The benefits to end users of the IEEE 802.3af/t POE (power-over-Ethernet) standard are clear: Without relying on an electrician, anyone who can string a Category 5 or 6 cable can inexpensively install both power and data ports for networked applications in locations with no ac-power plug. Users can simply add a POE power supply at a typical price of $39.95 in a network-equipment closet and connect it with Category 5 twisted-pair cable. Contrast this scenario with the cost for a new power plug—as much as $2000, including labor and parts costs—along with the time you waste in scheduling an electrician’s time.

A POE network has two main components: PSE (power-sourcing equipment), such as the network switch or central router, and PDs (powered devices), such as a VOIP (voice-over-Internet Protocol) phone or a wireless-access point, which connect to the network that delivers both power and data. Networks with a non-POE switch can still take advantage of POE by adding a midspan power supply, which injects power into the network independently of the switch or router, such as the enabling power supply in a network-equipment closet. Backing up the PSE with a UPS (uninterruptible power supply) ensures that your networked equipment, including VOIP phones, will continue to operate even if your office experiences a power failure.

VOIP phones and wireless-access points consume relatively little power. With these types of applications in mind, framers of the current POE IEEE standard, IEEE 802.3af, set the power limits for PDs at 12.95W. However, 13W is insufficient power for many types of equipment, such as multiport wireless-access points and PZT (pan-zoom-tilt) videocameras in building security and surveillance. The standards committee began work on a second version, 802.3at, or POE Plus. This version, currently in draft form, increases the maximum power the PD can draw to 25W. Semiconductor vendors have introduced a bevy of POE ICs based on the draft version, aimed at enabling higher-power POE PDs (Figure 1). (Also see “POE devices cater to higher-powered applications.”) However, 25W is still not enough for some applications, which can require power well in excess of 30W.

The IEEE 802.3at standards committee started with a limit of 30W before lowering it to 25W because of test data that showed that Category 5 or 6 cable temperature increased significantly when installers buried it in a larger bundle of cables and it was carrying as much as 750 mA of current. The committee’s concern was not about safety, because the possibility of fire in this case isn’t even remote, but rather that the signal quality of the Category 5 and 6 cables begins to degrade when cable temperature exceeds 60°C.

However, the cables for applications that require higher power are often not in a large bundles. For example, PZT security cameras, with features such as cold-weather heaters, IR (infrared) lamps, and steerable microphones, are often outdoors and have dedicated cabling. Designers of systems for these applications may consider using a nonstandard version of POE, in which they follow all 802.3at parameters except for maximum PD power draw. With an eye toward these nonstandard applications, all of the POE-interface ICs for the POE Plus version can support more than 25W of power.

Deviating from an industry standard can be a daunting move, but if your PD requires more than 25W, you may have no choice. You can take steps, however, to lower the risks associated with nonconformity. First, ensure that your device’s design really requires more than 25W. PDs of the past share a trait common to almost all electronics that emerged before the current surge in energy prices: inefficiency in their use of energy. Power supplies of five to 10 years ago were often only 80% or less efficient. You should now be able to buy or design a power supply with more than 90% efficiency, which may be enough additional power that your design won’t need more than the standard’s 25W limit.

If your PD has an efficient power supply and still requires more than 25W, however, you may next want to consider the network cabling. Will your application have to work with minimum standard cabling? If so, you must allow for fairly high losses in the cabling system. If, on the other hand, you can specify the cable for your equipment, you can go from Category 5 to Category 6, which generally has a higher wire diameter and correspondingly less resistance and wasted power.

“Over time, the cable will degrade as it carries more power because Ethernet cables are not really power-delivery cables,” says Mike McCormick, manager of Texas Instruments’ POE-products group. “They’re not solid-core cables; their primary use is for signal transmission. So, when we use them for power, they’re not the best conductor. You have to be aware of what type of cabling your customer will use—certainly if you’re going to something that’s beyond the standard. If your device is nonstandard, it may not make sense to stay with standard cabling.”

Make sure you get buy-in from your marketing department, which is often the entity that sets parameters such as going with a nonstandard and therefore more expensive cable. “I’ve met with customers who were designing outside installations of cable, and their marketing department says, 'No, we have to work with the most generic cable standards,’” says McCormick. However, there is no large installed base of outdoor data-communications customer-premises wiring. Limiting your product to “standard” cabling that’s for a different environment may not make sense.

Another option and one that the POE standard allows is to use four pairs rather than just two pairs. Keith Hopwood, vice president of marketing for Phihong Power, maker of midspan-power supplies (Figure 2), says that, although he has heard anecdotally of a few European installations that used two-pair cable, he has never encountered any installation that didn’t use four-pair cabling. Using four pairs increases the amount of power available at the PD and minimizes the heat in the cable. At some point, if cabling and installation costs become significant, you need to evaluate whether bringing in an ac plug might be the best approach.

The issue of resistive-cable losses can also be a surprise with nonstandard installations because the standard’s guarantee of available power at the PD no longer shelters your design. The standard guarantees 13W to every PD, and it also limits the cable run to 100m. Once you go nonstandard, you must calculate the power and voltage drop the cable incurs and determine the power that will be available at the PD.

If you’re designing PSE capability into a switch for compliant or noncompliant systems, you must ensure that the switch actually uses the power it allocates for PDs. Harmik Singh, business-development manager for Maxim’s POE products, emphasizes the importance of power budgeting for PSEs. “Power is scarce, and you don’t want to allocate more than the PD is actually consuming,” he says. If a PD identifies itself as a POE Plus type of device that will require 25W or more but consumes only 10W, the PSE should be able to determine that situation so that it doesn’t allocate the unused power. For example, Maxim’s MAX5952 PSE-controller IC measures the actual current that the PD is consuming, digitizing and reporting in real time the PD’s power consumption so that the system controller can allocate the actual power usage.

With all of the concerns that can arise with noncompliant POE installations, it’s worth asking: Why not have a flavor of the specification that accommodates higher-power devices? TI’s McCormick provides an answer: “In the networking world, it has held true that every option eventually becomes a requirement,” he says. A 96-port switch that must supply 50W/port is a more complex and expensive design than one that must provide 25W/port, especially because the number of high-power PDs on the market, although significant, will probably not dominate the market. For example, an eight-port router, with each port providing 50W, would burden the router with 400W, which most applications would not use.

Even as POE makes Ethernet increasingly attractive within offices and factories, its power-and-data combination also makes it attractive in exterior applications, such as security cameras and sensor networks. However, once electronic communication systems go outdoors, their susceptibility to power surges increases dramatically. Lightning strikes are the most likely cause of a catastrophic surge, but other random events can also send a debilitating surge through the network. The Ethernet cable serves as an antenna, picking up virtually any surge, discharge, or transient.

PD-interface-controller ICs come with a variety of protection schemes. TI’s family comes with minimal internal protection, assuming that designers will want to tailor their designs to the end applications. “When it comes to voltage suppressors and diodes, we believe you’re better off going to a [protection-device vendor],” says McCormick. You can get a higher voltage rating, he notes, and the devices can handle much more power than a mixed-signal device.

Other IC vendors offer integrated surge protection. For example, On Semiconductor’s NCP108X POE-PD-interface controller provides 3-kV cable-ESD (electrostatic-discharge) protection because the company fabricates the device on an automotive-qualified, high-voltage process. Akros Silicon’s AS1135 PD controller offers over 16.5-kV air discharge, 8-kV contact discharge, and 6-kV surge on-chip protection, with a low impedance path to ground.