Slashdot is powered by your submissions, so send in your scoop

 


Forgot your password?
Close
typodupeerror
×
The Internet

Four Simultaneous Access Points OK for 802.11b 73

Posted by michael from the slashdot-needs-a-wireless-topic dept.
jlouderb writes "ExtremeTech is reporting on a new analysis that shows that four of the 11 802.11b channels can actually be used simultaneously, rather than the just the three used today. This has big ramifications for multi-access point installations, especially in taller buildings. The analysis was done by the CTO of an 802.11b startup called Cirond and a white paper with all the details should be posted to their site later today."
This discussion has been archived. No new comments can be posted.

Four Simultaneous Access Points OK for 802.11b More

Four Simultaneous Access Points OK for 802.11b

Comments Filter:

  • Posted to site later.... (Score:5, Funny)

    by Chicane-UK ( 455253 ) <chicane-uk@[ ]world.com ['ntl' in gap]> on Saturday November 16, 2002 @07:22AM (#4684976) Homepage
    "The analysis was done by the CTO of an 802.11b startup called Cirond and a white paper with all the details should be posted to their site later today.."

    Not if /. has anything to do with it ;)

    • Re:Posted to site later.... (Score:2, Insightful)

      by Anonymous Coward
      One thing they are forgetting in all this "excitement" is that it does matter a great deal the distance between the four different channels. It matters how close you have those APs together. If they are right next to each other, they will probably overlap their signals by as much as 50% or so. (Although software takes care of this.)

  • Damn you Europe (Score:3, Funny)

    by InvaderSkooge ( 615857 ) on Saturday November 16, 2002 @07:24AM (#4684978) Homepage Journal
    Essentially, according to Burton's analysis, you can safely use four channels in North America, and five channels in Europe

    Before anyone thinks that Europe is better because they have more useable channels, don't forget that our 802.11b channels have much larger asses than theirs.

    • ...don't forget that our 802.11b channels have much larger asses...

      Are you talking maximum rectal diameter here, or buttockprint?

      If the latter, you may find that higher power doesn't always improve communication range, and in practice, the 802.11b protocol will in almost all cases use less power than the maximum allowed, making wattage a moot point. For access point density you want less power, not more.

      If the former, I've got no clue how to improve on the situation.

  • Why four is better than three (Score:3, Interesting)

    by ites ( 600337 ) on Saturday November 16, 2002 @07:38AM (#4685006) Journal
    Ask any map maker. You cannot tile a surface of arbitrary shapes with three colours. You need four colors. Since a WiFi channel acts as a map-filling color, the ability to use four channels gives real advantages.
    Well, it made sense when I thought of it. :-)

    • Re:Why four is better than three (Score:4, Interesting)

      by BlueUnderwear ( 73957 ) on Saturday November 16, 2002 @07:49AM (#4685024)
      Yes, but they are speaking about tall buildings here? So wouldn't they need to solve the 3 dimensional equivalent of the map coloring problem (a map is only 2D), and thus need more than "colors"?

      However I agree, 4 is still better than 3.

      • The general map coloring problem is pretty useless in a 3D world, there is no minimum # of colors, for example. Buildings, however, are not 3D, they're layers of 2D, so it would be solving a modified 2D map coloring problem.
        • Buildings, however, are not 3D, they're layers of 2D, so it would be solving a modified 2D map coloring problem.

          Wouldn't signals "leak" from one floor to the one above and below? Which would make it indeed 3D, wouldn't it?

          • Wouldn't signals "leak" from one floor to the one above and below? Which would make it indeed 3D, wouldn't it?

            That's why I said it's a modified 2D problem. It is a stack of 2D color problems, along with dealing with an adjacent 2D color problem. The "leakage" is always from one plane to an adjacent plane, each tiled, with no interaction across a plane. It is not truly 3D, as each plane simply interacts with one other plane, as far as a minimum answer goes, and while there is no limit to the answers for a 3D problem, there is for a modified 2D problem like this is. The maximum number of colors needed for such a modified 2D map is 8, with an alternating set of 4 colors every level.

          • Sure signals would leak from floor to floor, but floors are made of more significant material thanmost walls in buildings as well. Which means then atenuate the signal more. So, inherently less interference from the floors below (assuming all else is equal... it isn't...)

            According to one article I read on the antenas used for a particular access point it mostly radiated on the plane. Meaning there would be less signal from above or below, combined with the added issue of a floor/ceiling between them and your almost home free...

            YMMV...
        • Why is there no minimum number of colors for a 3D graph? I can't think of an example that would require an infinite number of colors, therefor the statement about there not being a minimum number of colors appears to be false.
          • Start with the simplest manifold, which is a sphere. That only needs four colors. Which should make sense because maps are 2d projections of a mostly spherical surface and maps only need four colors.

            Now add a topological "handle" to the sphere (poke a hole in it and make a torus or donut). Now you need six colors, though there's no proof of that, only a counter example for five.

            It turns out that for each "handle" or tunnel in the manifold you need zero, one, or two more colors. If the next handle doesn't interact (stick two donuts together at the edge), then you don't need more colors. If you poke another hole in the donut from the edge into the middle, you do need another color.

            Another way to think about it. Take a map that requires four colors to draw. Now add a tunnel in place between two areas of the map that have the same color and say that the border between the two endpoints is halfway through the tunnel (so that two areas with the same color touch). You'll have to add another color to print that map now. Repeat for five colors, then six, etc. You can always add another tunnel that will force two regions of the same color to be in contact, so for an arbitrarily complex manifold, you can't specify how many colors it will take.

            Wait, that actually is a proof. :)

            Regards,
            Ross
            • That's a really nifty way of describing the problem.

              Meanwhile, back in the real world, there usually isn't a physical equivalent to the proposed 'tunnels'. Unless some mischievous architect starts adding waveguides between distant floors, the volume associated with each access point will probably be something close to an oblate speroid. Under those limited circumstances, there is a solution to the colouring problem.

            • Fantastic!

              Great visualisation.

              But what if we are not talking about coloring the surface of arbitrary 3-d shapes (I assume that is a manifold?), but rather 3-d _volumes_? I.e. Without any non-local tunneling?

              Is there a minimum coverage needed for that case? Or am I too stupid to see that this is the same case as with manifolds?
      • Not if you use 3D colors.

  • hah (Score:5, Informative)

    by EmagGeek ( 574360 ) on Saturday November 16, 2002 @07:38AM (#4685008) Journal
    There is still frequency overlap if you use 4 channels... they're just assuming that the spatial parameters will guarantee that adjacent channels won't interfere.. They're basically banking on the APs being far enough apart to cause little or no interference. Of course, in order to set up the "4 AP" system, you'd have to perform a site survey, which would include numerical modeling of the structure of your building, site measurements of field strength at various locations, and in all, a couple hundred thousand dollars worth of engineering...

    Besides, if you want that many people on your network, 802.11b is not appropriate given it's slow link speed. I don't think 802.11b was ever meant to be the only netfrastructure in a commercial setting. My company uses it for conference rooms so the "important people" can bring their laptops in to waste everyones' time with presentations about schedules and upcoming meetings. :)

    Putting any more than about 10 people on a wireless network in a "we're trying to do actual work" setting is suicide... oh wait, this guy was with a startup, they probably just use it to play Quake...

    • oh wait, this guy was with a startup, they probably just use it to play Quake...

      With the funding they would got 3 years ago, they could have bought machines to play Quake 4! Instead they are stuck with that new pentium 3 and 2GB of RAM :(
    • 1) Why do you assume that since it's covering the whole building there will be a lot of people? I like the the network where most people use wired ethernet when they can, but they can fall back wireless when out of their office, say visting a coworkers office, or working in a LAB, or checking system status on a PDA.

      2) The number of people using the network is only 1 factor in a larger equation for determining the viability of WLAN. Depending on usage, a single WLAN Access Point may not be fast enough for 1 person. Or it could be fast enough for 500 people. All depends what they are using it for. Are servers used extensively?, for Files? Applications? Do people surf the net? Do processing jobs suck large files over the network?

      3) This article does not state that adjacent channels won't interfere. They are saying that they won't interfere VERY MUCH. If you are setting up a 3D grid of WLAN AP's, full coverage with a little interference using 4 channels is better then what can be done using 3 channels. With only 3 channels with 2 AP's per floor Some adjacent AP's will need to be on the same channel. That would be bad. With 4 channels, the overall network capacity is a little lower, but you have better coverage.

      4) I Agree, people like you should definitely pay somebody for a site survey. Probably you should hire someone for the entire network design, not just wireless. I'd be glad to do the job, I'll sub the work out for $10,000, and put the other $190,000 in my wallet.

    • Re:hah (Score:4, Interesting)

      by mindstrm ( 20013 ) on Saturday November 16, 2002 @08:30AM (#4685111)
      Where did he say frequencies don't overlap?

      Where did you read that this is about spatial separation?

      What kind of real business are you in where you think 802.11b can only handle 10 users "doing work". Network intensive work, perhaps, but many, many businesses use their network for email and surfing only.

      The article says they are talking about acceptable levels of interefernece that do not degrade performance... which is entirely possible with some channel overlap.. that's one of the benefits of direct-sequence spread spectrum, it's inherently redundant, many times over.

    • bah (Score:5, Insightful)

      by tulare ( 244053 ) on Saturday November 16, 2002 @08:41AM (#4685133) Journal
      Nothing annoys me more than the overly-precise. Let me take this point-by-point:

      Yer right about the method by which the author asserts that the extra channel is made possible by addition of the third dimension (or third floor, in the least-case solution of his problem). But then you go and blow it with hyper-precision. Honestly, in this day and age, your "site survey" will very likely consist of a tech wandering around with a laptop looking at the signal meter. If the company in question is particularly anal, that tech might be actually carrying a notepad and perhaps an actual signal strength meter. Numerical modeling? Not likely.

      Your second paragraph is predicated on the idea that the lan will consist solely of wireless. Again, not likely: not many people want to bother with replacing existing ethernet cards sitewide with 802.11 cards. More often, the wireless is for new computers and for laptop users, which is really a small proportion of all the computers in the building, no? As for myself, I have a wireless link in my office, which is fine for all sorts of "real work," and if for some reason I need to move big chunks of data, then and only then will I bother grabbing a 100-base-T link.

      As to your third paragraph, utter hogwash. We have been very successfully setting up reasonably-sized labs with wireless-only networking for some time. It's really nice to do this when buying new hardware - what a dream when the only cable you need is power! (big hint here: try to find a cheap - US$750 with air, er, 802.11 built-in - computer, with a unix installed, and no butt-full of spaghetti sticking out the back. Give up? Think different) No problems. None.

      "It's people like you what cause unrest"

      • Re:bah (Score:3)

        by Ponty ( 15710 )
        Spot-on. I go to CMU where we have one of the first campus-wide 802.11b networks that hosts, literally, thousands of users every day. It's really impressive. Performance is fantastic and reliability is high everywhere on campus but in the elevators (they're big Faraday cages.)
    • Nope. You could set 3 of the APs directly on each other and have no performance problems whatsoever. I have done it with Aironet gear, now Cisco. It's a good solution to getting more bandwidth to a room because different pc's can and will authenticate to different APs on different channels allowing for bandwidth sharing.

      Further more you can do slick multilink ppp setups by building long links with multiple APs on each side with parabolic antennas which all connect to a router on either side balancing links with Multi PPP.

      So you can easily get 33 megs out of a long >10 mile link out of this stuff.

  • Why the frequency spread? (Score:3, Interesting)

    by Big Mark ( 575945 ) on Saturday November 16, 2002 @08:04AM (#4685061)
    According to the article, the channel centres are seperated by 5MHz, yet the channels themselves span 27MHz.

    Now, I'm probably missing something really, really obvious, but why don't they just limit the spread to, say, 2.4MHz from the centre? Surely that way all the channels (11 or 13, depending on where you are) could be used and you would still have a small buffer zone between channels?

    Is is just that they can't make the frequency generators precise enough, or something?

    • Re:Why the frequency spread? (Score:5, Informative)

      by mindstrm ( 20013 ) on Saturday November 16, 2002 @08:32AM (#4685118)
      Channels are 22MHz wide, not 27...

      If they limit the spread to 2.4Mhz, almost 10 times less spectrum, then the data rates would drop and the succeptability to outside interference from other 2.4Ghz radiators would go way, way, way up.

      It has nothing to do with how precise they can make the equipment, and everything to do with playing nice in the 2.4Ghz spectrum.

      The reason there is channel overlap is because channel overlap is NOT as bad as everyone makes it seem.. poeple still tend to think in terms of normal radio bands.. where overlap destroys the signal. This is direct sequence spread spectrum, it can handle some level of noise.. hence the overlap.

  • Not particularly earth shattering... (Score:4, Insightful)

    by Raleel ( 30913 ) on Saturday November 16, 2002 @08:08AM (#4685067)
    Essentially, they are being "risky". A small section of overlap, in the weaker power section. I'm sure that they are not the only people to think of this. Certainly, they use of four aps does help cover an area more effectively, no question, and the diagrams do help show that :)

  • In other news... (Score:5, Interesting)

    by ihowson ( 601821 ) <ian@FREEBSDmouldy.org minus bsd> on Saturday November 16, 2002 @08:20AM (#4685090) Homepage
    Using seven of the available channels was ranked as "sort of OK", nine described as "highly dangerous" and eleven of the available channels ranked as "marginally suicidal", by a team of highly trained hamster analysts.

    If channel overlap is an issue where you are, you probably have too many damn AP's. Witness, rooms 424 and 417 of the EE building at Sydney University - two access points PER ROOM. Admittedly, they're large rooms (labs). I'm told that one is meant to be taken out of each and used elsewhere.

  • Ramifications not that big (Score:5, Interesting)

    by mindstrm ( 20013 ) on Saturday November 16, 2002 @08:24AM (#4685102)
    and the hundreds of RF engineers who have already spoken and said that only 3 channels do not overlap are still correct.

    802.11b Channels are 22Mhz wide, and spaced 5Mhz apart.... grab a pencil and paper and figure it out. You can't get more than 3 channels without overlap.

    The article lacks any real detail, other than a brief but accurate (typo aside, channel 1 goes to 2423MHz, not 2433Mhz) description of the 802.11b channel scheme.

    One of the benefits of DSSS is that you can deal with interference to a good degree. If you use four channels, as widely spaced as possible, instead of three, you narrow your bandwidth, but not by too much. I imagine the overlap could be reduced to between 2Mhz for the end channels, and 4Mhz for those in the middle, possibly only 2 for those in Europe where the spectrum allocated is wider. Given how DSSS works, this may not affect data rates noticeably... this is what they are probably going to talk about in the alleged whitepaper.

    Not sure why it's an article yet... there's no info yet.

    • Have you read the white paper? From reading your post I would assume NO! I am in the middle of installing a wifi system for our company and we are running into a lot of problems, but adding another channel will defiantly help me in my quest. I found the wire paper on their company web site (www.cirond.com. Is this company publicly traded?? I found the symbol CRDT and wanted to know if this was the same company?

  • Let's just hope... (Score:3, Funny)

    by JessLeah ( 625838 ) on Saturday November 16, 2002 @08:45AM (#4685138)
    ...they don't get a patent on using that fourth channel ;)

    <kiki>Stay good, cute lil' 802.11b startup! Stay good!</kiki>

  • Thank god for err;; c;;;ection (Score:5, Funny)

    by tulare ( 244053 ) on Saturday November 16, 2002 @08:47AM (#4685141) Journal
    I'm using 802.1sssjjjjssss;;;sjsfffeighbor insists that ''''ing the sa;e channesswill majjjjhe conn;;;ion stronger. Hessssery smart, don't you think?

  • 802.11b - learn the truth (Score:5, Informative)

    by puzzled ( 12525 ) on Saturday November 16, 2002 @08:58AM (#4685169) Journal
    There are so many nonsensical articles about what 802.11b can and can not do that I thought I'd set the record straight.

    There are eleven channels available in north america - 22MHz wide, spread from 2402 to 2483 MHz, with 5 MHz guard bands between them. Channels 1, 6, and 11 don't overlap, the others ... well, sit down and draw yourself a map.

    There are many other things in 802.11b besides DSSS 802.11b cells that you 31337 kids can h4x0r - I've got Western Mux Tsunami and Adtran Tracer T1 bridges. T1s are full duplex - these types of radios split the ISM band 50/50 - one end sends with the bottom half and listens at the top, the other side is opposite, and they use 100% of the spectrum.

    The other thing you'll find are FHSS systems in the ISM band. The most common is the Alvarion (previously Breezecom) Breeze Access II three meg access radios, but Cirronet's lower speed ISP products is starting to appear in rural areas.

    If you're working inside a building with full duplex T1 bridges or a hot FHSS somewhere outside its definitely going to make a difference, and that goes double if you're running an 802.11b system outside. Putting one of these things near an 802.11b AP is basically like sand blasting a soup cracker ... I've seen DSSS signal quality go from excellent to unuseable just by flipping on a Cirronet AP in the same area.

    Assuming you've got no problems to deal with other than your 802.11b, other's 802.11b, and building layout you've still go trouble.

    The 802.11b MAC layer is *broken*. If I pull up and start listening on a channel you're using, even if you've got WEP enabled, I can see your mac addresses and I can *issue disconnect requests* after forging your MAC and the AP *will honor the disconnect*. WEP is the equivalent of an ESP (encapsulated security payload) in IPsec and it protects your data, but the MAC layer needs something like the IPsec AH(authentication header) so that an intruder can't manipulate the MAC layer.

    Building systems always have dead spots. Always. 2.4 gig bounces like crazy when there is sheet metal (HVAC duct work) is in the area. You get reflected signals (multipath) which causes corrupt frames, you get dead spots due to the signal being blocked, etc, etc. You can add further misery by trying to use an AP with 'diversity'. Drop the word from your vocabulary - its 'perversity' mode - just take the time to monitor FCS errors on an AP with this enabled and you'll know what I mean - turn that stuff on in a situation where both antennas can see the same signal and you'll toast 50% of incoming frames *every time*.

    The 802.11b MAC layer is *broken*. If you want detailed knowledge I'd strongly suggest a read of the OReilly's 802.11 Wireless Networks book, but the game goes something like this. The channel you're in is a *shared* resource - that means you share with the rest of the world. 802.11b stations gain exclusive access to the channel they're on by settings a NAV (network allocation vector) in certain frames. Even if you have a WEP protected network the exposed MAC layer you're using will honor NAVs *from devices not on your network*. So when the same intruder who was disconnecting individual stations a few paragraphs back gets tired of that he can start issuing bent control frames that plug up the spectrum and bring your network to a crawl.

    I've barely scratched the surface here. If you see a pretty diagram and a lot of marketing buzzwords, understand that the reality is much, much more grim.

    • Re:802.11b - learn the truth (Score:5, Informative)

      by mindstrm ( 20013 ) on Saturday November 16, 2002 @09:09AM (#4685186)
      You are talking about the 2.4Ghz ISM band in general, not 802.11b.

      You say they use 100% of the spectrum. According to any regulations on ISM spectrum use I've ever heard, that's illegal. For the same reason, you can't time-sychronize a bunch of fhss transmitters to use the entire spectrum, as they would be classified collectively as a single device.
      If you mean it uses FHSS and hops around the entire spectrum, that's different than using the entire spectrum (I'm sure you know that but readers might not.). In FHSS,you set different transmitters to use different hopping patterns, so that interference is minimised.

      The 2.4Ghz ism band is used for lots of things, and many of them can interfere with each other, including leakage from your microwave oven.

      • Re:802.11b - learn the truth (Score:5, Interesting)

        by GoRK ( 10018 ) on Saturday November 16, 2002 @10:56AM (#4685519) Homepage Journal
        It's not illegal the way they do it. They use a DSSS frequency that is 33mhz wide on the upper half and 33mhz wide on the lower half. This is the same type of usage that you'd get colocating 802.11b dsss ap's on channels 1, 6 and 11. The same laws are not true in canada, though. You are technically only supposed to use a maximum of 50% of the band for your entire setup.

        most FHSS radios also use 100% of the spectrum, but they break it up into one of 74 or so chunks that transmit one at a time. The illegal thing to do is to synchronize let's say 74 fhss radios so that they transmit without accidentally hopping on top of one another. These laws are not the same everywhere, though, and in some parts of the world, it's indeed legal and a very good thing to run synchronized fhss radios. Incedentally, the alvarion fhss radios actually support this operation: but you actually have to use them in africa and such other places.

        Anyway, kudos to the parent poster. I have been arguing his points for years here on /. (ever since the editors and seemingly everyone else creamed their panties over the bastardized 802.11b). Really, deploying 802.11b for anything more complicated than a single or dual AP installation so you can walk around your house/office with your laptop is probably a bad move. There are a lot of better wireless technologies out there and a lot of them aren't even that much more expensive than 802.11b crap.

        ~GoRK


        • You *can* run 802.11b outdoors, with a lot of clients, in noisy environments - I have a dozen customers on an AP in a two mile wide valley with a bunch of other stuff - you just have to have a detailed understanding of the MAC layer and how to tune stuff to make it work. Today 95% of all WISPs are wildcatters, like dial up was in 1994. That will shake out to a handful of disciplined operators over the next few years.

          802.11b customers aren't where the big bucks are, but its going to be like ethernet - living on well beyond its appointed time, due to inexpensive gear.
      • Time synchronizing FHSS transmitters to use the entire spectrum also happens to defeat one of the main purposes of FHSS, which is to reduce multipath problems by changing frequencies before a second path reaches the receiver. I would expect that technique to result in a lot of interference.

        -John
        • Say what? Typical FHSS radio dwell times (that's the period at which the frequency changes) are in the tens to low hundreds of milliseconds. Radio waves travel at 3.0e+10 meters/second. Let's say you were running a ridiculously low dwell time of 2ms (which I don't think 802.11 equipment even supports, BTW), and you wanted to hop fast enough to avoid receiving a second path on half the incoming packets. That means there would have to be a 1ms time differential between the two paths. And THAT means there would have to be a 3.0e+07 meter length differential between the two paths, which is greater than the diameter of the earth. You show me a 2.4GHz wireless transmitter powerful enough that the receiver has to worry about multipath on the scale of 30,000km, and I'll show you a stiff fine and a jail sentence from the FCC.

          In short, you're full of it. FHSS does not hop fast enough to avoid multipath. Multipath interference comes in typically less than one symbol time after the main signal. If you hopped fast enough to avoid it, you'd never receive a single bit correctly.



      • Any radio that bridges T1s uses 50% of the spectrum on one side, the other 50% on the other side, and transmits 100% of the time.

        Its perfectly legal - look at Adtran Tracers in the ISM band, Western Mux Tsunami products with side band T1, etc.

    • This is why I always thought the FCC should have licensed out an 80 MHz band to a particular technology that could be used by anyone rather than a specific class of devices that can all wreck havok with eachother using a mix of technologies.

      It sure would have been nice if we had a cheap adaptive FHDSS system with a decent security layer.
  • 802.11g uses OFDM which has 52 subcarriers instead of a single carrier to send the data across the link. Since there is a proposed overlap of 7 mhz (ch1 and 4 for example overlap from 2416 MHz to 2423 MHz) on each side, more than 15 subcarriers will be jammed on each side. So for the middle channels (4 and 8), more than 30 of the 54 channels can be jammed. Since the subcarriers are carrying redundant information, you will lose either throughput or its ability to fight multipath (or both).
  • There are three possibilities: Pioneer's solar panel has turned away from
    the sun; there's a large meteor blocking transmission; someone loaded Star
    Trek 3.2 into our video processor.

    - this post brought to you by the Automated Last Post Generator...

Slashdot Top Deals

/earth: file system full.

Close