HUGE problem with MT

wirdo · Mon Nov 22, 2004 10:58 am

hi all,
i am using mikrotik for quite some time now and i am familiar with most of its downs and ups, but this one is no match for me. I have setup of 4 ap's half of them bridged another 2 routed and all of them connected wirelessly. Computers used are 1.7 ghz celerons with 256 ram, 64mb flash (i always use systems like this one and they work), cards are dlink ag520 and cm9's. Links are 23, 35, 37 and 20 km. Dishes are 1.2 meters parabolic with custom feeder. Cables are not more than 3 meters. The problem is that signal strenght on all 3 of 4 links is jumping all around and huge ping times occur and slow throughput. Signal goes from -69 to -82 with clear LOS. Never happened before so i cant understand whats going on. I replaced all antennas, cables and feeders thinking that they are fault part of the system, but nothing changed. Then i replaced dlink cards with cm9's (ar 5213) and the problem still there. Now the only thing i can think of is fault motherboards or bad power to computer, but since we use UPS'es power may not be the problem.
Please if someone had this problem before and solved it somehow, tell me how to fix that mess.
Thank you!

stephenpatrick · Mon Nov 22, 2004 11:08 am

Just a few thoughts off-the-cuff:

- pigtails to radios - changed them? often source of problem

- Fresnel zone clear?
- Links over water at all?

- What timescale the signal changing - jumping up and down rapidly, or slower changes? same effect both ends of a given link?

- any other radios in the same area (done a survey?)

Regards

Stephen

jarosoup · Mon Nov 22, 2004 7:01 pm

Sounds like interference to me. Have you tried changing channels around?

wirdo · Mon Nov 22, 2004 9:09 pm

thanx guys for suggestions.
jarosoup > yes i tried and on one of the links only 5210 ghz channel is working so-so with 72-75 signal. When trying to change close to 5300 signal goes up to -85 and sometimes loosing connection. weird

stephenpatrick > yes the cards were all dlinks at first, then after this strange behaviour we swapped 4 of them with cm9's - was the same thing. Pigtails we do it ourselves - soldered directly onboard on dlink's. Fresnel clearance is great (mountains) and no water at all. Signal is very very weird - no slower changes are detected - it can jump from -60 to -85 right away. Another thing is 10 db difference between rx and tx on some of the links.

ccrum · Fri Nov 26, 2004 6:31 pm

How solidly are the antennas mounted? Are they dishes? If so, keep in mind that dishes have very narrow beamwidths. The bigger the dish, the narrower the beamwidth. If the dishes are moving in the wind, you could lose signal very suddenly. With even a single degree of movement. I woul dtest the two units in a lab environment. If it truly is the board, then you will see the same problem even at close range. If it is something else, then at least you'll have eliminated one possible problem.

Cameron

marksx · Sun Nov 28, 2004 3:10 pm

Maybe base station atheros is changing output power...
on turbo mode (weak signal - low bandwidtch ex 12 - 18Mb @ -85)
power of atheros both turbo and without, are ~ the same.
if one of station will negotiate higher speeds ex. 48-54 MT will lower transmit power... this is ~ 10-12dBm lower (on mikrotik scan)
try use pooling clients or manualy set max speed reasonable is 12-36Mbps

maybe try to set manualy tx power (this is dangerous but..) value 17dBm I think is safe. Then write signal strengh on station's.

I'don't remember that you say about base station antenna...
37km with omini is loooooong distance...
Just say model and producent...

dwright · Tue Nov 30, 2004 12:58 pm

I have heard lately that there are problems with some of the 5Ghz pigtails. The genuine Horise pigtails are supposed to be the way to go. You can get them from one of these guys listed below.

Quotes from one of my mailing lists.......

I have to second Larry's motion. I tested quite a few pigtails from various
vendors using some very good HP Network Analyzer equipment and I saw about
1dB variation at 2.4GHz and up to 10dB variation at 5.8GHz between "good"
pigtails and "bad" pigtails.

but a lot of folks seem to like
pigtails from Roger Peters <rpeters@world-net.net>.

I have had good results with Hirose pigtails with those from Rob
Flickenger <rob@metrix.net> and Matt Westervelt <matt@metrix.net> (or
just sales@metrix.net).

Dan

yogi · Thu Dec 02, 2004 12:30 am

CM9 on both sides. What frequency? Must be 2.4 with "Signal goes from -69 to -82 with clear LOS" and the distance quoted.

What version of MT? How is the data rate set? These links worked before you did what? The difference from side to side is(-69/-82 side A side B X/X?

yogi · Thu Dec 02, 2004 12:44 am

[quote="wirdo"]thanx guys for suggestions.
jarosoup > yes i tried and on one of the links only 5210 ghz channel is working so-so with 72-75 signal. When trying to change close to 5300 signal goes up to -85 and sometimes loosing connection. weird

pays to read: adjust your datarate basic/supported to 6 megs to see if the latency settles. 2.8.18 non-legacy works ok with CM9's.

Adjusting the transmit power does Jack.

From 5100/5200 to 5700/5800 I lose 10 db's. I thought it was the antennas we use, seems you have a similiar problem. What db antennas give you -69 on 30+ k's?

wirdo · Thu Dec 02, 2004 1:45 am

some of you seems to me read last 2-3 lines of message and post replay (yogi)
anyway ar5212 works better than ar5213 with current drivers.
I always use rate set configured and limited to 36mbps on connections with NSTR enabled and more than 10km. 6 mbps supported is not an option since there has to be stable link at 36
in comparison one of my links is 45km 1.2 m solid dish on one end and offset 90cm on other with -68 - -70 signal works great.

Thu Dec 02, 2004 6:34 pm

We created a fix for the Atheors 5213 chipset cards that has slow performance. This is a test package and it doesn't contain the Winbox configuration support. Please report to support@mikrotik.com if it fixed the problem.

http://www.mikrotik.com/download/wireless-2.8.19.npk

dwright · Thu Dec 02, 2004 9:18 pm

From 5100/5200 to 5700/5800 I lose 10 db's. I thought it was the antennas we use, seems you have a similiar problem. What db antennas give you -69 on 30+ k's?

The pigtails that some of you may be using are only rated to 3Ghz.
This might explain why you are seeing a 10db drop from 5100/5200 to 5700/5800. The higher you go up in frequency the farther you are going away from the rated frequency of the connector(UFL), which causes your drop in db. It is worth a try to get the "Genuine Horise Pigtails" from one of the suppliers in my last post to see if that makes a difference.

Dan

wirdo · Thu Dec 02, 2004 11:53 pm

uldis, could you import this fix to wireless nstreme package - i only use this one for longer links, thanks

yogi · Mon Dec 06, 2004 12:53 pm

We created a fix for the Atheors 5213 chipset cards that has slow performance.
http://www.mikrotik.com/download/wireless-2.8.19.npk

I am hesitant to just adopt these drivers without more info. After loading them to 1 repeater with only 2 clients, there is slight signal gain on the client using MK with similarly updated drivers. The other client using MADWIFI drivers had no change at all. The repeater and client were previously using 2.8.18 wireless (regular not legacy).

My whole network consists of CM9's (5213), and would like to take advantage of any improvements. There was supposedly a timing issue on the CM9's with MADWIFI, but was resolved. If nobody can provide more insight I will try them on a longer backhaul with a more densely populated distribution.