Separate detection from the request for power
Detection asks whether the cable presents the electrical signature associated with a powered device. In TI's historical TPS2373 account, the DEN function participates in that initial signature while most internal circuitry remains inactive. This is an identification stage in the power relationship. It is different from supplying the operating load and different from communicating the amount of power the device requests.
The source also explains why the signature seen at the power interface is not merely one isolated component value. The input bridge and internal loading contribute to the detection relationship. Preserving the complete detection boundary prevents a controller label or a detached resistor description from being treated as a measured signature at the cable interface. No resistance, threshold or adjustment is selected by this qualitative explanation.
Classification follows another question: how much power is being requested before operation. TI's overview separates idle, detection, classification and operation rather than treating detection as a complete power negotiation. Keep that order of responsibilities visible in the named controller description. Recognizing an initial PD signature does not establish a class, an available allocation or a completed startup result for a board.
Keep requested class with its classification stages
The TPS2373 description assigns different classification stages to CLSA and CLSB. Their associated external resistive paths, together with the controller's own current contributions, create the classification signatures presented to the power-sourcing equipment. The signatures express the powered device's request. They are not observations of how much power a downstream converter or load is presently consuming.
Staging matters because the PSE generates a sequence of classification events and the PD presents the corresponding signature in that sequence. A broad class label can omit this interaction between the request and the events received from the supplying side. Retaining the early-stage and later-stage roles explains why the two classification functions are not interchangeable labels for one unqualified current entry.
The historical document then uses the received sequence to distinguish an agreed allocation from a lower allocation. Thus requesting and receiving are separate statements even within the hardware-classification account. TI also explicitly locates data-link-layer communication in the PD's Ethernet communication system, outside the TPS2373 itself. This scope gives no event-count decoder, resistor choice or class-power arithmetic, and does not assign an installed class from the controller's family name.
Read allocation indicators separately from controller capability
Allocation indicators report another layer of the relationship. TI describes TPH, TPL and BT as conveying information about the PSE type and allocated power. Their subject is the outcome communicated through the classification context, rather than simply the maximum capability advertised for the controller. An output's presence on a product description does not establish what state that output has in a particular assembly.
Demotion makes the distinction concrete. The PSE can allocate less than the PD requests, and the historical account retains that lower-allocation situation separately from the original class request. A high-capability controller may consequently participate in a connection whose allocation is lower. The family heading cannot replace the indication of what the supplying side actually offered in the documented exchange.
The indicators also have startup and supply-state dependencies in TI's prose. Their electrical presentation is not an unconditional statement detached from whether the controller is operating. Preserve the meaning of the indication without reconstructing its truth table, timing or logic levels. It supplies neither a measured output-power result nor evidence that an offered controller, board and downstream load have completed the exchange successfully.
Distinguish MPS from initial detection and classification
A maintain-power signature addresses continued presence after operating voltage has been applied. It is therefore different from the initial detection signature and the pre-power class request. TI's overview describes the PSE monitoring for this continued-presence condition and removing power when the relevant signature is lost. The function belongs to maintaining an existing power relationship rather than establishing the first detection identity.
The TPS2373 account includes automatic pulsed-current generation when the ordinary load path does not provide enough current for that purpose. The controller's MPS behavior responds to the detected supplying-equipment context. This explains why a low-power operating state can still require a separate presence signal: the useful application load and the electrical indication of continued attachment are different responsibilities.
TI also retains system conditions such as cable and bulk-capacitance context when discussing the signature observed at the supplying end. A controller feature therefore does not by itself establish the complete connection's behavior. The November 2018 document explicitly describes a draft-standard context; no present compliance conclusion follows. This explanation selects no pulse amplitude, duration, duty setting or component, and performs no actual power-maintenance assessment.
Sources and further reading
TPS2373 SLUSCD1C, revised November 2018, bounded detection/classification/indicator/MPS prose on physical pages 17–21,24–25 and 27. Its 802.3bt references remain explicitly Draft. Numerical limits, tables, pulse recipes, LLDP implementation and current certification are excluded.