AF24 connected to WS-12-250-AC - cable length measurement

[IntL-Daniel]

We have long time ongoing issue on our AirFiber24 (I believe that it is all AirFibers related) units connected to WISP switch (WS-12-250-AC, currently on 1.4.3rc7).

- Everytime we do a reboot or powercycle units (port bounce), the cable length measurement in AF24 WEB UI did not pass and shows "DATA Cable Length:
< 20 m" althogh we have a 35m long cable there.
- What we know is that current AF24 SNR and cable length mesurement (fw 3.2) are offline and are initiated only just before the ethrnet line come up.
- The 100% solution for this isssue is to Disable/Enable the LAN port on Netonix where AF24 is connected to (not the POE, just only a data port) or make a soft reboot on AF24 (make some changes in web UI and confirm Apply).
- We have also several AF24 units powered by original POE and then connected to other switches and there we never saw this issue.

So I believe that there is some issue related to the situation when the AF24 unit is powered directly by the POE in the switch. I am not sure if this is a AF24 fw issue or Netonix issue, we have no chance to check any other poe switch suitable for AF24. But even if there is an issue in the AF24 we cannot expect any positive feedback from Ubiquity (remember the current issue with pause frame storm) and would be nice if Netonix can "expect" this AF issue and can do some fix? It is long time anoying issue for us. Thanks for feedback.

[sirhc]

I have MANY AF24 units powered by Netonix switches at my WISP and do not have any issues.

Are you using an Ethernet Surge Protector, if so try removing it and see what happens.

[IntL-Daniel]

Yes we have surge protectors there (original from Ubiquity) and in case of warranty we cannot (permanently) remove them. Otherwise we have same protectors on other places (where AF is not powered by switch) and also never saw this issue.

[sirhc]

All the POE brick and the switch does is put voltage on the wires.

This voltage is the same rather it is coming from the switch or the POE brick.

Removing the Surge protector and seeing if the problem still persists is another clue as to the cause.

What possible thing could we do in the switch to change what the WEB UI in the AF24 is reporting?

All we do is turn voltage on or off on the pairs.

[IntL-Daniel]

I have MANY AF24 units powered by Netonix switches at my WISP and do not have any issues.

So if you do "port bounce" on port with AF24, it will come up with correct cable length measurment (I mean different than <20m)?

So just for sure..here is my cfg related to AF port on WISP:
* PS Poe Smart - disabled
* Flow Control Both enabled
* MTU 9600
* ISO diabled
* DS disabled
* Stats enabled

[IntL-Daniel]

What possible thing could we do in the switch to change what the WEB UI in the AF24 is reporting?
All we do is turn voltage on or off on the pairs.

In our case I think that the solution from Netonix side could be this - automatically disable/enable LAN port for 1 sec in a few seconds (here is the question how much*) after power to the port is enabled. (*I guess that 1-3s would be enough)

Edit: ...or maybe better - automatically disable/enable LAN port (not POE) for 1 sec whenever you detect LAN cable pluged to the LAN port on the WISP ?

I saw many screenshots here in the forum from other users where is "lan cable <20m" visible and everybody "live with this as normal" I think.

I can 100% confirm that this issue we never saw with original POE brick even connected with surge protectors to other (simple) switches. I cannot 100% confirm what would happen if we use original POE brick with LAN connected to Netonix or stay using Netonix as POE and remove our surge protectors.

[sirhc]

You know the more I think about this the more I remember seeing this in an airMAX AC radio as well?

I think I said something to Matt Hardy at UBNT about it, let me go look it up.....

This was a skype message I sent to Matt back on March 18th, 2016 and 2 pictures I sent him

You need to have the programmers fix the reported cable length on the PowerBeam 5AC 500. It reports <20 m when the actual cable length is 200+ feet. Not sure if this affects other AC devices. firmware v7.2(SC) (fear)

My guess is they are implementing their cable diagnostics while their PHY is still negotiating?

Not sure what we can do about it? I really do not like putting in a kludge like that if all possible.

[IntL-Daniel]

Are your screenshots of 5AC500 from unit powered by Netonix switch? If yes then it looks like the same issue!

But on other side...we have many 5AC500 units powered by UBNT ToughSwitch POE Pro a we could see correct cable length there (but we are on 8.0-beta15). But what I can do is to take a 5AC500 unit (feed only) and connect it to Netonix with 30m long cable and test it with 7.2 and 8.0-b15. But to do this I have to go to the 110m high building roof first... so maybe you can do this test easier? :-)

[IntL-Daniel]

My guess is they are implementing their cable diagnostics while their PHY is still negotiating?

It may be the right way! It looks like UBNT try to detect the cable length "in the moment" when Netonix still cannot give them the right answer?

[lligetfa]

My guess is they are implementing their cable diagnostics while their PHY is still negotiating?

It may be the right way! It looks like UBNT try to detect the cable length "in the moment" when Netonix still cannot give them the right answer?

Netonix does not have anything to answer to. The AF is solely responsible for answering its own probe.

The way cable length is measured is with TDR (Time Domain Reflectometry) Where is measures how long it takes for the pulse to reflect back. The switch plays no role in it at least not by design. It might inadvertently play a role in so far as the switch might also be trying to do a TDR measurement at the same time.

Anyway, does this cause you any real hardship or is it just an insult to your eyes? Cable length doesn't change so there is no need to look at it every day. The only time you may need to look at it is when there are ethernet issues at which time bouncing the port will get you the measurement you need.