PoE when load > 12.95W ..?

[[email protected]]

Is there any "standard" way to provide power that is more than 802.11af (PoE)
specifies at the power consumer end..?

Asfaik cat.5 cables use AWG 24 which should allow .8 Amps per wire. By useing
them in parallel one should be able to transfer
..8 Amps * 2 wires * 48 Volts = 76.8 Watt. Useing not only the spare pairs in
a 10/100 setup. 153.6 Watt should be possible.

If 802.11af is followed then limit is 12.95W. Which is to little. If a normal
dc/dc converter is used then the required power is supplied (large lcd).
But it will not follow specification.

Any thoughts on this .. ?

 

[Richard H.]

Is there any "standard" way to provide power that is more than 802.11af (PoE)
specifies at the power consumer end..?

Asfaik cat.5 cables use AWG 24 which should allow .8 Amps per wire. By useing
them in parallel one should be able to transfer
.8 Amps * 2 wires * 48 Volts = 76.8 Watt. Useing not only the spare pairs in
a 10/100 setup. 153.6 Watt should be possible.

If 802.11af is followed then limit is 12.95W. Which is to little. If a normal
dc/dc converter is used then the required power is supplied (large lcd).
But it will not follow specification.

Multiple links in parallel? A separate Cat-5 power feed?

There is a revision in the works to accommodate higher-power endpoints,
but that will take some time to reach market.

The whole PoE thing is quite a challenge on the head-end, where the
PSE's may have hundreds of ports pulling 13W each. It's driving
re-designs of power backplanes, supplies, chassis and facility cooling,
dedicated line power, etc. I suspect this influences the pace of the
spec changes too.

Cheers,
Richard

 

[[email protected]]

Multiple links in parallel? A separate Cat-5 power feed?

The main scenario is 10/100 Mbps setup with the 4-5 and 7-8 pairs for power.
The second one is all four with frequency seperation. And possible 1000 Mbps
operation.

There is a revision in the works to accommodate higher-power endpoints,
but that will take some time to reach market.

Know anything about it?, seems PowerDsine.com is the one actually doing it.
But I only find PSE chips..
Seems the current solution is to simple attach an dc/dc converter and be done
with it (for now).

The whole PoE thing is quite a challenge on the head-end, where the
PSE's may have hundreds of ports pulling 13W each. It's driving
re-designs of power backplanes, supplies, chassis and facility cooling,
dedicated line power, etc. I suspect this influences the pace of the
spec changes too.

One thing that pussles me is why not have the endpoint simple tell how
many Watts/Amps it wants with 0.25 Watt granularity. And then the PSE grants
the request if possible. One could also signal power requirements for power
outages. Useing something like a slow bitstream instead of complicated
measurements.

 

[Walter Harley]

Is there any "standard" way to provide power that is more than 802.11af
(PoE)
specifies at the power consumer end..?

Asfaik cat.5 cables use AWG 24 which should allow .8 Amps per wire. [...]

Sure, in free air. What about in the middle of a 2" thick bundle of cables?

 

[[email protected]]

Is there any "standard" way to provide power that is more than 802.11af
(PoE)
specifies at the power consumer end..?

Asfaik cat.5 cables use AWG 24 which should allow .8 Amps per wire. [...]

Sure, in free air. What about in the middle of a 2" thick bundle of cables?

Well you don't have to use it to the maximum. But there's sure is some more
power to be had.

 

[Richard H.]

The main scenario is 10/100 Mbps setup with the 4-5 and 7-8 pairs for power.
The second one is all four with frequency seperation. And possible 1000 Mbps
operation.

So, there appear to be two wiring options, including the one you list.
The other seems to be level shifting of the signal pairs. I'm not sure
which one is proprietary vs. the official "standard" (maybe both are).
Designs often include a selector for the input power pairs.

Know anything about it?, seems PowerDsine.com is the one actually doing it.

The newer spec? I don't think it's official yet.

ISTR that Linear is quite big in this space; they've advertised it a lot
in the trade rags. And, they make both ends of the link (PSE and PD
controllers).

But I only find PSE chips..
Seems the current solution is to simple attach an dc/dc converter and be done
with it (for now).

Only for the older style, where there was no sensing of the powered
device. Problem is, you plug in the wrong thing (like a non PoE
Ethernet card) and the endpoint goes POP! This was a problem with early
proprietary solutions, because some of the Ethernet card vendors (a la
3Com) didn't leave the extra pairs unconnected.

One thing that pussles me is why not have the endpoint simple tell how
many Watts/Amps it wants with 0.25 Watt granularity. And then the PSE grants
the request if possible.

Well, they do, I believe. There are several levels of power negotiable,
and the device signals the power requirement to the head-end during init
before power is applied (so the PSE can decide if it has capacity).
There's a signature resistance at the end device, and capacitance IIRC.

Check out http://www.linear.com Power Management | Hot-Swap | PoE
Controllers | View Table

Cheers,
Richard

 

[Rene Tschaggelar]

Is there any "standard" way to provide power that is more than 802.11af (PoE)
specifies at the power consumer end..?

Asfaik cat.5 cables use AWG 24 which should allow .8 Amps per wire. By useing
them in parallel one should be able to transfer
.8 Amps * 2 wires * 48 Volts = 76.8 Watt. Useing not only the spare pairs in
a 10/100 setup. 153.6 Watt should be possible.

If 802.11af is followed then limit is 12.95W. Which is to little. If a normal
dc/dc converter is used then the required power is supplied (large lcd).
But it will not follow specification.

Any thoughts on this .. ?

The standard way is to have a voltage between the
differential pairs. Some safety concerns limit this
voltage to a value far lower than the undamaged
isolation would allow. It now depends on your
cable and the treatment of it, on how high you
can go with the voltage. If you treat the cable
well, eg don't pinch it in a door or such, I guess
200V would be doable if you don't tell anyone,
for one-of, in a guarded environment, eg your lab.

For a product being used in and given to the public,
I wouldn't exceed the PoE rating, since these guys
thought a lot about everything.

Rene

 

[[email protected]]

The main scenario is 10/100 Mbps setup with the 4-5 and 7-8 pairs for power.

So, there appear to be two wiring options, including the one you list.
The other seems to be level shifting of the signal pairs. I'm not sure
which one is proprietary vs. the official "standard" (maybe both are).
Designs often include a selector for the input power pairs.

According to:
http://en.wikipedia.org/wiki/Power_over_Ethernet

The PD must be able to handle both options. But the PSE is the one choosing.

The newer spec? I don't think it's official yet.

ISTR that Linear is quite big in this space; they've advertised it a lot
in the trade rags. And, they make both ends of the link (PSE and PD
controllers).

Only for the older style, where there was no sensing of the powered
device. Problem is, you plug in the wrong thing (like a non PoE
Ethernet card) and the endpoint goes POP! This was a problem with early
proprietary solutions, because some of the Ethernet card vendors (a la
3Com) didn't leave the extra pairs unconnected.

Meaning senseing is needed.. , but the non-standard dc/dc approach to PD
should be workable at least. The PSE end might be different due it's the one
powering things and thus responsible to not blow things up.

Well, they do, I believe. There are several levels of power negotiable,
and the device signals the power requirement to the head-end during init
before power is applied (so the PSE can decide if it has capacity).
There's a signature resistance at the end device, and capacitance IIRC.

But the current power levels are to coarse. And there's no support for >12.95W.
Or backuppower specification.

Check out http://www.linear.com Power Management | Hot-Swap | PoE
Controllers | View Table

Seems www.linear.com have problems with their webservers today..
(Linux/WebLogic XMLX)

 

[Richard H.]

Meaning senseing is needed.. , but the non-standard dc/dc approach to PD
should be workable at least. The PSE end might be different due it's the one
powering things and thus responsible to not blow things up.

Exactly. I think you need the PD chip to do this signaling, so the PSE
knows to a) energize the line, and b) the power rating to supply. If
you were using the spare pairs and hacked the signature specs, you could
probably cobble a simple dc converter tapped off the line, but what if
the PSE is configured to level-shift the signal wires?

It seems to me that the PD controller is also needed unless you roll
your own solution (incl a discrete version of the PD controller), in
which case you decide how it works (but at the risk of creating a hazard
for anything else that might be connected to the port).

But the current power levels are to coarse. And there's no support for >12.95W.
Or backuppower specification.

True, the specs are very coarse. I'm sure it had something to do with
trying to keep it simple, probably due to limitations on the signaling
capabilities (rather than trying to actually create a simple standard).

I've heard there's a high-current version of the spec in the works, but
I don't know it's progress.

Unless you control both ends of the link, it would seem the cleanest
solution is to use the PD controller chips; perhaps multiple in parallel
if you need higher current, though customers may object, as the PSE
ports aren't cheap.

Cheers,
Richard

 

[[email protected]]

I've heard there's a high-current version of the spec in the works, but

I don't know it's progress.

I found the "answer":

There's an all-about-PoE site:
http://www.poweroverethernet.com/poe-plus-products.php

That mentions 'PoE plus' and the standard in the works IEEE 802.3at
As I understand it will make 56W operation possible in standard way..

Here's the company makeing it happen:
http://www.powerdsine.com/Developers/Products/PoE/ASIC/PD-83000.asp

And the chip id:
PD83000GAC-0100 LQFP-64 (i2c controlled it seems)

However their distribution chain seems somewhat not easy to deal with.