Can I downgrade firmware

[jurwiller]

so I still haven't came to a conclusion, but I believe that I can agree that this isnt a firmware issue, at least not in relation to the Ethernet/switch port. However, I have raised new questions ;)

late last week, I selected two micropop poles that appeared to be having the link issue more frequently than the others. On one pole (pole 12 for reference) I disabled PoE on the Netonix, and used the Mimosa power injector instead and used the same switchport on the Netonix for network access. This seems to have solved the port bouncing, although the occasional Jabber error is still relieved from time to time, I havent had link bounce once since I did this.

On the other pole (Pole 18 for reference) , I replaced the Netonix with another Netonix running 1.4.7 instead of 1.4.9. This did not resolve the Ethernet bouncing issues, but curiously enough it does appear to be happening less? at least I haven't had it go down and stay down on 1.4.7, just bounce.

I noticed on the newer firmware there is now a "calibration" for the power supply; how should I be tuning this part? All my switches are WS-8-150-DC's that are fed by a powerstream PST-SP48-150 that also charges a string of 4, 12v SLA batteries. the input voltage to the switch is usually around 52v.

has anything with the PoE side of things changed recently? is that anything that is touched by firmware? I suppose I could try an AC version I have to see if it doesn't like my DC feed? any ideas or suggestions?

[Julian]

Check the gland on the Mimosa's rj45 jack. there's a pretty well known issue with them as they age, which sounds about like what you're running into.

they "are not" user serviceable, meaning they can be replaced, but there goes your warranty, if you had one - Really not a tough fix.

PoE is 'passive' on our devices, meaning that firmware has very little control over the voltage present - so no, nothing changed on that end either.

Sensing is calibratable, meaning you can measure input voltage at the posts and type that in for the cal value - won't affect anything on the output side of the system.

[jurwiller]

these are brand new A5-14's. I bought them Oct 2017, and they just got put up on the poles March 2018. what is the issue with the gland you speak of? pulling the cable out or what?

[Julian]

I hear they have slop in them and the cables start backing out - This is from a wisp talk facebook group discussion dated 10/19/2017. I can't link externally, but searching 'mimosa gland' within the group's main page will give you the highlights.

From what I read, wrapping the cable in butyl mastic tape and sealing it to the body is sometimes the fix.

[jurwiller]

After investigating logs further, on the pole that i replaced the Netonix with a switch with 1.4.7; when i replaced it I had pulled a backup config of the switch running 1.4.9. With the problems I was having I had the Ethernet port set to 1G instead of Auto. it continued to bounce the rest of the day, and into Saturday until I changed it to 100M-Auto. It bounced a few times on that setting too, so I then changed it back to how the switch originally started with issues to Auto and it hasn't bounced since. This is with 1.4.7, and Mimosa 2.3.3. Mimosa 2.3.3 was still bouncing on auto setting when on 1.4.9. I dont know what to make of this!! I still occasionally have jabber errors count on the interface, bouncing or not but does not appear to be affecting performance.

[sirhc]

There is nothing between any of our firmware versions that affect Ethernet link speeds other than in new version we give you more options like 100-Auto.

The switch is based on a switch core (VSC-7427), all we do is allow you to make config changes to the core. All the switch core functions are locked in the core and we do not write that code.

Changing the link parameters such as AUTO, 100M-AUTO, 1G put limits on the port speed negotiation.

If you lock it to 1G then if any noise occurs on the line or the signals drop below threshold to support 1G modulation then the link drops which is why you always want AUTO which allows the link to reduce speed for ms during noise events.

Every time you make a port speed change or reboot or do anything to the port that causes Ethernet link to drop the l2 devices renegotiate link speed - roll the dice on a link with poor cabling or connections or noise on the line and based on the moment that the devices decide on speed is based on the cable noise at that moment in time - RANDOM

Things that reduce signaling on the cable (just like wireless links)
Connections and crimp ends insert signal loss

Cable kinks and sharp bends insert signal loss

Ethernet surge protectors insert signal loss and should not be used on "passive" POE switches that use PolyFuses such as Netonix and UBNT ToughSWITCH

Noise on the cable from power lines and RF broadcasts and such

Understand that Ethernet communications must have enough power to jump the magnetic coupling of Ethernet Transformers. There is no electrical connection between 2 Ethernet devices (except for ground coupling with POE which is why we scream over and over again to bond tower ground rods with electrical service ground rods).

By the way ground current traversing the Ethernet cable can cause issues even if the current is too low to fry components it will interfere with communications.


99% of all WISP equipment damage is from ground current (I own a WISP)


Here are some good posts on grounding:
http://forum.netonix.com/viewtopic.php? ... 279#p19279
http://forum.netonix.com/viewtopic.php?f=30&t=1816
http://forum.netonix.com/viewtopic.php?f=30&t=188
http://forum.netonix.com/viewtopic.php? ... =30#p13447
http://forum.netonix.com/viewtopic.php?f=30&t=1429

Read the posts above but the basics are as follow:
Tower ground rods must be bonded to electrical service ground rods HEAVY #2 wire.
I always add 1 or 2 “new” ground rods to older existing electrical services.

[jurwiller]

we know the importance of grounding, everything is tied to electrical ground; and in most cases we are setting the pole for the city and its all new power/ground done at that time, an electrician couples everything and checks resistance to ground on everything.

The pole that I have been running a mimosa injector on for the last couple days I decided to swap A5-14 and the cable today and go back to the netonix without the injector. Cable test is fine, and accurate on length (38ish feet) and the A5-14 didnt stay linked up for more than 20 minutes before it lost link again, and stayed down like I have had chronically. Instead of flopping between 1G and Auto like I had done before I just ran a link test and it immediatley renegotiated at gig.

the switch I swapped out with 1.4.7 on an entirely different pole with the same issue, (swapped nothing but the switch/firmware since I cant downgrade) has been running solid since i swapped that.

I feel like we are talking in circles here - we are not the typical mom and pop wisp (no offense) and I feel I have described our setup accurately and ruled out all other variables. based on the results from two separate poles across town from each other with exact same issues, the only thing that ends up fixing it is either using the mimosa injector or a switch without 1.4.9. I also have 20 other pole locations from last year that dont have this problem at all, this is just something that has came up since deploying these new 19 poles last month. I actually have port bouncing issues on all 19 new poles, its just a few of them that after some bouncing, stay down hard until you giggle the handle.

[jurwiller]

so regardless of firmware, regardless of a good switch chip (microsemi is gooood stuff) could it be anything in the manufacturing process gone awry with these? The 20 poles in service with no issues were bought early 2017 if not late 2016 in smaller groups. This new batch of 19 I am having issues with were actually purchased around October 2017. The switch I used to essentially try a downgraded firmware was a spare that had been on the shelf for months from the original orders early 2017/late 2016.

[jurwiller]

looking back, I don't think i got a clear answer on how to downgrade? If I try to load 1.4.7 on a switch with 1.4.9 i get some sort of folder/directory error. If i need to step it back, or go back even further let me know where I can get the right version.

[Julian]

So, I'm told that using the TFTP recovery instructions here viewtopic.php?f=6&t=409#p2398 will let you downgrade.

I still think you have something else going on that we're not seeing, like Chris said, our firmware has just about nothing to do with layer 2 mechanics. If you do revert to 1.4.7 and experience resolution, please do remain in contact. - Not ruling out some esoteric bug, but I can't see it as likely just yet, hope you understand.