Realize the co-line transmission of data and DC power on Ethernet

Power Over Ethernet (PoE) is a power technology that has just been ratified as the IEEE 802.3af standard. It takes advantage of the ubiquitous common Ethernet network to deliver DC power while transmitting data packets. The recently ratified IEEE standard looks set to be the first instance of a universal power outlet or socket being defined for global use. It means that a variety of connected devices that need continuous power, such as IP phones, wireless access points, and network surveillance cameras, no longer have to be connected to their own AC power source. It also means that these devices do not need to be placed near an AC power outlet, so power cables can be eliminated.

In a PoE system, the client device that receives power through the existing Ethernet is called a powered device (PD), and the device that provides power to the PD is called a power sourcing equipment (PSE). The power consumption of the PD is limited to 12.95W, and the PSE output is limited to 15.4W per RJ-45 port. Each PD can draw up to 350mA of continuous current if the Ethernet connecting cable and the physical layer equipment (PHY) transformer are well balanced.

The IEEE standard specifies different power ratings for PDs and PSEs to account for voltage drops along CAT-5 Ethernet transmission lines (which can be up to 100 meters long). Voltage drops are more significant over longer links, so the PSE output voltage must be higher than the nominal 48VDC to get the maximum power to the PD. As a result, voltages as high as 57VDC can be seen everywhere on Ethernet links.

Most PoE networks can be implemented using either endpoint or midspan PSEs. An endpoint PSE integrates an Ethernet switch and power supply in a single device and is placed at the other end of an Ethernet link. This type of PSE provides the easiest way to implement a PoE network because power is already present at the end of the Ethernet link. This type of Ethernet switch is sometimes referred to as having "inline power." As shown in Figure 1, an endpoint PSE is well suited for deploying a new infrastructure network.

Figure 1. Power is distributed across the signal wire pairs for both the endpoint PSE and PD devices. This type of PSE provides the easiest way to implement a PoE network because power is already delivered to the end of the Ethernet connection.

For existing Ethernet networks that cannot be upgraded in this way, a midspan PSE can be used to insert power into the Ethernet network. A midspan PSE can provide power through the "spare pairs" in the CAT-5 cable, which is the most cost-effective method if only a few Ethernet devices need to be powered. For example, a LAN with 4 to 24 ports can be part of a larger multi-port network system (see Figure 2). The endpoint PSE differs from the midspan PSE in the way it chooses to provide power - it can be on the same pair as the signal line or on the spare pair. Generally speaking, a PSE must be able to provide power through the signal line pair or the spare pair, but not both.

Figure 2, for the midspan PSE and PD device, power is distributed through the spare pair. The midspan PSE approach is more suitable for existing Ethernet networks that are not undergoing large-scale changes.

Although it seems simple, it takes considerable effort to design such a system. They must be backward compatible to ensure that old equipment will not receive 48VDC voltage on the Ethernet. IEEE 802.3af includes backward compatibility and includes the option to provide power over Ethernet, making it forward compatible. For a designer who is developing a product that is planned to run in a new or existing system - for Gigabit Ethernet or 1000BASE-T/TX, there are some insights from this article.
Cable Selection for Gigabit Ethernet
Gigabit Ethernet can work with end-point PSEs, but not midspan PSEs because it uses all four pairs of CAT-5 cables for data transmission. In contrast, 10BASE-T and 100BASE-TX use only two pairs of wires for data transmission (pairs 1-2 and 3-6), leaving idle pairs (4-5 and 7-8) for midspan power. Therefore, in order to provide line power for Gigabit Ethernet, an end-point PSE switch is required.
IEEE 802.3af supports CAT-3 cable because it was originally used in 10BASE-T systems. However, to maximize signal integrity in new network deployments, and given that each cabling run is typically 10 years old, we recommend using the highest grade Ethernet cable possible (CAT-5e or CAT-6). Gigabit Ethernet (specifically 1000BASE-T) requires CAT-5 cable, but some applications and Gigabit Ethernet switches that use CAT-5 are nearing their limits. As a result, the latest 1000BASE-TX standard requires CAT-6 cable, while the original 1000BASE-T standard required CAT-5 cable.
Detecting PDs
When a PSE is connected to an Ethernet network, the PSE must detect whether each Ethernet device requires power. Therefore, a PD must behave differently than older Ethernet devices. When detecting, the PSE makes a VI measurement while probing the signal line with a current-limited voltage between 2.7V and 10.1V. Table 1 lists the criteria necessary to detect a valid PD. The allowed 1.9V deviation is the result of the diode bridge that is typically used to control the voltage polarity. Since the PD must be backward compatible with midspan PSE applications, two of these diode bridges are required per PD (see Figure 3). The 10μA current offset is usually due to leakage within the PD. Table 2 lists another set of standards that any Ethernet device that does not pass is an invalid PD.






Power Classification of PDs
The earliest impetus for combining power with Ethernet was Voice over IP (VoIP). Because there are so many Ethernet devices (RFID readers, PDA chargers, mobile phones, and even laptops) that can use this convenient power supply method, the IEEE 802.3af standard includes an optional feature called Power Classification that allows the PSE to more accurately manage the power budget. Table 3 lists the different power classes that a PD can provide, and their corresponding classification signatures.



To implement the optional power classification method, the PSE first applies a detection voltage of 14.5V to 20.5V. The PD responds with a signature (classification current) that indicates to the PSE the maximum power that the PD can consume. This information enables the PSE switch to manage the maximum power available to the connected PD at any time. Another feature that you can implement beyond the IEEE 802.3af standard by choosing an appropriate PSE controller IC is to hardware-limit the output power of each PSE port.
Unless the network manager guarantees that all PDs will not be replaced with devices with higher power consumption, sometimes the power budget expected by the switch will be exceeded. In this case, the PSE will refuse to power the port until the PD's power classification meets the requirements.
Another flexible feature in emergency situations is that the PSE can decide which port receives power first or which port to disconnect first when the UPS or backup generator is running out of energy. This allows the switch to maintain power to the most important ports, such as fire alarms, access badge readers, some security cameras and access points, or other data circuits. Including this kind of fault recovery function in the PSE controller IC - which can be implemented in hardware or software programmable - can help minimize the power budget in emergency situations. In this case, you should look for a PSE controller IC that is software programmable.
Detection of disconnected PDs
After the PSE begins powering a PD, it must monitor the PD's Maintain Power flag in accordance with the IEEE 802.3af standard. The PSE must also detect if the PD has been disconnected. The standard defines both AC and DC methods for detecting a PD disconnection. For example, consider a situation where a PD is unplugged from a switch and a legacy Ethernet device is plugged into the same port. If the 48VDC power is not removed after the PD is removed, the legacy device may be damaged.
AC impedance measurements of a PD are generally more accurate than pure DC resistance measurements. A small common-mode AC voltage is passed down the Ethernet link along with the data signal and the 48VDC. You can then measure the AC current and calculate the impedance of each port, which should be less than 26.25kΩ (if the PD is not unplugged). The frequency of this AC voltage must be between 1MHz and 100MHz. Designers should consult the IEEE 802.3af standard for more details on the AC and DC methods of disconnection detection. Regardless of the method used, the measurement and subsequent termination of power must be fast enough.
Advanced Features in Chips
Of all the multi-port PSE chips available, the most common are PSE controllers that can control power on four ports. Look for those with an I2C-compatible serial interface with programmable registers for easy use with an MCU. The advantage of multiple operating modes is even more important in emergency situations.
For example, Maxim's MAX5935 offers automatic, semi-automatic, manual, shutdown, and debug modes of operation. Automatic mode allows the device to operate without software management. Semi-automatic mode (on request) continuously detects and classifies devices connected to a port, but only powers the port when specified by software. Manual mode is very useful for system diagnostics and allows complete control of the device through software. Shutdown mode terminates all activity and removes power from each port. Finally, debug mode allows detailed system diagnostics through fine-grained stepping of the device state machine.

Figure 3. Because Gigabit PoE network connections use all four wire pairs to transmit data, the midspan PSE method cannot be used to power the PD. The fallback modes of 100BASE-TX and 10BASE-T can be used for data transmission as usual. Therefore, PDs running on Gigabit Ethernet must be powered by an endpoint PSE switch.

Figure 3 is an example of PoE system design. This simplified block diagram demonstrates the connection between a Gigabit Ethernet PSE and a PD. Since Gigabit Ethernet cannot introduce mid-span power, the 100/10M Ethernet mode can only be connected to an endpoint PSE switch (the MAX5940 PD interface controller does not require a diode bridge, but can also work with a diode bridge when needed). Today's PD interface controller ICs (such as the MAX5941 and MAX5942) include a pulse width modulation (PWM) controller, although PDs generally include a DC/DC converter.