We have several KNOTs we are configuring for deployed usage. The primary power source will be PoE but there’s a risk that the PoE will fail or be disconnected temporarily, so we have a backup 5V battery power supply. When simulating a PoE failure, the unit successfully and seamlessly uses the battery backup without any issues. However, when PoE is reconnected, it does not switch back to using the higher voltage supply and continues to drain the battery.
/system health print
returns a voltage reading of 5V despite the ~50V PoE being reconnected. Is this the expected behaviour? A soft-reboot does not correct this but a hard-reboot does force the KNOT to prioritise the PoE supply.
Is there any way to force the unit to use PoE when available without a hard-reboot (power off → power on)?
Thanks for the reply. We would like to be able to use the PoE Out potentially so the PoE splitter workaround would prevent that. I’m unsure if we can set the PoE to forced on, I’ll look into that and if we can, see if it makes a difference. The source is either via a PoE+ injector or a dedicated PoE+ gateway (e.g. Unifi Dream Machine SE).
The PoE out will work just fine if you power the Knot via DC jack, PoE out is passive anyway, while PoE in (even if it is not explicitly written anywhere) is compatible with both 802.3af/at and with passive PoE (this latter as low as 18V).
In practice you can take the 24V 1.2 power supply that comes with the knot, feed it to a RGBPOE https://mikrotik.com/product/RBGPOE
and connect the knot at the other hand of the ethernet cable.
This would be plain “passive” PoE and you can experiment what happens with your 5 V supply.
If it works, good, otherwise you can use another RBGPOE to split the power coming from the cable to the DC jack (which would be exactly the same as directly powering the Knot directly from the included power supply) and see how the alternative 5 V supply behaves.
If it works, all you need is a passive (this time 48 V or so) power supply (or a switch capable of providing power “always on” bypassing the 802.3af/at negotiation).
Using (in pairs) injector and splitter avoids the possible issues with mode A and mode B, as the pairs on which the power is transmitted through the cable are determined solely by the splitter/injectors.
In theory an 802.3af/at fully compliant device should be able to use both Mode A and Mode B (in the protocol it is the PSE, not the PD that chooses which mode) but Mikrotik has been traditionally Mode B, so it is entirely possible that PoE in (passive) works only in mode B, cannot really say.
Thanks for your help. We’ve done some more experimentation and it appears the PoE negotiation is the issue as suggested. When we use a passive PoE supply the unit switches between different power sources seamlessly. It only struggles in the negotiation scenario. Some of our deployed setups use a PoE injector which is fine. One specific setup uses a PoE network switch so we’ll look at setting that to passive if possible. If not, we may add the converter suggested or use a PoE inector on that setup too.