After getting a RB750UP in, we began testing it. It works great except for if you try to power a Motorola Canopy unit from it using POE. If the Motorola is connected and the RB750UP is rebooted, the switch will not turn on the poe for that port even though it says it is on when looking at the ethernet interface, all other POE ports power up just fine. If you disconnect the radio after 5 seconds the switch works normal and turns on the poe for that port and the Motorola works just fine when you reconnect it. The switch powers Ubuiqiti and Trango units up just fine upon reboot. I think it is an issue with the current limiting circuit upon power up. It is easily reproducible. We have tried everything we can think of, we know the Motorola’s draw less than 500ma which the port is rated for.
Don’t the Canopy radios use “reverse polarity” on their PoE wires?
Are the ports POE on/off controllable through Mikrotik command terminal? If so, try a Mikrotik startup script that powers off, waits, and then powers on the POE on that port (as a workaround until you find something better to fix it)
Yes the canopy and trango are reversed. We made up cables accordingly and they work great as long as you connect the Canopy once the poe output on the mikrotik has fully switched on. We measured the Canopy and they seem to draw about 200mA vs the Trango’s which draw 130mA at startup which is still a far cry from the 500mA advertised. A startup script wouldn’t work due to as long as the canopy is plugged in, you can’t get the poe to come up on that port. Though once it is enabled and working you can plug in the canopy and it works great.
Sounds like a hardware problem with the Mikrotik, as you were thinking.
Something you might try: Parallel a couple of ports together (the poe pins). This would spread the load over multiple ports, perhaps enough so that they would work properly. If that works well, you could do that on the internal of the unit instead of the outside, as not to use up any usable ports.
I’ve not seen the board layout, but if there is something that actually limits current per port, it may be possible to bypass it internally on the board as well.
Have you reproduced this on multiple Mikrotik units?
It could be problem with other devices if it does not adhere to standart and does not have correct resistance - as a result it is detected that PoE cannot be turned to to not to burn out the port it is connected to. Load values you are calling should not cause any problems what ever you are doing on port or all of them.
some clarification from wikipedia:
A PD (Powered device) indicates that it is standards-compliant by placing a 25 kΩ resistor between the powered pairs. If the PSE (power supplying equipment) detects a resistance that is too high or too low (including a short circuit), no power is applied.
and
PSE detects if the PD has the correct signature resistance of 19–26.5 kΩ
this is the case of "P"ower feature used on these boards to make sure that no other equipment is burned needlessly (or only their Ethernet ports)
The standard you’re referring to is 802.3af, and that’s not what the RB750UP implements. Simple “passive PoE” doesn’t define any signature detection, so I’d expect any device that doesn’t draw too much power to work. As I can see, I’m not the only one with such troubles (the mysterious Bullet2 that won’t power up, even though two Bullet5s and one NanoBridgeM5 work fine).
I hope Mikrotik has made the MCU that controls PoE upgradeable in the field (the ATTINY461 chip itself does support self-programming, if supported by a properly written bootloader), so we will see some improvements later. I’d like to see support for poe-out=force option - “enable power, I know what I’m doing” (as in most of the simple, passive PoE adapters).
The standard you’re referring to is 802.3af, and that’s not what the RB750UP implements. Simple “passive PoE” doesn’t define any signature detection
Yes! Just because we don’t fully comply with the 802.3af standard doesn’t mean we can’t use the same detection method for the RB750UP “auto” detection. JanisK is correct on this one.
your magic devices just do not have the resistance required, so it will not be powered up. And where did you read that it is passive poe 
Also, it has overload protection and short protection. It is not a simple hack.
Also, i have checked a lot of 100Mbps interface routers and everything worked fine. As well as attempted to burn ports on RB1000 just to be sure that our customers does not damage their equipment.
Edit: so you think that “burn” or “kill” option would be useful?
oh and this is how it works:
http://wiki.mikrotik.com/wiki/Manual:Interface/Ethernet#PoE_out
Well, it’s “burn” or “kill” only if the wrong device is connected to that port. I had such an accident - wrong side of UBNT PoE connected to my good old IBM T41 laptop, it was gigabit and now it’s only 100Mbps… (Still auto-negotiates gigabit but then doesn’t work, have to force lower speed on at least one end.) So I’m more careful now, use patch cords of different color (orange) where PoE is used, etc.
I just want UBNT radios to work powered from the RB750UP, that’s all. It doesn’t have to work with “auto”, just with “on”. The strange thing is, some of the radios work fine with “auto”, and some don’t even with “on”.
If your PoE implementation has specific requirements, please document them clearly. Mikrotik has been using passive PoE for a long time (just like UBNT), so I was hoping to be either that or proper 802.3af standard, it turns out to be something in between.
Edit:
on - similar to auto setting just upper bound of detection range is removed.
would it be possible to change the lower bound from 19k to something like 1k ohm? Many devices that don’t support PoE have 150 ohms (two resistors 75 ohms each in series) between the unused pairs, so called “Bob Smith termination”, so it should still be reliably detected.
please measure resistance between the spare parts on devices that does not work.
If the Mikrotik is shutting the PoE off to protect itself due to incorrect resistance… then why does it work fine when you plug that device in AFTER the Mikrotik is booted? If it is the “wrong resistance”, then it should never work… and should shut off the PoE as soon as you plug it in.
How is the Mikrotik protecting itself from damage if it still powers this device if you plug it in after booting?
Because of that, the resistance theory doesn’t make much sense.
I measured the resistance across the POE wires of the Motorola that doesn’t work, its 2.5K - 3K Ohms, the units that are working fine are 2-3M Ohms. These aren’t even close to the resistance someone was talking about earlier.
resistance measured when device is off may be different from what it is when it is turned on. Devices with less resistance will generally have lower values displayed when measuring.
And board is not protecting itself, it protects device that is connected. Also if you pay attention to what is written in manual then resistance measurement is only done while port is not turned on. When PoE is turned on the port - short and overload is measured, this is why when you set ON and then plug in the device port is not turned off.
So, only self protection is turn on when port is on. (that is short and overload)
But we still need a way to make this work for all devices, other people are having the same issue, the port won’t turn on when equipment is connected, but once it is on it will power the equipment just fine. So maybe there needs to be an option to turn the checks off on a port by port basis for problem equipment so that way most ports can still maintain the built in protection but give us an option to still hookup all out equipment to and use the poe ports which is the reason any of us are buying this equipment to begin with.
I agree. The board should reduce the cabeling in our setup and it should give us the ability to do a power reset. I dont want to have some resistance checks to decide if port should be powered. I want a protection against overload and nothing else. This is no consumer product where the customer have to be protected from himselfs.
By the way: is the 500mA a hardware limit? If not please raise the limit to 750mA. With that we are able to use our existing 12V installations (<5m cable length and one 12V block backup)
we are working on possible solutions to turn on as much as possible devices. Most probably it will be additional setting added to enable unsafe mode. As soon as we receive the device that does not work properly (AFAIK it is or soon will be on route)
I’d like to know that too!
Is it possible to enable/disable PoE on each individual port?
thx
yes, it is described in manual (ethernet section)
but some highlights:
each port is controlled separately, so each can have its settings set.
can be configured through CLI, winbox and webfig (so everywhere)
there is one exception - if short is detected, then power is cut on all ports. It is done so to ensure that on short (that can damage the board) power down is done in fastest possible time.
What did you mean by as soon as you receive the device that does not work properly. At least for me in my post it was a Motorola Canopy unit. Are you ordering one of those up to test with or are you waiting on one of us to send you something.