Linksys Lies! WAPPOE(12) isn't 802.3af !!

I got a couple of sets of the Linksys WAPPOE12 Power-Over-Ethernet boxes for putting WiFi APs in distant places without needing AC power nearby, and checked very carefully to ensure that they were 802.3af PoE, as I didn't want one of the non-compliant PoE kludges, so I could be sure that everything was compatable, and safe to use with old and new equipment.

OOPS! I pulled the covers off the 'injector', the piece that connects the 48V to the spare pairs _only_ when the client device is determined to have PoE capabilities, and found

Not a single semiconductor (well, except the power LED), switch, current limiter, or anything. Just slap the 48V right across the spare pairs and hope nothing blows up. The WAPPOE injector is the same, with the exception of a blue plastic case.

Well, I guess I know why it was so cheap. Sigh. I dunno if I even dare try to see if the splitter will work on 802.3af switches, which supply phantom power over the data pairs...

Grrrr.... The vendor doesn't want returns, hopefully Linksys will be open to a refund.

Methinks the WAP54G uses the same LM2941 LDO regulator as the WRT54G. The radio runs on 3.3VDC. The regulator needs about 0.5VDC. You should be able to run it down to about 4.0VDC.

My BEFW11S4v4 running on 3.8VDC:

Nope. Short circuit protection. Maximum short circuit current is: 48VDC / 24 ohms = 1A That's divided between two pairs of contacts so the maximum connector current is 0.5A. That's slightly under the rating for the connector so the RJ45 connectors should be ok.

CAT5 is 3 ohms per 100ft per conductor. 300ft of CAT5 using 2 pairs of wires for power is equal to 9 ohms of loop resistance. Adding the

9 ohms to the series resistors yields 21 ohms of total loop resistance (assuming there's no additional resistance in the terminating end of the PoE).

The WRT54G sucks about 0.3A at 13.6VDC input. I'll assume there's a 3 terminal regulator inside the terminating end of the PoE thing. It would need about 15VDC on the input, at 0.3A to run a WRT54G. Grinding the numbers: Copper voltage drop = (0.3A * 21 ohms) = 6.3 VDC Max voltage drop allowed = 48 - 15 = 33VDC So, this should work just fine.

Assuming 0.3A per access point, this contraption should run: (48 - 15) / 21ohms / 0.3A per AP = 5.2 which means that it could power 5 access points for a total of about

1.5A current drain.

Well, the real 802.3af spec calls for 5ea different weird "classifications" of load. I won't pretend to undestand why this is necessary.

CLASSIFICATION: CLASS PD POWER(W) R(CLASS)ohms 802.3af LIMITS(mA) NOTE

0 0.44 - 12.95 4420 ±1% 0 - 4 Default class 1 0.44 - 3.84 953 ±1% 9 - 12 2 3.84 - 6.49 549 ±1% 17 - 20 3 6.49 - 12.95 357 ±1% 26 - 30 4 - 255 ±1% 36 - 44 Reserved

48VDC * 7ma = 3.4 watts. Yeah, that's about right.

I used to use DC to DC inverters to run a mess of AP's on top of a tower. Lots of RFI from the inverters so I had to shielded power supplies. Then I discovered the wide range of voltages that the Linksys stuff will run on. I dumped the DC to DC inverter and lived happily ever after.

Yeah, they lie.

Yes, I'm characterizing the WAP54G, looks like it'll do 4.20 to over

18.86V (as high as this supply will go). More news as it happens, plus graphs and such.

Those are 12 ohms, in series with the Ethernet input. Maybe some kind of matching for the input? Wierd...

On further inspection, there's a self-resetting fuse at SI1, with "R110 4289S" marked on it. Note that they are feeding in 48V, and using the "Power Splitter" at the other end to convert down to {12,5}V for the device they are driving.

My programmable load is down, so I can't characterize the 12V output, but since the WAP54G only draws about 3W (and the 802.3af PoE spec is somewhere around 15W), I'm sure they aren't stressing anything. In fact, at 48V, it's only drawing about 7mA (plus or minus the inefficiency of the switching supply in the splitter).

The whole point of this exercise was to have less power (heat) at the AP, but AFAICT the splitter dissipates more power than the AP, so I suspect I'll go back to using a spare cable or pair to power the AP...

The label says 12V?

Ignore comment about 12V, brain just engaged, it's not quite 8am. :)

David.

15.x watts from the sending switch / injector, 13.y available to the device (allowing for cable losses)

i dont have the exact numbers to hand here.

the power classifications are to allow a switch using 802.3af with a constrained PSU estimate aggregate load and shed any overload without being too conservative.

if you work out worst case PoE load on a big chassis switch with 802.3af, then the number get a bit painful (e.g. a cisco Cat6509 with 8 *96 port

10/100 blades works out at around 11.5 kWatts for PoE + power overheads + the switch itself).

power sensing lets you only supply PoE active devices - but a lot of end points dont need anything like 13W, and not many offices are already wired for these kinds of power levels in a wiring closet.

class 0 is what you get if the device / switch or both dont understand the classify scheme.

the others are about the various expected ranges of load - for example some Voip phones may report class 1 or 2.

some manufacturers make a big thing about finer grained power notifies using non standard schemes - Cisco for one.

Those aren't in the power supply circuit, but in the Ethernet circuit. Since these are doing "Midspan" power injection, they are leaving the Ethernet pairs alone and applying 48V on the 'spare' pairs.

5-terminal, LM2576 HVS-ADJ "Simple Switcher", plus choke, diode, and some caps. I can only imagine the WAPPOE and the WAPPOE12 have the same design with a couple resistors to change the output voltage.

Anyone make a CAT5 splitter to allow power over the spare pairs, or am I going to have to make my own? Linksys almost does (a couple of their WAPPOE "power injectors" back to back), but that's kinda pricey.

8*}

Not quite true, just characterised the "splitter". The power supply is a regulated 48V supply, and the splitter draws 9.42mA DC with no load, or about 0.45 watts.

When connected to the WAP54G V.2, which is a 3W load, the whole assembly draws 95.0 mA, or about 4.56 watts. At this load, the switcher in the splitter is about 66% efficient.

Still, not much point to having a power supply feeding some random voltage to a device that'll accept something like 5-25VDC, so I'm going back to splitting the cable between 10/100BaseT Ethernet and DC power on the spare pairs for these particular devices.