I have a wireless bridge that I drops for no good reason.
We're using RB600A's on both ends with MT Mimo cards (2.4) and two 12dB antennas on both sides.
The radios are only about 1000' apart.

Stats:
signal strength: -53
TXRX CCQ: 100%
TX/RX rate: 300/300Mbps

The stats move a little but basically they are in the pocket.

The problem is that once every hour or two this link just drops.
It comes back in a few seconds, but this causes some problems.

What could cause this?
My only guess in interference but I don't really see any.

What could cause this?
My only guess in interference but I don't really see any.

The only way to rule out interference as the cause is with a spectrum analyzer, frequency scans/usage on wireless gear are only of use to indicate how much 802.11.. compliant equipment is on that band and will not reveal any interference or non-compliant gear that maybe also present ?

I also get this on 1 link.

RB532A+R52 <-> RB433AH+R52

Differening vintages of ROS each end.

WDS seems to be the culprit on that particular link.

Are you using WDS as well ?

yes, I'm using WDS. I have router OS v4.1 on both sides using RB600A's.
I never thought that WDS could be the culprit. Why?

I dunno, but my link is WDS and -57 both ends with zero interference.

Drops quite regularly, and always has, but it's just for my house, so never really had a need to investigate it a lot.

The same problem on 5 links with different distances (1 km - 40 km), with different cards (R52H, XR5, Alfa AWPCI085h), with zero interference. All links withe Nstreme and WDS. Great signals (-65, -70), great bandwith (28 Mbps for 40 km) but badly unstable - few disconnects in day. Usualy it tooks about 25-30 seconds, but sometimes (fortunately rarely) such disconnects may continue several hours. I tried differnet versions of RoS from 3.30 to 4.10 and I know that 5beta cannot resolve this problem. I have only one stable link without such disconnects. The only difference from the rest ver. of RoS 3.22 on one side and 3.25 on another. I think this will be my next step to try.

The same problem on 5 links with different distances (1 km - 40 km), with different cards (R52H, XR5, Alfa AWPCI085h), with zero interference.....

To rule out Interference then as i mentioned previously is to use a sprectum analyzer, till then you cannot know for sure if you have zero interference?

..... All links withe Nstreme and WDS. Great signals (-65, -70), great bandwith (28 Mbps for 40 km) but badly unstable - few disconnects in day. Usualy it tooks about 25-30 seconds, but sometimes (fortunately rarely) such disconnects may continue several hours. I tried differnet versions of RoS from 3.30 to 4.10 and I know that 5beta cannot resolve this problem. I have only one stable link without such disconnects. The only difference from the rest ver. of RoS 3.22 on one side and 3.25 on another. I think this will be my next step to try.

I had cpe's losing association intermittenly one evening and on the AP freq scan usage and signal to noise was almost the same as the other channels,i moved freq and problem disappeared?

Differening vintages of ROS each end.

The only difference from the rest ver. of RoS 3.22 on one side and 3.25 on another. I think this will be my next step to try.

To eliminate issues which may occur with using differening vintages, versions use one only version of ROS on the network?
I am using 3.30, nstreme but not using WDS, routed network with almost no bridging, which is working very good for me

I used different frequencies with the same effect, I scanned air with REAL spectrum analyzer, so if I say "zero interference", that's realy what I mean. All my unstable links working with same ver. of RoS (3.30 with 3.30, 4.5 with 4.5, 4.10 with 4.10) and only one link, which has differnet versions of RoS on each side, extremly stable (there wasn't disconnects at all! Besides this link was installed more than year ago).

In my network I have a situation when I must use bridging instead of routing and must use wds to be able use some protocols of layer 2.

I used different frequencies with the same effect, I scanned air with REAL spectrum analyzer, so if I say "zero interference", that's realy what I mean. All my unstable links working with same ver. of RoS (3.30 with 3.30, 4.5 with 4.5, 4.10 with 4.10) and only one link, which has differnet versions of RoS on each side, extremly stable (there wasn't disconnects at all! Besides this link was installed more than year ago).

In my network I have a situation when I must use bridging instead of routing and must use wds to be able use some protocols of layer 2.

Good to hear you have used a spectrum analyzer as a lot do not, as mentioned i use V3.30 everywhere and not like in case (3.30 with 3.30, 4.5 with 4.5, 4.10 with 4.10), not saying this is the issue but unfortunately only one way to eliminate ?

And i use the same radio cards, UBNT XR5 for links

Have you eliminated the possiblity of electrical interference at each location

Is there a time frame when disconnects occur or is totally intermittent

Disconnects are totally intermittent.

Disconnects are totally intermittent.

I assume you have spaced apart on each mast sufficent distance between any AP and PTP and what type of antenna's are used

Drops quite regularly, and always has, but it's just for my house,

It seems to dro

p about 3 times a day for about 30 seconds.

Enable wireless debug logs on both sides and see what they say

In my case sometimes this is "extensive data loss" and sometimes "too many poll timeouts".

Looks like I found solve for my problem. I decreased frame limit value from 3200 to 3000 (framer policy = best fit). Bandwith of my link also decreased aproximately for 2-4 Mbit/s, but stability have increased. I also did the experiment: with "framer policy = exact size" and "frame limit = 3200" my link kept no longer than 5 seconds. Can it be the problem with recieving big frames? Link I'm telling about has -68/-70 dBm signal strength and CCQ 100/96 %.

Have you tried to increase the HW Retries from the default 4 to, say, 8 or 10...?
This simple change stabilized my links to a 99.9%!

Cheers!

Changing HW.retries and disconnect timeout was the first thing I did for increasing stability. It decreased disconnections from several times in a day to one.