IEEE 802.3bt: Not all detection schemes are equal
The next IEEE 802.3 Plenary Session, to be held in San Diego, California, is only a few weeks away. During the meeting, I will be presenting on an important topic that has yet to be investigated thoroughly: Can you safely apply power to equipment connected to a PoE-enabled switch when performing detection on all four pairs in parallel?
In order to answer that question, let’s look at some of the consequences of detecting PDs on all four pairs (also referred to as both alternatives) in parallel.
CONSEQUENCE 1: The PSE cannot be sure what it is connected to.
The PSE will not be able to differentiate between a single PD connected to both alternatives (through a double diode bridge) and a PD connected to one alternative with a high impendence on the other alternative. This results in the PSE applying power to open circuits and isolated terminations. Second, parallel combinations of two invalid impedances (one on each alternative) can appear as a valid impedance when detected in parallel. For example, a PSE performing parallel detection will see a valid signature if a 50 KΩ impedance is connected to each alternative. Furthermore, the current balance between the two alternatives will be equal, the same as a when a single PD is connected with double diode bridges.
CONSEQUENCE 2: The detection accept and reject criteria in the IEEE standard must be modified in order to maintain backwards compatibility.
This is because one compliant implementation of existing Type 1 and Type 2 PDs is a “dual PD” structure where once one alternative is powered, the PD on the other alternative presents an invalid detection signature. If a PSE detects this type of PD on both alternatives, the two 25KΩ impedances in parallel will appear as a 12.5 KΩ. This necessitates an additional detection acceptance range around 12.5 KΩ (see figure).
Adding the new acceptance range introduces a number of problems as well.
When the original PoE standard was written, a tremendous amount of work went into defining a detection range that minimized the risk of powering equipment that was not designed to take the power. In addition, the current PoE standard allows both PSEs and PDs in the mark state to present signatures as high as 12 KΩ. This means that when detection is performed on both alternatives in parallel, PSEs may be detected as valid PDs and power may be applied. As PSEs are only required to be able to withstand 5 mA of back-driven current, the PSE receiving the power will most likely be damaged.
Adding a detection acceptance range that is the same as PDs in the mark state creates another type of problem. In this case, the mutual identification protocol that is part of PoE can be thrown off and PDs will not determine the amount of power they are receiving properly. Here is a presentation that covered this topic during the IEEE 802.3at task force.
With all of these problems, it is probably clear that I do not believe detecting all 4 pairs in parallel is sufficient to safely power network equipment. What do you think? Are there problems I didn’t discuss? Are there solutions to these problems that would make parallel detection safer?
This post originally published on TI’s Power House blog.