Hello guys. We recently bought a pair of RB 112 to use on a low volume backhaul to a nearby (2 miles) tower.

we are using CM9, osL4, 2.9.18, 17dbi "backfire" antennas on 5.7.

The backhaul seem to work just fine. I am happy about consistant 4-6ms ping times. However throughput is another story. Running a bandwidth test from one to the other we get an average of 2.5Mbit both/udp and about 3.1 both,TCP. We can't seem to get anymore.

Any tips or ideas for improving this, or is this a resource limitation? We have 7Mbit of bandwidth on the sector we would like to use.

Well, I would ditch the backfire antennas, those things are just about worthless. The pac-wireless 2ft solid dishes are inexpensive, and work quite well (other real dish antennas are quite good as well).

After changing antennas, I would test modulations (data rates), one at a time. If you start to see latency fluctuations (or unexpected registration drops), you have gone too far. I would then back it down two steps from the point you start to see trouble, and lock down both the basic and supported rates. With luck, you may not have to back down from 54Mbps at all.

5mhz and 10mhz channels may be worth exploring as well; as you may be able to get just as much throughput, with a better signal level (+3dB for 10Mhz, +6dB for 5Mhz), and less spectrum usage. This is especially appropriate if you are only trying to get near 7Mbps, 54Mbps @ 5Mhz should easily net you 6.5Mbps real throughput, if you can keep your signal clean and strong enough.

I would not use Nstreme, since the CPUs in those RBs are a bit on the weak side. Disable connection tracking, don't attempt to do any queueing or firewalling, and otherwise strip down your configurations as much as you can.

Also, be sure you are not testing to or from the radios when you do your speed tests, only go through them.

--Eric

I have not tested the rb112 yet, but did you try to test with 2 systems on both ends?

Intel p4 like system -- rb112 -- rb112 -- Intel p4 like system

The rb112 has less cpu power than the rb532 series. (175mhz instead of 333mhz), but I still think that the rb112 is more powerfull than a wrap board or a rb200 series board from where the tests are better than your results.

Regarding the cpu-power; when I do a test on the rb532's I get approx. 10-11mbits both and approx. 17-18mbits receive or send. I'm curious about the result when you test this.

when I tested from p4 to P4, I got 6.5-7mbit both as average. So, an improvement, and pretty much all we will need for this sector. thanks for your suggestions, I will consider different antennas.

I also played around with the rates, and the antenna power. I locked into 12, and manually adjusted the power to 30 at this rate. Things are quite stable now, and fast. A great bargain for the price! These are auctually outperforming my motorola 5700BH10 units.

I hope it will hold!

cheers

We are testing a 112 and can only get around 0.5 Mbps. The AP is a 532 which gives decent throughput to other clients. Atheros chipsets in AP and in 112. We tried b and g modes. SNR is 30. We also tried using a linksys wireless router at close range as an AP, with the same result: lousy throughput.

adjusted the power to 30

Not sure that a great idea, I think default is best.

Your card may work for a while but it may overheat and give you problems

We are testing a 112 and can only get around 0.5 Mbps. The AP is a 532 which gives decent throughput to other clients. Atheros chipsets in AP and in 112. We tried b and g modes. SNR is 30. We also tried using a linksys wireless router at close range as an AP, with the same result: lousy throughput.

OK, we deployed the thing and it does better in the field than it did in the shop. Maybe there's multipath in the shop or something.

It still is not impressive: plain 802.11b, 42 dB SNR, 1.2 Mbps up and down.

We are testing a 112 and can only get around 0.5 Mbps. The AP is a 532 which gives decent throughput to other clients. Atheros chipsets in AP and in 112. We tried b and g modes. SNR is 30. We also tried using a linksys wireless router at close range as an AP, with the same result: lousy throughput.

OK, we deployed the thing and it does better in the field than it did in the shop. Maybe there's multipath in the shop or something.

It still is not impressive: plain 802.11b, 42 dB SNR, 1.2 Mbps up and down.

42dB SNR is way too much, the best results we get is staying around 25-30dB. Also running G on other channels than 1,6,11 also gives lousy performace. I have several RB112's in both P2P and P2MP (RB532AP) links on both 2.4 and 5G and all draw around 10-11Mbit TCP (a little better on 5G) and up to 16-17Mbit UDP.

Funny story, we made a 6km link using 23dBi Panels from CyberBajt. Had like 55-60SNR with SR5's on default TX power and almost no traffic came through. Played around with TX power settings while running bandwith tests simultaneously. Ended up with 13dBm power on SR5's and 25-33 dB SNR (depending on weather) and we push full speed. Also have experiences with 2.4 customers with +40 dB on AP performing like shit until they turned their antennas away! to get a 20 dB SNR and things worked full speed.. I cant explain why it is so, but i guess shouting too loud to each others isn't specially good for communications.. Maybe that's why there's so few women in this business

Just my 5 cents..

/Henrik

We are testing a 112 and can only get around 0.5 Mbps. The AP is a 532 which gives decent throughput to other clients. Atheros chipsets in AP and in 112. We tried b and g modes. SNR is 30. We also tried using a linksys wireless router at close range as an AP, with the same result: lousy throughput.

OK, we deployed the thing and it does better in the field than it did in the shop. Maybe there's multipath in the shop or something.

It still is not impressive: plain 802.11b, 42 dB SNR, 1.2 Mbps up and down.

42dB SNR is way too much, the best results we get is staying around 25-30dB. Also running G on other channels than 1,6,11 also gives lousy performace. I have several RB112's in both P2P and P2MP (RB532AP) links on both 2.4 and 5G and all draw around 10-11Mbit TCP (a little better on 5G) and up to 16-17Mbit UDP.

Funny story, we made a 6km link using 23dBi Panels from CyberBajt. Had like 55-60SNR with SR5's on default TX power and almost no traffic came through. Played around with TX power settings while running bandwith tests simultaneously. Ended up with 13dBm power on SR5's and 25-33 dB SNR (depending on weather) and we push full speed. Also have experiences with 2.4 customers with +40 dB on AP performing like shit until they turned their antennas away! to get a 20 dB SNR and things worked full speed.. I cant explain why it is so, but i guess shouting too loud to each others isn't specially good for communications.. Maybe that's why there's so few women in this business

Just my 5 cents..

/Henrik

Well, funny interesting story.
It's true what you are saying about loud, I'll try some test in my network.
Thanks.