I’m probably going to get some heat for this, but I have to get something off my chest. At Cisco Live this year, I saw a technology that was really flexible, with amazing controllability potential, and just cool: PoE-based LED lighting. Rather than connecting light fixtures to mains power and controlling them via a separate control network, it’s all one cable. Network and power, with the efficiency of solid-state LED lighting, with only one connection. However, after several vendor conversations, I can’t escape the conclusion that the idea is inherently, well… dumb.


Okay, Not Dumb, Just Math


Before Cree®, Philips®, or any of the other great companies with clever tech in the Cisco® Digital Celling Pavilion get out their pitchforks, I have to offer a disclaimer: this is just my opinion. But it is the opinion of an IT engineer who also does lots of electrical work at home, automation, and, in a former life, network consulting for a commercial facilities department. I admit I may be biased, and I’m not doing justice to features like occupancy and efficiency analytics, but the problem I can’t get past is the high cost of PoE lighting. It’s a regression to copper cable, and worse, at least as shown at Cisco Live, ridiculous switch overprovisioning.



First, the obvious: the cost of pulling copper. We’re aggressively moving clients to ever-faster WLANs both to increase flexibility and decrease network wiring costs. With PoE lighting, each and every fixture and bulb has its own dedicated CAT-3+ cable running hub-and-spoke back to an IT closet. Ask yourself this question: do you have more workers or bulbs in your environment? Exactly. Anyone want to go back to the days of thousands of cables in dozens of thick bundles?  (Image right: The aftermath of only two dozen fixtures.)


Second, and I’m not picking on Cisco here, is the per port cost of using enterprise switches as wall plugs. UPNP is a marvelous thing. A thousand-plus watts per switch is remarkable, and switch stacking makes everything harmonious and redundant. Everyone gets a different price of course, but the demo switch at Cisco Live was a Catalyst 3850 48 Port UPOE, and at ~$7,000, that’s $145/port. Even a 3650 at ~ $4000 comes to $84 to connect a single light fixture.


It’s not that there’s anything inherently wrong with this idea, and I would love to have more Energy Wise Catalysts in my lab, but this is overkill. Cisco access switches are about bandwidth, and PoE LEDs need little. As one vendor in the pavilion put it, “… and bandwidth for these fixtures and sensors is stupid simple. It could work over dial-up, no problem.” It’s going to be tough to sell IT budget managers enterprise-grade stackable switches with multi-100 gig backplanes for that.


And $84/port is just a SWAG at hardware costs. Are you going to put a rack of a dozen Catalysts directly on mains power? Of course not. You’re going to add in UPS to protect your enterprise investment. (One of the touted benefits of PoE lighting is stand-by.) The stated goal by most of the vendors was to keep costs under $100/port, and that’s going to be a challenge when you include cable runs, IT closets, switches, and UPS. Even then, $100/port?


Other Considerations


There are a couple of other considerations, like Cat 3+ efficiency at high power. As you push more power over tiny network cables it becomes less efficient, and at a certain output per port, overall PoE system efficiency drops and becomes less efficient than AC LEDs. There’s also an IPAM management issue, with each fixture getting its own IP. That ads DHCP, and more subnets to wrangle without adding much in terms of management. Regardless of how you reach each fixture you’ll still have to name, organize, and otherwise mange how they’re addressed. Do you really care if it’s by IP you manage or a self-managing low-power mesh?


DC Bus for the Rest of Us


What this initiative really highlights is that just as we’re in the last gasps of switched mobile carrier networks, and cable television provided in bundles via RF, we need to move past the most basic concept of AC mains lighting to the real opportunity of DC lighting. Instead of separate Ethernet runs, or hub-and-spoke routed 120VAC Romex, the solution for lighting is low voltage DC busses with an overlay control network. It’s the low voltage and efficient common DC transformation that’s the real draw.


Lighting would evolve into universally powered, addressable nodes, daisy-chained together with a tap-able cable supplying 24-48VDC from common power supplies. In a perfect world, the lighting bus would also support a data channel, but then you get into the kind of protectionist vendor shenanigans that stall interoperability. What seems to be working for lighting or IoT in general is more future-proof and replaceable control systems, like wireless IPv6 networks today, then whatever comes next later.


Of course, on the other hand, if a manufacturer starts shipping nearly disposable white-label PoE switches that aren’t much smarter than mid-spans, mated to shockingly inexpensive and thin cables, then maybe PoE lightening has a brighter future.


What do you think? Besides “shockingly” not being the worst illumination pun in this post?