Industry Forums > Wireless Service Providers > ARGH! Ubnt radios freezing!

reply to bburley

Re: ARGH! Ubnt radios freezing!

reply to voxframe
-30C this morning: Nanobridge frozen

My first one: I have almost all Nano's

why do Nano's stay alive when Nanobridge do not

It was at 23 lowered it to 20

We will see tonight

That means no more Nanobridge for me

reply to voxframe
Documenting all of these failure points with respect to temperature is very valuable. Having accurate temps is helpful. These are cheap enough to scatter around as necessary:

»www.ebay.com/itm/New-Digital-LCD···1a5e46ab

A couple of questions come to mind...

It would be helpful to know if both the radio and the POE PS are in the same temp environment, or if, for example, the PS is located indoors. A little more clarification as to the particulars for the device installation (since each install is unique) would also be helpful.

Another thing is the careful use of terminology, since common words might have a literal meaning with respect to the technical issue we are discussing. For example, folks might say "the radio is frozen". Does that mean "like ice" or that data transmission has ceased?

At any rate, good thread...
--
Nothing makes an American want to do something more than telling them they can't.

reply to voxframe
temp will be -30 tonight

ps is inside: link had red light on aircontrol: could not reach radio

when on premises: I could reach radio: had to default and reprogram

looks like rf side dies in the cold

tonight will be a good test

reply to voxframe
Quick and dirty ways to bring back links...

-Lower TX power to 21 or less and it should keep the link stable - ish. (This doesn't always work, but it does help quite a bit if you can catch this before it happens)

If your link is dead, but you are on the LAN side of the radio and can get into the interface...

-Run Airview for a good 10-20 minutes, the wider the sweep, the better it seems. This, I'm assuming, makes the unit work harder and gets the CPU to heat up juuuuust that extra bit to get the link to establish.
-Reboot the radio over and over. Does the same effect as above, but doesn't seem to be as efficient at it, but it will do it.

Once the link IS up, you need to keep it up. Pound the hell out of the link continuously. I actually have a dedicated "Heater" computer set up to just generate loads of traffic on problematic links to keep them rolling. This is a delicate balance, too little and the link drops, too much and the link is saturated and useless.

reply to John Galt
The PoE on one end of my problem link is indoors. The other end is in a wooden box with a thermostaticaly controlled 100W Light Bulb for heat. Not as good as indoors, but certainly better than outside.

I do not believe that the PoE's were of any influence. I also believe that the ethernet ports do not contribute to the problem, although I can't be at both ends at once.

Rebooting (or changing a setting) will bring the link back, but only for a minute. This method gave me communication long enough to reduce the power level on the remote link. I only did this after experimenting with power levels on the local link to be sure that I wouldn't permanently lose communication.

When I successfully lowered power levels on both sides to 21 dBm, the recovery was almost immediate. No heavy traffic or other warm-up factors were needed.

The Nanobridge M5 is in the front of the dish and completely outside. There are three Rocket M900 AP's on the same tower, but they are mounted inside the bottom of the UBNT sector antennas which are also surrounded by RF Armor shields. The AP's did not suffer from this problem.

reply to voxframe
I have a couple dozen BM5 and a similar number of RM2. Not a single lockup in -43 a couple of winters ago. It hasn't been quite as cold since, and still no temp-attributable problems.

I have only a couple of power supplies in an unheated shelter, but those aren't UBNT branded.
--
db

I just started wondering about the length of the cable run. The resistance of the cable would decrease as it got colder, but the change is probably not significant.

If the RF Power Amplifier draws more current as it gets colder, it may reduce the voltage at the end of the cable. If this is the reason, then voxframe's 'heater' computer should make it worse instead of better.

At this point I cannot tell if one side of the link is worse than the other. My cable lengths are 75 and 100 feet.

reply to voxframe
The problem we're having is not pulling too much current as to cause the units to be "under powered" (Otherwise they'd simply reboot or jam entirely).

This is (Our ours anyway) is a case of the wireless radio freezing up and being unable to communicate/stabilize/lock/whatever. The actual CPU, LAN side, and the rest of the unit, including the radio controller, all work fine no matter what. It's just the radio itself that craps out in the cold.

Which is also why when we up voltage it makes no difference, or on short runs it makes no difference.

reply to voxframe
Our first good cold snap last night. 1-3 degrees F (depends on who you ask).. Wind chill put it at -10-15F. We lost a battery back up for 7 hours last night, so a whole site without power, in the cold. Not a single one of them did not come back on line and work properly.

All short runs of cat5 (under 50'), all with 24V power supplies. Some higher current UBNT, others standard UBNT supplies. Rockets, bridges, and a bullet at that site. Power supplies are in one enclosure, with a RB493. Enclosure is insulated, but vented with louvers on the front (from a removed computer exhaust fan) An old purcell small enclosure. The battery backup was in a separate enclosure. So we have 9 POE's and an RB 493ah in that box.

For whatever it's worth, if anything.
--
»www.wirelessdatanet.net

reply to voxframe
Are any of you considering changing out the UBNT for other companies backhauls. Up to this point I had resisted using UBNT for anything other than an hotspot at a customers site. I was getting really close to caving in to the low cost backhaul options they offer on some new projects coming up. I am really glad this thread came up because we had -18 last night actual. I hate chasing my tale on equipment issues. This would drive me insane if I had them implemented and had to babysit them to keep them working.

reply to voxframe
We've had a COLD spell the last few days, today's high is -22 c, lows around -32, w/windchill around -40 to -45!!and things seem to be humming along pretty good.
I'm not putting tons of traffic through either....

Nano's or Nanobridge: I was wrong the other day: mine(NB)froze up around -25C:

I lowered the power on that unit and although it got colder (tonight -30C, no windchill) it stayed up

Why does lowering the power fix that? I would have thought higher power: more heat?

I would guess too much power = heat = melting = water = drips = freezes where not hot enough ????

reply to Anav
With my own personal experience, plus the evidence offered in this thread, I think it narrows down to something in the design in the RF Power Amplifier or perhaps a timing/design problem in a support chip/circuit closely related to the RF Power Amp.

If it was a simple support component, it should have been an easy fix for UBNT, but instead of fixing it, they give the advice to lower the power level setting to 20 dBm. This suggests to me that the solution involves re-design or sourcing a new chip supplier which makes the fix more difficult and therefore far less likely to happen given the small percentage of customers who get cold enough to see the problem.

So far I have found that the magic number is 21 dBm works and 22 dBm does not work. I still need proof that even colder temperatures will require the power level to be lowered more.

I know that this affects the Nanobridge M5, but I can't remember what other models were mentioned that have this problem.

I went back to the beginning of this thread and found Bullet M5 and Powerbridge M5 mentioned with high latency that matches the symptoms.

The "M5" Series must be the common element.

Well speak of the devil, -23 in the breeze, just had a NBM5 link go out, this is its first winter, and is the coldest day/night here to date.

I did save it by reducing the power
--
OptionsDSL Wireless Internet
»www.optionsdsl.ca