Power over Ethernet (PoE) and Smart Buildings: Part II

According to the National Fire Protection Association (NFPA), electrical and lighting equipment is the third largest source of commercial fires in the United States. Typical root causes are old or defective wires, overloaded circuits, loose connections, faulty fuses, unbalanced electrical loads, and many other electrical or lightning strike problems. These can lead to overheating, sparking, and ultimately a fire.

By: Bob Card, US Marketing Manager, Advanced Solutions, ON semiconductor

According to the National Fire Protection Association (NFPA), electrical and lighting equipment is the third largest source of commercial fires in the United States. Typical root causes are old or defective wires, overloaded circuits, loose connections, faulty fuses, unbalanced electrical loads, and many other electrical or lightning strike problems. These can lead to overheating, sparking, and ultimately a fire.

Mains power carries alternating current over short and long distances through three insulated copper wires: live, neutral, and ground. The live wire carries an AC potential difference (120 VAC or 230 VAC). The neutral wire completes the circuit and is held at or near ground potential, or 0V. The ground wire is a safety wire that grounds the circuit in the event of a fault. In short, along with fuses and circuit breakers, the mains uses 33% of its total copper mass, the ground wire, for safety.

Power over Ethernet (PoE) and Smart Buildings: Part II
Figure 1: Cross section of a 2.5 mm2 solid copper power cord (left) next to a 23 AWG solid copper CAT6 cable at the same scale (right).

Power over Ethernet (PoE) transmits short distances (up to 100 meters) of direct current over Ethernet cables between Power Source Equipment (PSE) and Powered Devices (PD). According to the PoE standard, up to eight copper wires are used to transmit DC power, including the return path. In short, PoE doesn’t use any copper wires for security. Logically and architecturally, the PoE standard moves security control from the copper wire (mains power) to the silicon. There are two benefits here: Silicon is much cheaper than copper, and you can code the silicon. But you can’t encode copper.

2-Pair Power Supply vs. 4-Pair Power Supply

Ethernet uses an RJ45 connector, which has eight contacts. These contacts are divided into four differential (diff) pairs (Figure 2). In 10BASE-T (10 Mbps) and 100BASE-TX (100 Mbps) networks, only two of the four differential pairs are used to transmit data, and the remaining two differential pairs are not used. In a Gigabit Ethernet (1Gbps) network, all four differential pairs are used for data transmission.

Leveraging existing 10/100/1000 Ethernet infrastructure, IEEE 802.3af (now called PoE) provides 350 mA/pair, 57V max, and IEEE 802.3at provides 600mA/pair, 57V max (referred to as PoE 1), utilizes these unused pairs to provide power, enabling two alternative modes; alternative A or B.
A. Alternative A (PSE), or Mode A (PD) transmits power on separate Pair 2 and Pair 3.
B. Alternative B (PSE), or Mode B (PD) transmits power on separate Pair 1 and Pair 4.

Meanwhile, PoE 2 or IEEE 802.3bt runs 4-pair power at 960 mA/pair using all four different wire pairs, with a maximum of 57 V. The power supply terminal can transmit up to 90W of power.


Figure 2: 2-pair power supply vs. 4-pair power supply

IEEE 802.3bt (90 W) classification

The Ethernet Alliance further divides these four types into eight distinct categories, as shown in Figure 3. For Power Sourcing Equipment (PSE), each PoE 2 class (5-8) is divided by 15W power difference, and for Powered Device (PD), each PoE 2 class is divided by 11W power difference, finer The division of categories and types optimizes the energy efficiency of multi-port PSEs and provides various powers for connected PDs, especially as the number of connected PSE ports increases, the energy efficiency improvement is more obvious.


Figure 3: IEEE 802.3bt classification

IEEE 802.3af/at/bt power supply stage

PoE power delivery between PSE and PD follows five distinct phases, as shown below and in Figure 4.
Phase 1: Detection
Stage 2: Classification
Phase 3: Launch
Stage 4: Running
Stage 5: Disconnect

The PSE contains an Rsense resistor in series with the return current path to measure the current drawn by the PD. There is also a 25k pull-down characteristic resistor on the PD to inform the PSE for detection.


Figure 4: Stages of PoE Power Delivery (Source: Ethernet Alliance)

Phase 1: Detection

When the PSE and PD are connected via an Ethernet cable, the PD provides a 25 kΩ pull-down resistor to the PSE (Figure 4 right). The PSE then took two current measurements over a 500-millisecond period.

1) Apply voltage V 2.8 V and measure I
2) Apply voltage V 10 V and measure I

By calculating ΔV/ΔI, a PSE can be accepted as valid if its measured resistance value is from 19 KΩ to 26.5 KΩ. Otherwise, the PSE MUST treat the detection as invalid. The benefit of making differential measurements is that any surrounding noise (the source of the noise is called aggressor) will be common mode to each measurement and will therefore be rejected (common mode rejection).

Stage 2: Classification

During the classification phase, the PD announces its required class signature, or power requirement, to the PSE. As shown in Figure 5, the classification phase is divided into five category events or time slots.

1) Class Signature 0: 1 mA to 4 mA
2) Class Signature 1: 9 mA to 12 mA
3) Class Signature 2: 17 mA to 20 mA
4) Class Signature 3: 26 mA to 30 mA
5) Class Signature 4: 36 mA to 44 mA


Figure 5: Class features generated by PD

The figure captures the class signatures (rows) required during each class event (columns) to determine the PD class (1-8). For example, a class 7 PD will provide 40 mA during class event 1, 40 mA during class event 2, and 18 mA during class events 3 to 5. The PSE measures the PD’s current sink during each time event to understand the class of the PD.

The PSE is responsible for applying the voltages described in Figure 6 below to the line, while the PD is responsible for drawing corresponding up to five different currents called class signatures.


Figure 6: Class Signatures and Current Levels

automatic classification

As shown in Figure 5, category event 1 is longer than other category events. This is specific to 802.3bt, not 802.3at or 802.3af. If the PD is also 802.3bt compliant, the PD can change to class signature 0 (1 to 4 mA) within 81 ms of class event 1, which informs the 802.3bt PSE that the PD is also 802.3bt and supports automatic classification.

After the PD is powered on, the PD runs at its maximum power for about 1.2 seconds. The PSE measures the power of the PD and adds some margin. This new power level is the power provided to the PD after the PSE is optimized.

Automatic classification optimizes the power distribution of the PSE. For example, if a PD requires a maximum of 65 W during operation, the PD will confirm to the PSE that it is Category 8 to guarantee that the PD gets 65 W. Without automatic classification, the PSE will allocate 90 W to ensure the PD gets 65 W. With automatic classification, the PSE may only measure 66.5 W (short cable length, line loss about 1.5W), +1.75 W headroom = 68.25 W distribution. Compared with the original 90W, the power saving is 21.75 W or 25%. While this may not seem like much, if the PSE switch has 8 802.3bt ports, auto-classification can optimize each port (with different wire losses depending on cable length), saving hundreds of watts in total energy efficiency.

Phase 3: Launch

During the startup phase, the PSE is responsible for limiting the inrush current to 450 mA for Category 1 to Category 4 and 900 mA for Category 5 to Category 8.

During the startup phase, the PD is responsible for limiting the load current to 400 mA for classes 1 to 6 and 800 mA for classes 7 to 8.

Stages 4 to 5: Run, Disconnect and Maintain Power Signature (MPS)

Power Sustain Signature (MPS) is a keep-alive function in which the PD draws periodic current pulses from the PSE to inform the PSE that the PD has not been disconnected. If the PSE does not receive the PD’s MPS after every 400ms, the PSE must disconnect the PD’s power.

IEE 802.3bt PD application block diagram

Figure 7 depicts a typical 802.3bt powered device (PD) application diagram. From left to right, transformers AC couple Ethernet 10/100/1000M data to nearby processors. Full-wave rectification is done by GreenBridge™ 2, which consumes less power than traditional silicon diode bridges. Onsemi’s PD interface chip NCP1095 (pin 7) provides a 25 kΩ detection pull-down resistor, while pins 2 and 3 determine the power requirements of the PD through the Class category (external resistance value), in the classification event after connection communicated to the PSE. Pins 6, 8, 9 and 10 control surge and provide over-current protection (OCP) through external Rsense and MOS transistor gates, respectively. Three bits of status information to the external processor are provided on pins 13, 15 and 16. Pin 14 (PGO) notifies downstream DC-DC devices when the PoE power output is stable. Pin 4 allows the NCP1095 to power up from the local auxiliary supply, while pin 6 controls autoclassification, a new feature of 802.3bt.


Figure 6: 802.3bt application block diagram

ON Semiconductor also offers the NCP1096 controller, which integrates external FETs and Rsense.

You can code the silicon

Relatively speaking, fuses, circuit breakers, and grounding wires are less flexible for preventing electrical fires, especially when compared to the capabilities of IEEE 802.3bt. The power supply functions provided by IEEE 802.3bt, such as classification, automatic classification, surge control and MPS, are far superior. For example, with utility power, rodents hiding in walls or ceilings can easily start an electrical fire without warning. In contrast, if the PD does not provide an MPS to the PSE every 400 milliseconds, the PSE automatically cuts power to the PD.

We can easily imagine that coding the PSE to catch unexpected disconnects would trigger an early warning sign to the IT department, potentially preventing a catastrophic event like a building fire. At the same time, classification and automatic classification intelligently distribute the exact power required by a load. This is a very safe and efficient way to distribute electricity. As mentioned earlier, silicon is much cheaper than copper, and you can code silicon, but you can’t code copper.

The Links:   CLAA150XG02S CM600HA-24E