EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: SiliconWizard on October 11, 2019, 01:18:54 pm
-
Would some of you guys happen to know how much power can be drawn from the port of a typical Ethernet switch (not meant for PoE)?
I'm typically looking to draw under 100mW, and not requiring any additional PS with the possibility of plugging to any switch would be great.
I haven't really studied PoE in details yet, so maybe the obvious answer is: no problem. Or maybe: forget about it.
Any thoughts?
-
From electric point of view:
Ethernet is +/- 2.5V into 100 ohm load, but isolated via transformer so no DC without PoE, and to get AC you'd have to have something to send traffic to that port. And there is a good chance switch will just interpret it as link down, or as error and dont do anything on that port.
From system administrator point of view:
If you connect a POS that tries to do that in MY network, I will find you, and I will kill you. :-DD
-
You can't draw power from the active data pairs because that will upset the impedance and you can't inject DC onto them, either, because the transformer at either end will just short it out. I suppose you might be able to use the Tx pair for (+) and the Rx pair for (-) but that sounds really sketchy and, again, you have to be really careful not to upset the termination impedance at either end.
It's best to use one or both of the normally unused pairs in a CAT5/6 cable but then you are looking at a proper PoE solution...
-
It's best to use one or both of the normally unused pairs in a CAT5/6 cable but then you are looking at a proper PoE solution...
Except the base PoE standards have a huge hole, they require a minimum power draw which eliminates lots of interesting low power "edge" nodes. Not sure if that was addressed in the 802.3bt update.
-
Except the base PoE standards have a huge hole, they require a minimum power draw which eliminates lots of interesting low power "edge" nodes. Not sure if that was addressed in the 802.3bt update.
The cheap PoE adapters solve that by adding a load resistor, the somewhat better ones automatically disconnect the resistor if the load draws enough power by itself. I suppose a supercap could be a way to make a lossless solution.
-
Thanks.
I thought that true PoE would be the only way, but just wanted to see if there was any possibility with non-PoE networks.
And yes, unfortunately, PoE was not really designed for very low-power devices, so in the case of low-power, there's a significant waste of energy. That sucks.
-
I'm typically looking to draw under 100mW, and not requiring any additional PS with the possibility of plugging to any switch would be great.
I'm not sure how you'd ever maintain an Ethernet link using only the power available from the received data.
Surely even maintaining the link would require 100% of the received power... conservation of energy and all that?
On a long cable, you'd need to transmit more power than you'd received - and if you can do that, you have a device much more interesting than any IoT thing...
-
Thanks.
I thought that true PoE would be the only way, but just wanted to see if there was any possibility with non-PoE networks.
And yes, unfortunately, PoE was not really designed for very low-power devices, so in the case of low-power, there's a significant waste of energy. That sucks.
I'm interested in this too. Did you find a solution? I'm reading this pdf (attached). See "Power Disconnection Scenarios" it's very interesting :-+
It seems there's a lower current threshold of about 7.5mA which at 48V would be 360mW, but the PSE waits 300..400 ms before shutting power down. If you can manage to draw power in pulses of >10mA with 20% duty cycle and a period of less than that it won't shut down, and that would be 10e-3*48*0,2 ~= 96mW. Not sure if I'm understanding that right?
Another thing I don't know (I have zero experience with PoE) is if the PSE will keep trying to power up again and again in a loop after a/the first no power disconnection event/shutdown. Do you happen to know?
"All you need to know about POE.pdf":
-
If I'm reading that paper right all you need is to
- provide a right value resistor so the power deliver can sense it is a PoE device
- turn on load only when > 30V (so the sensing feature can detect that resistor)
and adding "classification" (choosing the power requested) would require only adding a current source and some comparators to enable it at "right" time (and you probably want it if you want to have many devices on one switch because of switch's power budget that can be smaller than "max power x number of ports"
-
The thing I'd want to power draws 200 mW, I'm afraid the PSE will shut power down, because that's less than 7.5mA @ 48V.
-
Thanks.
I thought that true PoE would be the only way, but just wanted to see if there was any possibility with non-PoE networks.
And yes, unfortunately, PoE was not really designed for very low-power devices, so in the case of low-power, there's a significant waste of energy. That sucks.
I'm interested in this too. Did you find a solution?
Nope, I've given up on PoE so far.
I'm reading this pdf (attached). See "Power Disconnection Scenarios" it's very interesting :-+
Thanks for this app note. It's short and informative.
It seems there's a lower current threshold of about 7.5mA which at 48V would be 360mW, but the PSE waits 300..400 ms before shutting power down. If you can manage to draw power in pulses of >10mA with 20% duty cycle and a period of less than that it won't shut down, and that would be 10e-3*48*0,2 ~= 96mW. Not sure if I'm understanding that right?
I think you are. This seems to be exactly what the "DC modulation" paragraph and figure show.
Another thing I don't know (I have zero experience with PoE) is if the PSE will keep trying to power up again and again in a loop after a/the first no power disconnection event/shutdown. Do you happen to know?
I don't know for sure, but I think it will cycle forever indeed. (See: "3 The PoE Provision Process ")
Upon disconnection, voltage and power shut down. Right afterwards, the cycle is restarted (1).
Note that the cycle itself restarts. It won't "power up" the device again after a disconnection event. It will first start again with the detection process. So one way you could implement your device so that if you want it to shut down at some point and never get powered again, you could just programmatically disconnect the required detection resistor.