In order for any industry standard to reach wide adoption, two required characteristics are: interoperability and safety.
In the case of Power over Ethernet (PoE), the technology’s reputation for both, heavily rests on how the source (Power Sourcing Equipment or PSE controller) detects an IEEE802.3 compliant load (Powered Device or PD controller). Or, more accurately, rejects an invalid one.
The first step in attempting to power on an active load is the proper detection of a 19-26.5KΩ resistor on the PD side of the CAT-5 cable. Per the current IEEE specification (Clause 33 of 802.3at), the PSE must properly detect a resistor within that range and may choose to accept (or reject) anything in the ranges of 15-19KΩ or 26.5-33KΩ. Anything <15KΩ or >33KΩ is to be rejected and not turned on. It is critical that detection occur reliably and consistently in order to prevent excessive voltage from being delivered which could possibly result in damage to port-level components.
While the standard defines what must be done in order to pass this first step, it does not mandate exactly how.
As a result, the detection function of PSE controllers varies widely from one IC to another. The biggest differences are: (1) how many times the PSE controller attempts to discover the load’s resistance, (2) to what accuracy, and (3) within what time duration.
What is frightening about this situation is that some implementations can result in “fooling” industry-accepted PoE Test Suites into thinking a PSE controller is safe when, in fact, under the right wrong conditions it would attempt to classify (and perhaps even power) an invalid load.
One of these conditions is the so-called “plug/unplug/plug” (or shaky hand) test. During this test, at low and invalid resistance levels in the range of 11-15KΩ, some well-known PSE controllers will incorrectly conclude that a valid PD load exists and move to the classification step. An example of this behavior is captured in Graphic 1 below.
Graphic 1: Improper Detection of a low-resistance load using 2-point detection
In this example:
- An invalid RDETECT value of 11.5KΩ was present on the load,
- The load was plugged, unplugged and plugged back in within acceptable IEEE802.3at time frames, and;
- 2-point detection methodology was implemented.
Avoiding this condition, and others like it, is critical to the future success and adoption of PoE. To this end, TI is taking a 3-pronged approach toward driving proper detection (and power classification) of PoE loads.
- Participation in the IEEE802.3bt ( 4PPoE, four-pair PoE, >51W PoE) committee with a particular focus on challenges associated with proper detection of four-pair systems. You can read our latest proposals on this topic here.
- Supporting work to create a formal, and more stringent, PoE logo and certification program. You can read more on those exciting developments here.
- Releasing PSE controllers, like the TPS23861, which meet not only the “letter of the law” but also the spirit of it – with innovations like four-point detection (AKA,”never fooled detection”) which prevent accidently powering up an invalid load.
Graphic 2: Four-Point Detection feature of the TPS23861
As you can see in comparing graphic 1 to graphic 3, the old adage “measure twice, cut once” doesn’t quite work in the world of PoE. Here, you need to “measure four times, power once.”
- “IEEE 802.3bt: Not all Detection Schemes are Equal” Blog
- TPS23861 Product Folder
- “Detecting PDs Over All 4 Pairs in Parallel: An Analysis” – Presentation from 802.3bt Plenary Meeting, July 2014
- “IEEE802.3: Compliance Revisited—Two Exciting Announcements” Blog