OpenWrt Forum Archive

Topic: "The WRT Radios are high performers:" fact or fict

The content of this topic has been archived on 1 Apr 2018. There are no obvious gaps in this topic, but there may still be some posts missing at the end.

I'm swimming in WRTs now doing a lot of things with them.  One area of weakness appears to be the radios, when they are used in a WISP client (CPE) environment.

Side-by-side with an alternative method (266Mhz laptop, Hermes-based Orinoco PCMCIA car) the WRT consistently has come off giving weaker reported signal quality values, and also performs more poorly on iperf traffic tests, by a fairly significant margin.

Maybe the radios have some limitations inherent in their design (e.g. fixed ACK timeouts set to a short roundtrip time) that prevent them from performing well in this environment.

I don't know how much I can trust wl to be truly telling me about my connections.  The "status" command will stick at some quality value and then never change, while at the same time I can see the noise and rssi values changing using the individual wl subcommands.

And when I sample the radio continuously the values I see flucuate wildly, even when I deploy them in areas where I know I'm very RF-clean.

I'm not complaining, mind you, but if I could get them going in that sort of mode it would be a major win for me, as they perform so beautifully in so many other ways.

I'm wondering who else might have experience in this arena.  I've read the mesh stuff.  My environment is vanilla 802.11b infrastructure ISP.  The clients are interspersed at distances of a few blocks to a few miles from the access points.

I'm glad to post more of my experiences if they're of interest.  I have several working under load now for some weeks but it hasn't been a smooth trip.

Hi Brian

I found the same thing, and I've heard before that the Broadcom wireless chipsets (which the WRTs use) don't have great receive sensitivity.

I have a 12km link (via a 24dBi antenna on my roof) for which I originally used a Cabletron RoamAbout (hermes) pcmcia card.

The RoamAbout reported -85dBm which was enough for a 5.5Mbit link.  I replaced it with a WRT54GS
which reported -89dBm, locked itself at 1Mbit, and only gave me 15-20kB/s throughput.

I now use a Senao (aka Engenius) AP/client which gives me a nice solid 11Mbit link.

I use the WRT54GS behind it to run frottle and routing (BGP). In fact I now have another link with another Senao and use the same WRT for routing both - by putting one of the LAN ports on its own VLAN, effectively giving me two "WAN" ports.

Still cheaper, more reliable and above all quieter & cheaper to run than my original PC + pcmcia cradle + roamabout setup.

Over shorter distances the WRT works well on its own, but if you want to do long distances, use something like the Senao for the actual link, and treat the WRT as a cheap linux router.

FreeNetworks has a list of Receive Sensitivity levels for 802.11b for various chipsets. It doesn't mention the Broadcom, but it seems to me the Broadcom needs at least -80dBm or so for 11Mbit

From my experience:

Previously i used a cheap MR814v2 Netgear router which I dumped because the low NAT timeout was a big problem for me (SSH sessions, etc.)

With my Notebook (old Airport card in an IBook) the Gnome Wireless Link Monitor reported a Signal Strength of about 92%. With the new WRT54GS a signal strength of about 87% is reported (with the WRT54GS the strength seems to be unstable - sometimes it is reported as 83% sometimes it goes up to 90%.). I didn't modifiy the tx output with wl, but use the default values.

(Of course I'm still happy that i now use a WRT54GS/openwrt, because it works much better than my old router, and the signal strength is still strong enough to cover all the locations where I might want to use my notebook.)

/gst

It's a Broadcom card, they're not renowned for their signal reception capabilities...

The discussion might have continued from here.