RB333 2.4GHz Half-Duplex Issues

I have been using Routerboards and MikroTik OS for about two years with very good results. I have mainly used RB532 boards and R52 cards for connecting outlying offices to a central location. These outlying offices only have a couple IP phones and computers. Because these have performed well I decided to try and create a backup link for our main link which is currently being handled by Western Multiplex (Tsunami) 5.8 GHz, 11Mbps full duplex equipment. The main link is 20 miles in distance from the central location to our main office. At the time RB333 were the fastest Routerboards so I bought a couple of those with some R52 cards. I was aware the R52s may not be powerful enough for 20 miles but I decided to give it a try, especially since I am using 2.4 GHz 30 dBi Grid Antenna.

I first set this equipment up in a test environment and everything worked fine. I went ahead and installed the equipment on our 100ft towers. These are my own towers, no one else uses them. I was getting a signal to noise ratio of about 25; it fluctuates slightly but not a lot. When I tested the link with my laptop by passing large files in both directions and using a soft phone at the same time, everything was fine. But when I decided to see how it would handle all office traffic it got bad fast; CCQ was all over the place and throughput plummeted. Phones calls were full of static and would cut out. Terminal Server connections would freeze. I replaced the R52s with Ubiquiti SR2 400mW and I saw an improvement in signal strength and signal to noise (up to 35). But, once I put it on the main network it got bad again. CCQ bounces between 10% and 50%, and throughput again plummets.

This main link handles a lot of traffic. We have over 20 IP phones, an email server, web server, sql server, and terminal server all located on the far end. Before I put this up I had wondered if a half-duplex system would be able to handle this type of traffic, even with QOS implemented. I have enabled Nstreme and connection slightly improved but not to what is needed.

Summary:

  • Using RB333 with SR2, running 3.10
  • using 2.4 GHz, tried 10MHz channel and 5MHz channel
  • 2.4 GHz 30 dBi Grid Antenna
  • 20 mile link, 100 ft towers
  • Signal to Noise is about 25, 35 if I use 400mW radios
  • CCQ fluctuates around 70% to 75% (using 400mW radios), better at 10MHz and 5MHz channel width with lite traffic. Heavy traffic brings it down to about 20%
  • tried both enabling and disabling Nstreme, no polling, disabled CSMA

I wondered going into this if a half duplex system would be able to handle this. The 11Mbps full duplex system with no QOS handles it fine. I have configured both bridging and routing, but with no difference in quality. My next thought was to create a Nstreme Dual connection. Is the half-duplex the problem or is there something I may be missing?

Thanks.

Sounds like interference

Interference was one thought I had but I had checked for it and found little. We are basically in an unincorporated area so there is very little 2.4GHz usage, especially at that height. I did a scan and found no one on our frequency. The 5.8GHz and 2.4GHz antennas are over 6ft apart, and the radios are over 50ft apart.

Also, wouldn’t interference be the same with lite traffic as well as heavy traffic. This problem only occurs when more people get on the network. When more VoIP phones are in use and more people are remotely connecting to the servers the link degrades rapidly. This is why I was wondering if I need full duplex or dual nstreme to handle this type of traffic.

What other types of interference could there be?

Sounds like a bandwidth problem.

What is the terrain like? I have a non-LOS 17 mile link with worse signal and decent results before. 19dBi on each end 165’ on one end 140’ on the other.

I’m not sure 100’ is high enough…

Kijomabb wrote:

wow, thats a deep comment.


Sorry Bill, I didn’t mean to insult your intelligence. I was actually referring to the original poster and his whole office full of traffic as well as IP phones over an RF link.

For your problem, I think I would look at getting a higher gain antenna mainly to improve received signal strength and I have run into a few patch antennas that were marked incorrectly as far as polarization. You might try rotating one of them 90º. I’m sure you’re aware of channel spacing, in case you are using multiple radios in any of your boxes.

We have a business that is running 12 viop phones plus regular internet traffic, including some hosted servers they connect to. Hardware is rb333 XR5, 10mhz channel ad works fine.

At 20 miles at 2.4 ghz your freznel zone is some 80ft and earth curvature is 50, your received signal strength should be in around -50. If it is not you do not have enough height to run that link at 2.4 ghz, unless those towers are on high points.

Problems from interference can get worse with higher through puts as more packets are corrupted.

Erik

Ack timing could be the culprit as well. I tinkered with ack on my 6.6 mile link, partially corrected my high bandwidth/low ccq problem (only had a problem during heavy testing, live traffic was never enough to cause a problem).

I enabled nstreme and my ccq is hardly ever below 90%, regardless of the bandwidth across the link.