All of my AF5Xs and AF24 are running on the older version of the
GIGE-POE-APC and they work fine at gigabit.  We have MANY.  Just FYI.

On Sat, Jul 29, 2017 at 9:28 PM, Chuck McCown <[email protected]> wrote:

> Thanks, all info like this is very helpful.  GigE is more akin to RF
> engineering than digital.
>
> *From:* Adair Winter
> *Sent:* Saturday, July 29, 2017 8:19 PM
> *To:* [email protected]
> *Subject:* Re: [AFMUG] new GigE-APC issues
>
> If we can feed you any data to help troubleshoot, please let me know what
> that would be. in several places we've had to remove the modules. It only
> seems like it's airfiber 24's and 5x's we ever have problems with. I mostly
> chock it up them and possibly crummy ethernet ports. When we have trouble
> it doesn't seem to matter what cable we use or how long the run is it. We
> have lots on UBNT Carrier cable and recently many on Primus CAT6, doesn't
> seem to make a lot of difference. Removing the modules always fixes the CRC
> errors though.
>
> On Sat, Jul 29, 2017 at 9:10 PM, <[email protected]> wrote:
>
>> It tests perfect on the lab....  but as we all know....
>>
>> Full speed, no CRC or any other type of errors.  Long cables, short
>> cables etc etc.
>>
>> I generally only hear from cambium users.  And not that much, so it is
>> not happening on all of them, but some.
>>
>> I do have 3 components coming that  have better specs that I am switching
>> to.
>> It will improve NEXT and return loss.
>> If that is the cause then that will be the fix.
>>
>> There is also a thing in the POE inserters called the "Bob Smith" or BS
>> termination.
>> There really was a guy named Bob Smith that did the original calculation.
>>
>> There is another guy that has more recently published a different
>> mathematical treatment of the BS termination that uses a different
>> impedance.  I am going to try this and see if it improves things as well.
>>
>> -----Original Message----- From: George Skorup
>> Sent: Saturday, July 29, 2017 4:10 PM
>> To: [email protected]
>> Subject: Re: [AFMUG] new GigE-APC issues
>>
>> I had very few issues with the previous ones. The RJ45 tab facing at the
>> PCB was a minor inconvenience. Sync over power + gigabit didn't play
>> nice with those, but GigE otherwise worked fine.
>>
>> I tested a couple prototypes of the new revision for Chuck. Sync over
>> power + gigabit with a PowerInjector+Sync and a 450 AP worked fine. Then
>> we replaced that AP with a 450i because of the 450's PPS limit. I didn't
>> notice it at the time, but the 450i never came up at gigabit. We didn't
>> really need >80Mbps downlink anyway. It would usually run about 55-60 at
>> night which was where the 450 was crapping out. Well, more like 50 and
>> 9-10k PPS.
>>
>> Now I've got some ePMP GPS radios for PTP (not for sync, just the GigE
>> interface). Same symptoms as the UBNT prism thingamajig. The link comes
>> up at gig, but flaps and constant FCS errors. The 450i just goes, gig,
>> gig, gig, nope.. 100mbps at bootup.
>>
>> Pulling the SS out makes it work, but I can't do that. We lost that 450i
>> AP because someone went and bypassed the SS because they were apparently
>> concerned about the CRC errors and didn't f'ing tell me about it, even
>> after I said I'd debug it on Monday. Welp, a storm rolled in the Friday
>> before and.. there went the AP. Good job! In Steve Jones' words, I have
>> only one comment: fuckwits.
>>
>> I'd say just stay tuned. Chuck is working on it. As he put it, GigE is a
>> different animal. NEXT, FEXT and all that fun stuff.
>>
>> On 7/29/2017 4:20 PM, Matt wrote:
>>
>>> Seeing same issues on Gigabit ethernet ports with Cambium PxP450i and
>>> also SAF licensed gear.  Shorter cable runs help but just not possible
>>> on some towers.  Bypassing the GigE-APC always fixes it.
>>>
>>>
>>> On Sat, Jul 29, 2017 at 12:40 PM, George Skorup
>>> <[email protected]> wrote:
>>>
>>>> I think I've seen this mentioned a few times now. Maybe Steve and a
>>>> couple
>>>> others?
>>>>
>>>> The latest revision (D?) is causing CRC/FCS errors and random link
>>>> drops.
>>>> Taking the SS out makes all problems go away.
>>>>
>>>> Definitely more issues at gigabit. I have multiple sites with some UBNT
>>>> and
>>>> ePMP stuff (no sync over power). Some I've turned off gigabit
>>>> negotiation
>>>> and get only a few FCS errors here and there. Running gigabit sees lots
>>>> more
>>>> errors and adds a 2-3 second link loss randomly which is a big issue on
>>>> a
>>>> PTP link.
>>>>
>>>> Got a couple sites with 450i APs. One that I was working with the other
>>>> day
>>>> is a new radio (because the old one blew up). New cable, new GigE-APC,
>>>> new
>>>> PacketFlux PowerInjector. The link tries to negotiate gigabit, but falls
>>>> back to 100 after a few tries. Turned off sync over power just to see
>>>> and
>>>> still no gigabit. Remove the SS and it works fine.
>>>>
>>>> Floating the shield makes no difference. Running the site on battery
>>>> also
>>>> makes no difference. Some FM sites, some water towers, some elevators,
>>>> some
>>>> regular towers with only our stuff on it.
>>>>
>>>> The stuff I was working with yesterday is a new UBNT Prism PTP radio,
>>>> new
>>>> GigE-APC, new CCR1009. This site also has an Exalt G2-11 fed by an older
>>>> GigE-POE-APC and I get no errors on that. It used to be plugged into a
>>>> MikroTik CRS210 and I didn't see any errors on that either.
>>>>
>>>> I had far fewer issues with the older cards. Unfortunately I don't
>>>> think I
>>>> have any good ones left to do a comparison. I do have a couple that got
>>>> some
>>>> water in the jack, maybe I can clean those up for testing. I've used
>>>> maybe
>>>> 10 of the new ones now and they all seem to be causing issues. And
>>>> before
>>>> you ask, yes, I've tried dumb gigabit switches at a few sites
>>>> temporarily
>>>> and still see the random link loss with the SS in line.
>>>>
>>>>
>>>>
>>
>
>
> --
>
> Adair Winter
> VP, Network Operations / Co-Owner
> Amarillo Wireless | 806.316.5071 <(806)%20316-5071>
> C: 806.231.7180 <(806)%20231-7180>
> http://www.amarillowireless.net
> <http://www.amarillowireless.net>
>
>
>

Reply via email to