I wanted to recount a recent situation that we put a lot of time into. I hope this helps some other folks out along the way.
We recently put MT gear up on a 300’ tower. We had the following setup:
- PtP Backhaul with RB333 / XR5 / 28dbi Solid Dish w/ Radome
- 5 GHz AP with RB333 / XR5 / 90 deg 5 GHz Sector
- First Client: RB133 / XR5 / 26dbi Grid Dish
The PtP backhaul worked flawlessly from day 1 and we are able to push 60+ Mbps UDP over the link.
We initially put the client up and were able to connect and pass good traffic, -64 / -64 and good CCQ, so we left the site thinking all was well. About 6 hrs later, the client dropped off and would not remain connected. The client would actually connect briefly and drop on and off every few seconds. At this point, we attempted a variety of tactics in an attempt to get the client to reconnect and remain connected including:
- Changing frequencies - every single frequency we tried provided the same behavior
- Changing client hardware - we tried a variety of hardware configurations, antennas, routerboards, cards, etc - nothing helped and the behavior remained the same. When changing to new hardware, we could sometimes get it to connect for a period of several hours only to have it drop off and not reconnect.
- Using the frequency scans in an attempt to detect noise - all channels reported noise floors of -95 to -103 - in addition, every frequency we tried (and we tried them all) provided the same results
- Fiddling with wireless settings - we tried almost every configuration of wireless settings in RouterOS (NStream, no NStream, 5GHz, 5GHz Turbo, 5GHz-10MHz, 5GHz-5MHz, etc, etc.) no amount of fiddling changed the situation
We worked on this for about a week on and off trying everything we could think of. Even though the noise floor on both sides was always reporting -100s, I still suspected that we might be dealing with some sort of wide spread noise. Our next approach was to attempt to swing the 90 deg sector to the north away from part of the city that we were overlooking, but still keeping the client within the 90 deg.
When we swung the 90 sector so that the city was out of it’s view, the client signal remained steady at -64 / -64 and the noise floor remained at -100s and the CCQ remained in the upper 80s. No indicators seemed to change, but immediately after swinging the 90 deg sector to the north, the client connected and has not dropped since for around 4 - 5 days and we are able to pass full rate traffic. We are able to connect the client on almost any channel now also. There was a clear issue before swinging this antenna that disappeared after we re-aligned it.
It seems that the swinging of the antenna somehow made it deaf to some sort of undetectable wide band noise that was affecting every single 5 GHz channel.
Even though there are some tools in MK that are supposed to allow you to detect noise in the area, they are really not effective at all in some situations. I wonder what a spectrum analyzer would have revealed in this situation. I suspect that we may be fighting some similar noise in some other areas.
Hope this is interesting to some folks. If anyone has any sure fire way to detect noise on a sector, that would be helpful.