I am working with POE powered devices, and one use case we see is that our device is connected to a POE injector, but behind the POE injector, there is a POE switch that can do LLDP negotiation.

(The thing here is that a POE injector does not normally support POE software negotiation (LLDP), and that it does not have any IP address.)

According to the POE standard, after the boot sequence the DUT should check for LLDP support, and try to do fine grained negotiation using POE-LLDP.

So, if doing this the simplest possible way, what can happen is that the device receives 25.5W through hardware negotiation (from the POE injector), then the DUT connects to LLDP (in the switch) which might tell it that it has only received 12.95W, (since that is what the POE switch can provide).

Then the DUT suddenly may think it has too little power to work, and shut down.

Also, since the 802.3at (and I assume bt) standard seem to be a bit relaxed in what it allows when it comes to hardware versus software negotiation it is not clear to me how these use cases should be handled.

Recommending that customers do not connect a POE switch behind a POE injector is already done I assume.

We were discussing that if the LLDP signalling indicates a lower power than what we have already received though hardware negotiation, we might discard the LLDP signalling as it is probably from the wrong node, but how did the ones who wrote the specifications think this should be handled?

Are there other recommendations outside the specification on how this should be handled?

Is there any reliable (or at least sometimes helpful) way to detect when a POE switch is connected behind a POE injector?

  • I'd have to look at the (too many) POE standards, but isn't it up the PSE to decide a PD is using too much power and cut it off, not the PD? If your device can do passive / hardware negotiation, that's what it should default to in cases where LLDP differs. This is an unfortunate situation for a device that can limit it's own power usage. (eg. an AP)
    – Ricky
    Nov 11, 2023 at 2:38

1 Answer 1


Using a PoE injector that is certainly possible, since the injector is L1 only and passes through all data, including LLDP on L2.

I don't think there's a 100% clean way out for the PD - there's simply no way to detect that scenario. Usually, I wouldn't expect seeing an additional injector in front of an endspan PSE. The only reasonable use case I see is an 802.3af PSE with an -at injector to increase power. I would simply advise to disable PoE and possibly LLDP on the switch in such a case.

Alternatively, the PD could just ignore L2 when L1 negotiation already provides sufficient power - after all, L2 is optional for up to 25.5 W.

  • That matches our normal way of handling it quite well, but some talk about using LLDP to negotiate DOWN power (from 25.5W to 18W for example), so that more devices can be attached to the same POE switch. Then there would be a reason for the PD device to use LLDP even if it has enough power. Nov 10, 2023 at 10:18
  • I'd say when you explicitly point out the problem to the user and give sound advice you're pretty much out.
    – Zac67
    Nov 10, 2023 at 11:40
  • I can see this as a common problem... switch can't supply enough power (unsupported, too many devices, line too long, etc.) so a mid-span is used. However, most enterprise grade switches can be configured to disable POE and/or LLDP on ports where this needs to be done. "Lesser" switches may have no control over it.
    – Ricky
    Nov 11, 2023 at 2:42
  • Yes, it may be more common than we initially thought. If the standard (or some recommendation) asked the power requested and allocated through HW negotiation to be shown when the DUT / PD connects through LLDP after boot, it might be possible for the PD to know if there is a point in trying to negotiate towards LLDP. Is there no such recommendation / requirement in the standards? Could we lobby in some way to change it in that direction? Nov 13, 2023 at 14:35
  • I'm not sure whether IEEE 802.3 would see that as within their scope - there are simply too many fringe cases to try to catch them all. Currently, there's no PoE maintenance running, the last one was 802.3cv-2021.
    – Zac67
    Nov 13, 2023 at 15:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.