I can't seem to power RB5009 using POE on ether1

Hardware MikroTik hardware general
1

I have a rb5009upr+s+in for a few weeks now and getting along well with it. Particularly like the power of queue trees and how effectively they work with cake.

I am continuing to use my USW-Flex-2.5G-8-PoE because of the number of 2.5G ports I need to connect my wifi 7 access points.

I have found that I can power the rb5009 by poe using ports ether2 to ether8, but have had no luck powering it using ether1.

Powering the rb5009 by ether1 would be good as it could also be the 2.5G trunk between router and my switch. The switch can provide poe++ but note that the ether1 port maybe expects poe+ and there is no way to change the port profile on the switch to fix it to poe+. I just presume it should negotiate the power requirement.

I’ve tried brand new high quality cable with no difference. When I plug the patch cable into ether1 nothing powers up and the router remains dead.

Is there something I’m missing here, something I can do or try or is this something that I just have to live with?

2

POE-in on ether ports is different for ether1 and 2-8.
ether1 uses 802.3af/at (POE+) whereas ether 2-8 use Mode B (not POE+, I believe its passive POE).
If your switch can not provide POE+, it will not work on ether1.

Why not simply use the power brick which came with RB5009 ?
Then you can still use ether1 as trunk (but not for power).

Or if you really want to thinker with it, there may exist adaptors which can be used to get power out of a POE line and feed it in separate wires. Those can then be used to feed RB5009 using 2-pin terminal. Not sure anymore but I seem to recall I saw it being mentioned somewhere...
Be careful though it needs to support 2.5Gb then. That might become a problem.

3

Strictly speaking, the ether2-8 should only accept Passive PoE in, (as mode B), so unless there is a setting for Passive PoE out on the switch (AND Mode B) , the RB5009 should not power on from any of them if 802.3af/at/bt is provided.

Roughly, there is usually a pre-detection of the connected device on the power supply side:

  1. a resistance below 3 KOhm or so means that the device is NOT PoE compatible and power should NOT be applied to that port
  2. a resistance above 3 KOhm or so means that the device might be Passive PoE compatible so power may be applied to the port (if settings are correct)
  3. a resistance above 26 KOhm or so means that the device might be 802.3af/at/bt compatible so power may be applied to the port (if settings are correct AND power negotiation is carried successfully)

So what you report is strange.

Many devices (but not Mikrotik) use Mode A for passive, in the 802.3af/at/bt scheme, the one "in command" is (should be) the PSE (the switch in this case) that "tells" the PD (the RB5009 in this case) that it is going to supply power either in Mode A or Mode B and the PD should comply with that.

I am not at all familiar with the settings available on that USW-Flex-2.5G-8-PoE device, maybe you got it working "forcing" passive PoE out and the switch actually supports Mode B.

But if the negotiation is not carried properly (no matter if the fault is on the switch or on the router side) there won't be voltage out, this is by design in 802.3af/at/bt.

What is happening (maybe) is that the switch initiates the connection but this does not go on or doesn't complete in the allowed timeframe:
https://en.wikipedia.org/wiki/Power_over_Ethernet#Standard_implementation
The correspondence is:
PoE = 802.3af
PoE += 802.3at
PoE ++= 802.3bt

But negotiation should start "from the bottom", i.e. first test for af, then for at and finally for bt, so your switch that provides PoE ++, i.e. 802.3bt should be -in theory - compatible withe the router, that should accept 802.3af/at (this is clearly a mistake in the specs of the RB5009, a device that "idles" at 14 W cannot be powered by 802.3af that maxes out at 12.95 W ) 802.3at should be needed.

If you have mains power near the switch and not near the RB5009, probably your only way out is to use an external 802.3at power injector, or a 802.3at splitter near the router but 2.5 Gb versions of these devices are usually industrial (please read as "not cheap").

If you need to use PoE OUT from the RB5009 the use of the 802.3at power injector is ruled out, see:
RB5009 / PoE-IN *and* PoE-OUT at the same time

(the RB5009UPr+S+IN can seemingly provide PoE out only if powered form DC jack or 2 pin connector)

BUT, if you power the RB5009 through PoE (802.3at splitter to either power jack or 2 pin connector) you would only have 25.50 W - 15 W = 10.50 W available for PoE out from the RB5009.

So you would need an 802.3bt 2.5Gb splitter and that would be even more costly.

4

RB5009 ether1 is not a truly 802.3af/at compliant PD - for that 1500V isolation is required, which means an isolated (flyback or forward) DC/DC converter, not just a cheap step-down with common negative ground connected to the metal case. I’d suggest to check for ground loops, does the 5009 power up from PoE on ether1 if that’s the only port connected (with unshielded cable!) and the metal case is isolated (not touching anything conductive)?

Cube60Pro ac is another example of such (almost but not quite PoE standards compliant) device, the mount is electrically connected to the negative of PoE power. The isolation requirement applies to the PD only, the PSE (switch) many ports are usually not isolated from each other (sharing a common power supply, which might have positive or negative or floating ground).

PoE is just a bit of analog electronics, really not rocket science - yet so many devices (not just MT, and even relatively new products) get it wrong, resulting in a total mess (24V passive sometimes with detection that sometimes works, 48V passive, 48V active, or Cambium 30V passive with reversed polarity) and having to be very careful to not release the magic smoke by plugging the wrong device into the wrong power injector.

5

Thank you to everybody that has replied. There are clearly some very knowledgeable people on this forum with very feasible technical explanations of what might be occurring with my setup.

I will try marekm’s test for ground loops and whether that works and will consider the options.

It may be that I have to revert to powering the device with the original power brick it’s not a big problem.

I’m just trying to reduce the number of large power bricks, cables and heat generated to a minimum to keep everything neat and tidy and wife approved :white_check_mark:

6

If the goal is to reduce the number of power bricks, it is also worth to look at their efficiency - use larger power supplies, try to find the most efficient ones you can afford (the higher price may be paid back rather quickly by energy savings). For example, I often use Mean Well SDR series DIN rail mount power supplies (two 240W units, with 24V and 48V DC outputs, some devices like netpower 16P use both) - not the cheapest ones, but rated for 94% efficiency (many older units are more like 80%, which means much more energy wasted as heat).

7

I had a look around, there is a (Unifi) power injector that could do, UACC-PoE+-2.5G, which has an affordable cost (but - if it works it will only be a "brick" moved from near the RB5009 to near the switch):
https://eu.store.ui.com/eu/en/category/accessories-poe-power/collections/pro-store-poe-and-power-adapters/products/uacc-poe-plus-2-5g

As expected the 2.5 Gb splitter is more difficult to find, but you could use a 48V to 24V 2A one, brand Revodata can be found around or below 20 Euro, example:
https://www.amazon.it/REVODATA-compatibile-adattatore-dispositivi-PS5724TG-RJ/dp/B0F2H8T7MJ?th=1

I have read here and there of people successfully using on 2.5 Gb connection PoE injectors or splitters rated for 1 Gb connection, but cannot say how reliable these reports are.

8

Same here, cannot power RB5009UPS+S+IN on ether1, which is what I need for 2.5G trunk, switch is TP-Link SX3206HPP, which is 802.3 bt/at/af capable 60W per port. No problem in powering with ports ether2-ether8, but not what I need with 1G port trunk.

This should of work on ether1. Yet another disappointment.