The TX/RX levels are OK (between -60 and -70) on both cards and on both routers.
One by one the radio links (sets A & B in the scheme) work perfectly. The nightmare is that whatever channels and modulations we choose, when both radiocards are enabled in one router they simply work awful. We got steady 50-60 mbit/s in one direction tcp traffic on both card radiosets, but when we activate both radiolinks they work with 30-40% packet loss and rates drops to 10-12 Mbits at max. Sometimes even if there is no traffic through one of the radiolinks the other set is working badly…
The “funny” thing is that we have at least 3 links that use the same scenario and work perfectly well and another 3 that doesn’t work at all.
We have changed the radio cards with various atheros 5211 (the same from Proxim MP11.a), 5213 and 5314 (3-com PCI)
Any ideas or suggestions are welcome. I am eagerly awaiting your comments on the case
please see my previous posts on spacing of frequencies utilized within the same AP/system, or same rack/location. That may or may not be related to your problem.
Anything nearby or possibly deformation of dish/feed that might be upsetting the polarization? Also what BurstNet says is definitely worth checking into. We have a couple of sites w/4 5ghz 29dBi dishes stacked w/only inches to spare between due to real estate shortages and getting a working combination of frequencies on those was quite a dance. If we were demanding more performance from the links I’m sure we would not be happy.
burstnet - we have tried frequencies from 6100 to 4900, and it DOES NOT affect the phenomena… which is that: once both cards are active, the signal gets f***ed up
dbostrom - as for the nearby anntenas - there are locations that doesn’t have any nearby interference and still get the same problem…
wirdo - and what about the duplexers = what exactly is a duplexer?
Try doing something as simple as feeding your pigtails through thin copper tubing, and be sure to ground the tubing.
This will ensure that crosstalk between the pigtails is greatly reduced within the enclosure.
Also be sure to use good cabling, try seperating the LMR by at least a few inches and do not run it Side by side.
Be sure your LMR is well grounded.
A well grounded shield is exponentially more effective than an ungrounded sheild.
first time I see this thing. Anyway I know pacific wireless has some dual parabolic antenna. Those ones look better (never tried yet) than this one. The specs are not present on the site, and I do not even understand how to connect/use/spec this object. I prefer to have precise docs about what we use in production
Intresting case you have i have a simalary problem i have 4 cards in one P4 PC, That works perfect with noise -101-105 db all in 5ghz band…
But my second PC that are identical with the other one that i have noice -92-96 and the bandwith is bad. i wondering if it have somthing with the PCI bus freq or CPU freq… and the radio cards… I will try a 3 PC for testing
I think, that your problem is related to IRQ-sharing - and not to RF.
When you are using an Intel platform, irq’s are a sparse resource. Therefore remember to disable any IRQ consuming resource you aren’t using. Sometimes, it is possible to change IRQ-usage in the BIOS.
Anyway, always pay attention to the system-resources-PCI dialog in MT. You can’t expect that your system works when you are sharing IRQ between different resources.
