Slashdot Mirror
Broadcom Accuses Atheros Of WiFi Pollution
eggboard writes "We just posted a story at PC World about 802.11g chipmaker Broadcom's claims that the high-speed 108 Mbps mode available in rival Atheros's AR5004G chipset disrupts all nearby Wi-Fi networks. The Turbo mode, part of Atheros Super G, uses two Wi-Fi channels (5 and 6) to double bandwidth, but Broadcom says this can lead to 'an enormous degradation in the speed of nearby 802.11b and 802.11g networks.'. D-Link and NetGear are shipping Super G-based devices. If Broadcom is right, Atheros gear would pollute neighbors' networks. If wrong, they're putting out a pretty heavy marketing smokescreen just before Comdex - where Broadcom says they'll be demonstrating the Atheros problem."
Okay, so what happens when two Super-G networks are within interference distance of each other? Do they both drop down to 54Mbps speed, or what?
IIRC, the only regulations for those frequencies are power restrictions. Beyond that, you can do whatever you want. Even if it breaks everything else.
It may be that the FCC would consider two seperate broadcasts can only use half the power each, but I kind of doubt it.
autopr0n is like, down and stuff.
I live on a suburb with several (20+) radiomodem access points - no cables, no ground lines, just several "clusters of houses" hooked to their antennas. Connections to several ISPs.
The network connection sucks.
At first it worked fine. I'd say it worked great. People heard it works great so they began installing the equipment themselves. The lines began disturbing each other, but it still worked okay. More people installed this, and the network quality began to suck really. So some of them, to overcome the noise, installed signal amplifiers for their antennas. Result? Everyone without amplifiers simply lost their connections completely. So people began installing amplifiers en masse, which resulted in that connection sucks for everyone again. My packet loss ranges from 10 to 60%. TV signal gets disturbed. Radio mice and keyboars don't work. Great, just great. And the ISPs just can't come to agreement on putting one, good, shared ground line.
Ether is a limited resource and wireless in larger amounts will suck, no matter what.
45 5F E1 04 22 CA 29 C4 93 3F 95 05 2B 79 2A B2
Broadcom *do* produce OSS drivers - just not for their 11g stuff. I'm told that's due to patents (not sure how true that is, though).
They also released the source to their DOS configuration utility to their network cards (the files had 'copyright broadcom' and 'not for redistribution' written on them, however).
I don't know about Atheros, but ever since my 2-floors-down-neighbor put up a D-link "b" access point, I can't connect to my bedroom (g) access point from my own living room, even though I've separated the channels (2 and 11). I am forced to connect to his.
Does anybody test for interference with these damn things?
With 12 more channels as of this past week, why would one fight over the crowded 2.4 GHz band? Unless one doesn't have a product for the less crowded 5 GHz band.
Normally I'm inclined to agree, but not in this case. Having worked for companies that have dealt with Broadcom, and talking to some of their engineers myself, I know a bit about the company itself.
Their engineers are very smart people (most of them hold degrees above the typical Bachelor of Science), and I've seen their source code (for cable modems), and it's well written. Their spec sheets (again, for their cable modem products) are generally top-notch, although I'm a bit miffed at them about the whole Broadcom 3415 tuner chip issue (with the patent infringement suit against Microtune), and the whole Linksys/Cisco situation. I'm also not keen on their interview processes either (myself and a couple of my former co-workers have all interviewed there in the past), but that's not related to this issue.
That aside, I'll address the statement from you that I copy/pasted... If the hardware sucks, having GPL'd drivers for it is of no use (at least not to me). If the Atheros chipset is causing interference (and I reserve my judgement until I hear more about the issue), then as far as I'm concerned, I'd rather buy other hardware that actually works.
-- Joe
Oh, yah, I forgot to discuss C.
All the consumer 802.11* products today are half duplex. That is, they can't transmit at the same time they're receiving, and only one party can talk at a time while still allowing the signal to be demodulated successfully. (Vivato is doing some really neat work with phased arrays to receive from multiple people at once, but that is high-end $10K+ hardware).
Channel arbitration, or deciding who can talk when, is expensive, and isn't perfect. A certain amount of the time, two wireless transmitters will decide to talk at the same time.
802.11 uses a CSMA/CA protocol to prevent this "doubling". CS is Carrier Sense-- side A can tell when the other side is transmitting.. but in this case usually only when side A is not transmitting itself. MA indicates that it is a multiple access channel. CA=collision avoidance specifies that a protocol is involved where one side transmits for a little while, and then stops and checks to make sure there's no one else transmitting.
The amount of time needed to spend on the test carrier is related to the propagation delay between stations-- in other words the speed of light and distance, as well as the switching times from transmit/receive and the amount of listening time necessary to determine if someone else is transmitting, as well as various other probablistic answers (the odds of someone else both starting to transmit and also checking for someone else at the same time, for instance-- adjusted for propagation delay, of course).
All of this arbitration, waiting, etc, wastes bandwidth, and thus lowers real world throughput.