POE Smart

[sirhc]

So I have been mulling this over for awhile and I think we are going to put in a feature called "POE Smart".

If a switch is cold booted (Power applied) it will simultaneously test all ports that have POE configured and look for "cross shorts" and cable problems.

If it finds a "cross short" which will burn out the Ethernet transformer it will refuse to turn POE on, log the short in the switch log and turn the port POE option to OFF or maybe some special option called "SHORT" or "FAILED" and possibly send a text alerts (SMTP).

If the test reveals anything else such as cable pairs of different length which indicates the cable is damaged probably due to stretching or kinking during installation, abnormal termination which indicates that that an end is probably crimped poorly it will still turn the POE ON but log it in the switch log and possibly send a text alerts (SMTP).

If the user is turning POE on and it detects that no cable is there (0 meters long) or no device is plugged in which reports all pairs open it will also no turn the POE up and will turn it to OFF in the config and possibly send a message.

***********************

If a user is in the UI or CLU and turns POE on to one or more ports it will check those ports and do the same tests with the same outcome except pop up a warning as to why ports were not energized and report any possible problems on ports that it did enable POE on.

The delay is about 5 seconds but this could prevent a WORLD of problems for the user and us.

[JustJoe]

But why turn it off if it doesn't find something plugged in? I would find that really annoying.

Why not keep periodically repeating test mode and only turn it off if it finds something bad, and get out of test mode if it finds something good?

That's basically the idea I sent to Dave in a PM, but he never replied to it.

[wayneorack]

But why turn it off if it doesn't find something plugged in?

Because, you are not supposed to be plugging things into a hot port!

Chris has made strong recommendations several times against plugging a device into a port that has POE turned on. A miss-alignment of pins could lead to switch or device damage.

[sirhc]

Wayne is correct. I keep hearing how people plug stuff in live all the time. Most famous line is "I do this all the time on my old "ToughSwitches".

Yes and so did I but there is a HUGE difference between .5A and 2A.

To provide 1.5A on 48VH at 55C the Poly Fuses must be rated higher then 1A per pair at room temperature because remember Poly Fuses are NOT an exact science as they Trip based on their temperature and they are rated at 1A which means they are rated to HOLD 1A indefinitely already being warm.

The highest rated Ethernet Transformers we have been able to find and use in our current production runs are 1A per pair verses the original .73A.

Now remember that if you are turning up a port the Poly Fuse is COLD and will have the ability to handle an inrush FAR ABOVE 1A for a short period which = DEAD TRANSFORMER if there is a short.

Now I have tested several dead shorts with ToughSwitches (at least 1 by mistake) and every time it resulted in a completely dead ToughSwitch but with our switch it results in a DEAD PORT not a dead switch, Matt Hoppes can back me up on this test as I am pretty sure he found this out the hard way as well.

[mike99]

I really like the idea but for unplug port with PoE on, why not test it until something is plug in like 802.3af/at ? Some kind of mix between passive and 802.3af/at that you select your voltage but won't power on until something is plugged on. The downside vs af/at would be that it would power on even if device don't support PoE.

[sirhc]

The check takes 5-6 seconds to complete and the function was not really designed for this so constantly checking is not a great idea but we will look into it. Since this feature is from the core (we did not write it) we are considering opening a ticket Vitesse to see if the test can be made quicker but we doubt it.

The switch is passive POE so we are not trying to be 802.3 or we would have put that into it but we do want to protect the switch, the user, and us from oh crap mistakes because people are going to dead short ports and claim they came that way. I personally have flashed and tested EVERY SINGLE switch that we have made so far at the end of the line where I sit with my laptop and assign the MAC address. Now by default I plug into port 1 to do this. Now usually I check all the ports capable of 48VH just because I know how easily these ports can be fried with that much POE. Well anyway I have already received several switches where the end user swears the switch came with port 1 dead and when I get the switch back I find port 1 Ethernet transformer fried? I mean come on, so I am realizing that this scenario will only increase as more switches are shipped. As I said I had fried a couple ToughSwitches because I crimped an end wrong so I am not saying I do not make bonehead mistakes which is my my older brother Doug always called me bonehead.

Now one condition I did not think of is if the switch detects watts consumed drops to 0 for several seconds then the switch will also turn POE off as this indicates the user unplugged the device live or it died and either way you do not want power on the line.

[mike99]

Even if it was done every 5 minutes, I could deal with that. I'm asking this is because we often install switch at top of silo (harvestor) or elevator and technician don't want to bring a laptop at the top of it to power on ports on the switch.

What we do right now is power one backhaul with a external PoE injector and once I can access it from office, I power on the backup link. Once it's connected, we remove the PoE injector from main link and plug it directly and then power on the main link.

That not a big problem but not ideal neither so when I read that you can test before power on, I thinked that maybe their could be a way to work around this. I'm sure we are not the only one that would like to be able to plug antenna with power preconfigured on so that could be a nice feature.

[WisTech]

I'm all for this Chris.

[lligetfa]

Yes, we need Chris to protect us from our own stupidity.

Problem is, stupidity knows no bounds.

viewtopic.php?f=6&t=271

[JustJoe]

No, wait, stop, let's do some thinking outside the box and not sit on some NIH attitude as a couple of people here are doing.

Maybe the full "cable test" function is overkill, so what if it was stripped down to the safety test essentials ...

OK, in the Email exchange with Dave, while I thought we could do a bit of brainstorming, I first brought up the idea, that since he has programmable MOSFET switches controlling POE power, could he essentially use them, plus the voltage/current sensors and the system processor to act as a programmable PWM voltage/current regulator? He explained that no, that basically the voltage/current sensors were not on a bus structure that was not fast enough to handle that type of continuous feedback regulation.

So I then said:

But imagine that you could probably program a free-running, narrow enough pulse width for an "on" 48VH port (for example) so that even without feedback monitoring, it would be just enough to power an unplugged port for an "ohmmeter" or "continuity" test function. (But not enough to burn out the Ethernet transformer.)

The processor wouldn't have to continually monitor that. It could come back every 3 seconds or so to see if the charge voltage on the per port output capacitor dropped more than an open port, but less than a shorted port? Only then would it would run it full on all the time. At least until such time that the current dropped below a certain level, indicating unplugged port, so that it could switch back to the "ohmmeter" state.

Sure, this would not fix the case of a bad cable randomly shorting out while in use. But maybe it could give it a chance to get past initial crooked plug in of a RJ-45 or a totally shorted new cable ... or new radio?

Unfortunately, Dave never had a chance to reply back since March 11, and I have had to deal almost continuously with a sick family member, so I also didn't have much time to try followups. I have long felt that the burned transformer exposure is a perceived hit to these switches being fully embraced as carrier class ...And I dooo want them to succeed as such and have spent a bit of time trying to come up with ideas around it that also encompass the idea of very low power loss for solar installations.