budget. Conversely, do not use a PSE-side number as if it were measured at the converter input.
Evaluate peak and startup demands separately from average consumption. A processor boot sequence, radio transmission, or motor event can violate the available envelope even when average power appears acceptable.
Controller integration and external components
| Function | Question for the component shortlist |
|---|---|
| Detection and classification | Which source types and requested power classes are supported? |
| Input protection and hot-swap path | What is integrated, and what external parts remain? |
| DC/DC conversion | Is a converter controller included, or is a separate stage needed? |
| Power indication | How does system firmware learn the granted power? |
| Auxiliary input | How are adapter power and PoE prioritized and isolated? |
| Low-load operation | How is the required maintain-power behavior handled? |
A PD interface IC is not automatically the complete power supply. Identify the bridge, protection network, magnetics, converter, isolation requirements, and thermal path from the actual design.
The Ethernet controller discussion concerns the data function. PoE power reception does not replace the MAC or PHY.
Startup and reduced-power operation
Before releasing the BOM, write the intended response to each source capability. Specify which functions start, which are delayed, and which remain disabled when full power is not available.
A useful acceptance matrix includes supported PSE types, two-pair and four-pair cases where applicable, cable resistance, low input voltage, startup load, and repeated disconnect/reconnect events. Include an auxiliary adapter if the product has one.
Test the transition between reduced and full operation rather than only steady-state output voltage. Firmware should not enable a high-power load before the design has established that sufficient power is available.
This is a system verification plan, not a declaration that selecting a controller establishes Ethernet or safety compliance.
Substituting a part in the complete power stage
A candidate with the same package and power-class label may differ in classification signaling, gate drive, startup current, timing, power-good behavior, or supported external components. Review the exact ordering code and suffix in the current manufacturer documentation.
For an IoT hardware program, retain the source compatibility matrix and available-power policy with the schematic and firmware. For an alternative-component evaluation, repeat the operating cases affected by the change.
The procurement specification should end with a usable system envelope: input conditions, approved source capabilities, output loads, permitted degraded modes, and verification evidence.
Frequently Asked Questions (FAQ)
Does a 90 W PoE source provide 90 W to the application load?
No. Source power, power available at the powered-device interface, and useful converted output power are different quantities. Account for cable, input-stage, conversion, and auxiliary losses at the correct boundaries.
Is a PoE PD controller an Ethernet PHY?
No. A PD controller manages power reception and related functions. The Ethernet data path still requires the appropriate MAC, PHY, magnetics, and supporting circuitry.
Can a higher-power PD operate from a lower-power source?
Only if its design recognizes the available power and restricts operation accordingly. Do not assume full functionality or automatic safe fallback without validating the controller and system behavior.
