Use Multiple Channels for Faster Wireless Networking

icypyr0 writes "The Register reports: 'Current dual-mode 802.11 'a' and 'b' access points use only one of Wi-Fi's 11 RF channels at a time, with users taking turns. The Engim chipset can 'see' all 11 at once, and can use the three non-overlapping ones (1, 6 and 11) in parallel, increasing total throughput and enabling features to be incorporated in silicon that are usually implemented, at extra cost and performance degradation, in software.'"

Comments

Err... it is a cheat

[arivanov]

Err... My polish is crap, but unless I am mistaken they seem to have used a 500mW aplifier and a 27dbM antenna to boot.

What's next? Sticking it in the middle of Aresibo and claiming half a light year range?

Geek factor 9.3 useful factor 1

[obi1one]

(error 1337 destination page /.ed) Is increasing the range of current wireless networking equipment really what is needed. I know i personally am turned off from wireless not because of the lack of range, but the lack of speed. There is certainly some cool factor being able to get that masssive a distance, but I dont see this as making wireless more desirable to anyone.

Re:Geek factor 9.3 useful factor 1

[_avs_007]

I work with a lot of wireless APs and client cards in our lab.

If you are talking about G type stuff, stay away from Linksys, they have the crappiest range. When I used it in my house, I would get 68db with an Intel 802.11b AP, but the Linksys G router/AP yielded 77db, and that was only going through 2 walls.

I replaced it with a Netgear WGR614, which uses the Intersil Prism GT chipset (as does the D-Link we tested), and got much better range. Similar to straight B. ~68 or 69 db in my master bedroom.

In our office environment, the Linsys G would drop signal after walking past the conference rooms. The Netgear G allowed us to almost walk around the entire floor. I connected a signal booster, and found it to be next to worthless, as it did not improve range. If it did, only by 5 ft or so. It still dropped signal as I walked past the conference rooms.

The measured actual throughput was 4.5mb/s with straight 802.11b, and 21mb/s with the Netgear G.

Quite suprisingly, I had the best results with the Netgear WAB102 Dualband A/B, which is the only A/B AP that uses Atheros second generation A. Tom's Hardware had a write-up on this. Atheros had a whitepaper. I bought 3 of these, and verified the claims.

With a Linksys A+G card, (which uses Atheros 5001X+, as does Netgear WAG511), I got slightly weaker signal strength in my master bedroom 70db), but throughput killed both B and G. I was measuring 24mb/sec throughput in non-turbo mode, and 45mb/sec in turbo mode. In the office, I was able to sustain 7-11mb/sec at the opposite end of the building. The Netgear G was only able to sustain 1-2mb/s. Inside the conf rooms, Linksys G had no signal, Netgear G sustained 7mb/sec, Netgear A in turbo mode sustained 24mb/sec.

In the office, the range of this second generation A actually exceeded that of B, which is something Atheros pointed out in their whitepaper. They said while true A can't go through walls as well as B, the 1st generation A was not performing up to its capabilities. Kind of like how Shannon's law states what is the maximum amount of data that can be carried across wireless, but current technology does not even begin to approach this limit.

I've tested various client cards from Orinoco, to Cisco Aeronet, Prism 2 and Prism 3 cards, and various Atheros based cards. I that the AP affected range more often than the client cards. Though I have found that anything based on the Atmel chipset to be crap. The USB 802.11 card from Linksys (V2.6) uses this chipset. Unfortunately, the Netgear WAB102, uses Atmel for its B, so its B is crap as well. I just use the A portion of it anyways. But the new Netgear triband router, I beleive uses Atheros for all three bands, it just costs an arm and a leg.

The Linksys Triband AP, only uses Atheros for the A, it uses Broadcom for B and G, so its G sucks just like the Linksys B/G stuff.

Somewhere I also read that Linksys will not support any turbo modes in their AP/Routers. (though their A+G client card still supports turbo). Both Atheros and Intersil have planned turbo features. Atheros already had 108mb/s A support in turbo, allowing 45mb/sec throughput by using multiple channels. They already have support for hardware compression, so are promissing a future firmware update that will flip this on, that will allow a turbo mode to sustain 90mb/sec throughput. Its called SuperA. They just released SuperG, which uses multiple B/G channels and compression, to allow 108mb/sec, and I think 45-60mb/sec throughput.

Intersil's turbo technology is called Nitro. Similar (but incompatible) with Atheros's technology.

If we continue to see developments such as this...

[SwansonMarpalum]

It will only mean that last mile solutions will become more plausible for those who don't live within a couple miles of their CO. This is a Good Thing, as having Dial-Up and Satellite as your only options is pretty unbearable.

Partial translation..

[Anonymous Coward]

Not a full translation.. just some of the more pertinent bits..

"Nalez.y podkres'lic', z.e realizowane ?a;cze by?o typowo naziemne (w odroz.nieniu od ?a;cza zrealizowanego pod koniec 2002 roku z balonu stratosferycznego w Szwecji przez firmy Alvarion i Swedish Space Corporation)
Co rownie waz.ne, wszystkie elementy uz.yte w eksperymencie firmy INTERLINE sa; niemodyfikowanymi urza;dzeniami/osprze;tem doste;pnymi na rynku(parabola 1.1 metra i wzmacniacz 500 mW) - a w Szwecji uz.yto anteny parabolicznej o s'rednicy 2.4 metra i wzmacniacza 6000mW.
Na potrzeby eksperymentu wyznaczono dwie skrajne lokalizacje: Wroc?aw i Hala pod S'niez.ka; - najwyz.sza; gora; Karkonoszy - na po?udniowy wschod od Jeleniej Gory.
Odleg?os'c' w linii prostej: oko?o 110 km."
==>
"It should be noted that this connection was ground based (in comparison to the record achieved at the end of 2002 by a Swedish company which used a hot air balloon)
What's also important is the fact that all the equipment used here is unmodified and readily available off the shelf (a 1.1m parabolic dish and a 500mW amp), unlike the Swedes who used a 2.4m parabolic dish and a 6000mW amplified.
For the pruposes of this experiment we used two locations, Wrcolaw and Hala pod Sniezka - the highest mountain in the Karkonoszy - south-east of Jelieni Gora.
Distance in a straight line - about 110km"

"Lokalizacje
Jednym z kluczowych etapow eksperymentu by? wybor lokalizacji dla stacji tworza;cych planowane po?a;czenie punkt-punkt. Pierwsza z nich, to 10 pie;trowy wiez.owiec na jednym z wroc?awskich osiedli. Druga lokalizacja, kluczowa dla ca?ego eksperymentu to hala pod S'niez.ka;, obok schroniska Dom S'la;ski na wysokos'ci 1400 m.n.p.m."
==>
"Location
One of the key decisions to be made was the location of the end stations in this point-to-point link. One station was the tenth floor of a Wroclaw block of flats. The second station was the hall under the Sniezka.. near the Dom Slaski shelter, about 1400 meters above sea level"

"Sprze;t
Do przeprowadzenia eksperymentu wybralis'my naste;puja;cy sprze;t:
* Anteny - PARABOLIC maxi, 27 dBi - produkcji INTERLINE
* Punkty doste;powe - INTEL Pro/Wireless 2011 Access Point - produkcji firmy SYBMOL
* Kable i konektory - kable BELDEN H-1000, H-155, RG-316, wtyki VITELEC
* Karty radiowe - Lucent ORiNOCO PC Card Silver/chipset Agere, ZCom XI-300/chipset Intersil
* Wzmacniacz - 2.4 GHz o mocy 500 mW"
==>
"Equipment
For this experiment we chose:
Antenna - a 'PARABOLIC Maxi 27dBi' by INTERLINE
Access points - INTEL Pro/Wireless 2011 Access Point - manufactured by SYBMOL
Cables and Connectors - BELDEN H-1000, H-155, RG-316 cable and connectors by VITELEC
Radio cards - Lucent ORiNOCO PC Card Silver/chipset Agere, ZCom XI-300/chipset Intersil
Amplifier - 2.4GHz at 500mW"

Damn.. I'm bored.. anyone else wanna finish this?

DonP.

Re:110kms? The world record is already 310km.

[MikeHunt69]

acknowledged by Guinness (as in world records not as in beer).

The are actually the same company. (seriously)

Re:Damn measurement standards..!!

[RabidMonkey]

DAMMIT!

apparently pressing ctrl-enter submits ... same as how I submit tickets .. stupid habits die hard.

lets continue that.

Sears tower = 442m
so we have 248.8688 sears towers

Boxcar = 43 feet avg (source)
Avg freight train length=45 cars (some other site that won't load but is cached)
Avg lenght in feet = 1935 feet

google tells me theres 0.3048 metres in a foot so we have the avg freight train being 589.788m long.

That means we have 186.5077 freight trains (not counting engines) end to end


length of football field (cdn) = 100m
length of football field (us) = 109.1m (source

That leaves us with 1,100 cdn football fields, or 1008.2493 american fields.

a us dollar bill is 156mm long (source
so that gives us 705128.2051 dollar bills, end to end

asian elephants can grow to 340cm (3.4m) (source)
so thats 32352.9412 elephants

I think thats good enough for now. back to work.

Re:Any ideas

[mustangsal66]

If you have a big fat wallet. Looking either Breezenet gear ($1000 US per radio) which freq hops for security but at 3MB transfer.

For only around $10K, you can get Tsunami gear 100Mb PTP. It runs 5.4Ghz, just don't stand infront of it.

Wirless side note. I work with last mile wireless gear. It's cool, but also on our tower is XMRadio. These fookers run at 2.478 (Yeah they don't bother to tune their antenas much). Do the math here... 200Watt radio, 12-14DB gain antenea. Licensed at 200Watts at the antenea. Our spectrum analyzer pulled them in as the stongest signal WITHOUT and antenea. Our RF tech figured out they were running at about 1800Watts at the antenea.

The good news is there are no birds nests on the tower at all.

So come on down and have a cookout.

no trimode?

[MoFoQ]

no 'g'?

but it's a good idea.
Hopefully, other chipset makers (TI, etc.) follow suit, which in turn will reduce costs (thx to competition).

Of course, if it can do all three, 'b', 'a', 'g' separately, and each with the parallel-ness, that would kick ass.

Re:overlap?

[pe1rxq]

Hadn't thought about that...
We have 1 to 13 here in the Netherlands.
(Besides its not that hard to use all fourteen)

Jeroen

Re:Firmware

[The Clockwork Troll]

The immediate benefit to legacy clients is that they can be partitioned into 3 channel groups, each with its own full allotment of 802.11g bandwidth.

You would only need a hardware upgrade if you wanted each client to be able to make use of multiple channels simultaneously and reach that 50Mbps throughput figure quoted in the article.

Otherwise it's a solution for reducing bandwidth contention in heavily trafficked networks (and protecting 802.11g users from bandwidth degradation by 802.11b clients, as mentioned).

Re:overlap?

[jerde]

Are 1,6, and 11 the only channels that don't overlap?

You can get away with using four without much problem. I use 1, 4, 8, 11 for my wide-area 802.11b network.

You have to plan out in 3 dimensions when you have multiple access points like that. Often the strongest signal available to a roaming user is above or below them, rather than on their floor.

With only 3 channels available, it's too hard to map them out. With 4 you can at least guarantee that no two adjacent access points are on the same channel.

- Peter

Re:Valid Uses

[LostCluster]

Large old office buildings that arn't wired for ethernet, large warehouses, and people who live on large plots of land.

Common bond: Areas where for the range of WiFi, you control nearlly all of the territory. Therefore, you're sure you're not getting in the way of anybody else because there's nobody else arround.

A densely packed dorm or appartment house is not the place to do that. It's to the point that in 2-family dwellings, the families need to agree on a consumer bandwidth sharing plan between each other so they don't buy devices that'll collide. Something like "You go on channel 3, I'll go on channel 10 for WiFi, and we'll reserve 900MHz for phones so don't get a 2.4GHz one..."

Re:Tragedy of the commons forming!

[bbdd]

mod up the parent. this is a big problem.

i live in a tightly populated suburban area and i can see 8 or 10 APs immeditately around my home using simple tools like netstumbler. this doesn't count APs that don't announce their SSID, as well as who-knows-how-many 2.4ghz cordless phones.

my 2.4ghz cordless phones get interference all the time from what i can only assume are everyone elses APs and phones around me.

simultaneous multiple channel use is a bad idea, unless you are out-of-range of others...

Re:overlap?

[pe1rxq]

Thats what you get with unregulated pieces of spectrum, everyone can put its crap on it....
You could use a spectrum analyzer to monitor 2.4GHz band to see if anyone in your neighbourhood is using it.

Jeroen

tradeoff

[l33t-gu3lph1t3]

While combining channels may be a somewhat decent idea, it has a serious drawback: you lose your usable channels! I believe that maximum feasible number of users on any one access point is about 50, right? That's on channels 1,6,11. Sure you get more speed by using just one channel and sharing the resources of all 3, but hell, now only about 17 people can use the network!

I can see the FCC loving this.

[Halvard]

The licence that 802.11, A, B, G, etc. falls under says that if you inferfere (and you are discovered to be the culprit), you can't use it. I'm paraphrasing but I suspect that if you saturate the bandwidth available in the frequency and you get outed, then you'd have to stop using at least all the frequencies or the offending radios.

Department

[BandwidthHog]

From, uhh, *what* department?

Ahem.

Re:Department

[BandwidthHog]

Holy shit people, who did I piss off with that inane comment to earn two Overrated mods? I mean, c'mon, some doofus actually burned a mod point on a comment already at 0? Wow. Must be the sig.

Dlink has a similar idea

[FigWig]

Dlink's AirExtreme G products advertise speeds up to 108 Mbps. Their literature implies that they use two wireless channels to achieve this speed.

I have their wireless card & router, but can't get the 108 speed because of some legacy b adapters in the network. The G speeds are quite nice - even though 802.11b's theoretical speed is higher than my DSL bandwidth, it's actual performance was quite dissapointing.

Re:Sounds like

[quantum bit]

It's not illegal by the FCC because anything goes in 2.4 GHz so long as you don't go over the power limits... there's no bandwidth-footprint limit that keeps you from using everything between the lines.

So if my neighbors get one of these I just need something that will broadcast random noise at the maximum allowable power level over the whole 2.4Ghz band, with a directional antenna. Then we'll see how long it takes for them to give up and take it back to the store because it doesn't work.

same effect using two wifi cards?

[stiller]

Maybe I'm missing something, but could you not achieve the same effect by whacking in two wifi cards and using some loadbalancing scheme in software? I don't see why this would be any slower than a hardware implemented solution, I think the CPU cycles spent are hardly the bottleneck here. Having said that, I think this is a silly solution that only wrecks the whole shared spectrum idea of wifi. If I were a wlan network admin, I would find (packet analyze) and block these guys. (not applicable on public networks ofcourse).

scanning networks

[Gyorg_Lavode]

Would a card like this, integrated with airsnort and appropriate drivers allow quicker characterization of the network traffic in an area?

Re:overlap?

[PetWolverine]

With 4 you can at least guarantee that no two adjacent access points are on the same channel.

Is this a consequence of the four-color theorem? It sounds distantly related.

Re:overlap?

[wankledot]

Yes, except with wireless there are no solid boundries, so it's not completely the same. You could have 6 non-overlapping channels are still have interfearance problems if they are stepping over each other. . It's very possible to have overlapping pieces, but as long as you manage your power levels and design it well, you should be OK.

Re:Some comparrisons regarding G

[bru_master]

This is good info to have, Keep in mind if you are trying to impliment wireless in a corporate environment, crappy range is a good thing. I have common problems with Cisco and Symbol AP's having such good range that too many users will pick up the signal in the cube farms of todays corporate america.

I try to place my AP's so that 25 users will access them from there desks or conference room. Some conference rooms that are very large I will place 2 or 3 AP's on different channels with the power turned all of the way down so it will balance the user load between the them.

Since there are only 3 non overlaping channels it is often a chore to design wireless in a room where the same channels dont overlap with each other. Poor range solves this problem. Think of how to put 5 AP's in a room with 3 channels, it can be done, poor range is the key.

We urge our users to use their wired connection and use their wireless when in meetings or on the road at other corporate offices.