Frequently Asked Questions about the Brismesh Inc. and their answers.
Brismesh Inc. is an incorporated association interested in establishing a non-profit, wireless network over Brisbane. Our main goal is to obtain decent, free bandwidth between our homes through a cooperative wireless network.
This approach is particularly attractive because firstly, theoretical wireless bandwidth (108Mb/s for extended 802.11a) dwarfs the speeds available from commercial wireline technologies like modems (56kb/s), cable (10Mb/s) and ADSL (1.5Mb/s). Secondly, there is no monthly rental cost for air, although you do have to cough up for the initial hardware.
First, add yourself to the node database. Second, subscribe to the mailing list. Third, become a member of the association. Then, find and talk to the people around you, and buy some hardware to connect to them.
Access to the "commodity" internet is never going to be free. E.g., if you want to surf overseas web sites, you'll probably transit the undersea cables paid for by Telstra and Optus, and ultimately someone has to pay for that.
Currently, because of the class licence conditions on the use of the microwave spectrum, and non-commercial carrier licencing, Brisbane Mesh must notionally remain independent of the "commodity" internet until some legal questions are resolved. [Letters sent to ACA October 2001] [Wireless Broadband Enquiry under way 2002]
You might be able to organise internet access through other nodes in the mesh, but the legal status of this activity is quite grey.
There are many uses of a local, high-speed cooperative network, including:
If local businesses and/or government organisations (like council, libraries or universities) become involved, their services could also become available.
Not necessarily. Carrier licencing constraints and restrictions on the use of the class-licence spectrum make the establishment of for-profit wireless networks quite expensive. A cooperative community network that avoids these licencing costs can certainly coexist with a commercial wireless network and is probably more attractive.
See also:
Great! Monitor the mailing list and announce your talents! You can volunteer to attend (or organise) one of our get-togethers.
Here are some that kind of fit that description:
Also worth noting is DCnet ISP which is part of Banana Shire Council.
Yes. Other non-commercial wireless metropolitan area networks in Australia include:
There are more... mail the FAQ maintainer if you think some should be added to this list.
The big ones [in 2000] are:
But the number of these is huge, and constantly growning. More complete lists are maintained by various people:
The current management committee is listed here.
Lots of people are putting significant effort into their own local activities, too.
Put simply, work gets done on facets of the project by the people that spend the time and energy. With that comes a natural, de facto leadership of their area.
An IRC channel has been created on
pacific.au.austnet.org port:6667, channel
#BrisMesh.
There is also the mesh mailing list and the
forum.
Other places:
Ministerial determination 20 Sep 2002: WiFi escapes carrier rules (The Aust.), Wireless regulation encourages innovation (DCITA), WiFi providers escape carrier rules (Syd W)
It doesn't appear to be. As long as we
People who operate nodes in the mesh are also encouraged to formally specify an "operations policy" that makes explicit how they allow their network node to be used and what their expectations and requirements are. A sandard "peer agreement" is in the works.
There are some counterpoints to the legalities of community wireless broadband:
My general feeling towards these counterpoints is that (from an individual's point of view) we are simply friends within a community, exchanging data with each other, not causing interference outside that community and so, quite simply, it is none of the government's business. However, government regulation that is distant and non-onerous to bona fide participants is very welcome, even necessary, especially if it effectively protects the community from abuse.
A separate issue is about the legality of what content appears on the network. Here are some links I am collecting in the FAQ.
No.
We are concentrating on wireless networks that transmit in the 2.4GHz and 5GHz bands. These frequencies are licenced for public use.
The 2.4GHz band usage is also summarised in this table (which includes NZ spectrum) and as follows: channels 1 through 11 allow for 4W EIRP; channels 12 and 13 allow 200mW EIRP and channel 14 is not available.
Yes, please. We are looking for a pro bono assessment of our current and planned activities, especially with regard to its legality, and our liability.
Also, a watertight, standard peering agreement/disclaimer would be nice. Ideally, one that protects us from each other and reinforces trust in the common cause.
You really need to shop around, prices change all the time. Here are some pointers to places that can help:
This is hard to answer in a FAQ. The best we can do is mention some of the cards used by people in the mesh and include their comments. Any card that you do buy should have an external antenna jack. Without a jack, you can't easilly add a separate, high gain antenna.
More card descriptions at:
Various card throughputs have been tested with Netperf.
More links:
From time-to-time bulk purchases are arranged via the mailing list. Other Australian wireless groups also often organise bulk buys (notably Melbourne Wireless).
If you want to see current prices, check out our retail prices database.
IEEE 802.11 is a set of standards published by the IEEE. Other 802 standards include 802.3 (ethernet) and 802.5 (token ring).
There are a couple of relevant parts to the 802.11 standard. They are
While 802.11b is a published standard, it does provide some options which means that some 802.11b cards won't talk to other 802.11b cards. To avoid this problem, you can look for a Wi-Fi compliance label, meaning that the device has gone through an interoperability test matrix against other cards and appeared to work OK.
802.11a offers higher throughput (54Mb/s), but at 5GHz is limited (by law) to 1W power, which makes it a poorer choice for long-distance links. However, one vendor extension to 802.11a has doubled the theoretical throughput to 108Mb/s. The other benefit of the 5GHz band is that it is (currently) less cluttered.
The proposed 802.11g standard (54Mb/s at 2.4GHz) has recently progressed [Nov 2001]. When ratified, it will also be backward compatible with 802.11b.
For more information:
Something else to keep an eye on is 802.16 ("WirelessMAN") which was announced and reported on in September, 2001.
DSSS and FHSS are different ways of dividing the 2.4GHz band up into channels. DSSS (Direct Sequence) aggregates adjacent sub-bands into wide frequency channels (that overlap) while FHSS (Frequency Hopping) uses all the sub-bands individually in particular hopping patterns. The SS part (spread spectrum) refers to the general technique of spreading signal power over a large band to defeat some types of noise or interference including jamming.
Both DSSS and FHSS are mentioned in the 802.11 standard. However, they are completely incompatible! You need to ensure that cards intended to talk to each other use the same thing. All 802.11b cards are DSSS.
For more information:
If all radio cards are sufficiently close together, and there is relatively little traffic, then ad hoc mode allows any card to transmit directly to any other card. This strongly resembles a bus network topology, like ethernet uses.
Due to the fact that sometimes not all the radio cards in a zone can hear each other, infrastructure mode designates one card as the "master". The master then arbiters the channel for all of its "clients". Communication is then between the master and a client. This kind of resembles how a SCSI bus works.
See below for information on IBSS and BSS.
If you're referring to 802.11 WEP (Wireless Equivalent Privacy), then no. First, WEP has already been broken. Second, if you encrypt a link, you make it harder for new nodes to join, and the channel key would have to be distributed or made public anyway. Thirdly, you can't guarantee that your data won't be routed over un-encrypted links, or that the node owners aren't sniffing the traffic that they route.
Using WEP provides a false sense of security, which in itself can be dangerous.
A far better approach is to use end-to-end security. This is where the applications implement the security without relying on the network (e.g. SSH or SSL (as in HTTPS)), or the hosts under your control create a VPN (virtual private network) perhaps by using IPSec.
Articles of interest:
Alternative security schemes:
I'd like everyone to have two cards! But well, they're expensive, and you can certainly get by with just one.
Remember: the less cards (and antennas), then the less efficiently the spectrum is used, and the lower the aggregate capacity of the network.
If you're a client node, you only need 1 NIC [network interface card] and maybe an external antenna to connect to a Backbone node. A backbone node requires at the very least 2 wireless nics (maybe one nic and one AP [access point] since no one has confirmed that a nic can be used as an AP yet.)
For a backbone node to properly manage a large number of leaf nodes, it should run its card in BSS (Basic Service Set) mode. This allows much more efficient usage of bandwidth through better collision avoidance, and makes leaf node attachment simpler.
If you have a purpose-built Access Point (AP), then BSS is almost certainly available on it. However, if you are building an access point out of a general purpose computer (like a PC) maybe running Windows or Unix, then depending on the card you have, you may not be able to turn on BSS mode. You might be able to get away with Independent BSS ('IBSS', or sometimes called 'ad hoc') mode if your leaf nodes can all hear each other (unlikely.)
(BSS is also sometimes called Infrastructure Basic Service Set mode just to confuse everyone.)
There is also 'ESS' (Extended Service Set) which is when two or more APs are plugged into a wireline network, and the wireless hosts roam between them. The APs are said to be configured in ESS mode.
Lucent has not yet published how to enable infrastructre BSS mode on their ORiNOCO/WaveLAN cards. This means you shouldn't buy them for the purpose of being a backbone access point unless of course it is part of one of their router products which uses proprietary software (but then IPv6 is not yet available, reportedly.)
Ben Wu tells me:
I once asked the enginners at Integrity and Symbol whether you could just hack the Client Orinoco driver to run in BSS mode. They just laughed and said, 'well, then how can Agere sell the router software?' D'oh!
This section describes the cables, connectors and mounting hardware required to connect your adaptor or access point RF socket to an external antenna. The general rule is that this equipment best be impedance of 50 Ohm, whereas antenna equipment for televisions is 75 Ohm (75 Ohm may work but would not follow principles and formula used for 2.4GHz operations).
It is important that you have purchased a network adaptor that supports an external antenna. Many PCMCIA (a.k.a. PC-Card) adaptors only allow use of the built-in diversity antenna and have no external RF socket.
Upon close inspection you will find that the RF connector is somewhat different to the industry standard range of connectors (thus being called proprietary). You require equipment to allow your external antenna (which may feature an N type connector) to connect to this proprietary RF connector on the network device.
If anyone ever wondered why WLAN cards have all these annoying special connectors, it's actually legislated. Various countries (i'm not sure if Australia is one. I'm pretty sure the FCC in the States enforces it.) have requirements that class-licence (ie, 'unlicenced') equipment be shipped with connectors "not available to the general public". That's also why the connectors on mobile phones are so funky.
Yes, the logic of that escapes me too.
The actual FCC rules say that class licensed RF equipment must be
designed to ensure that no antenna other than that furnished by the responsible party shall be used with the device.
It will be easier to obtain connectors for these proprietary connectors as the technology becomes more mature. These "not available to the general public" connectors can be acquired by doing a bit of work. Some networking manufacturers offer ways of obtaining their connectors as a genuine spare part to their product range. There are also third parties who manufacturer after-market connectors, often at cheaper prices.
Some device families use the same type of connectors through their whole range (particularly those from same OEM - for example, Lucent is the same as Compaq). Some manufacturers change their connectors across their product families (for example Cisco offer two types of external connector). You may need to identify the correct type of RF socket on the back of the device, examples include SCX, MMCX, MCA, SMA, TNC, 7/6 DIN & 7/8 EIA (ref: Pacific Satellite).
|
An RF connector needs to be in place at each end of the pigtail. On one end a suitable connector for your network adaptor, the other end an industry-standard RF connector (generally "N" type). Some proprietary connectors can be sourced from x.net.au for about $60 or Pacific Satellite for about $110, although other sources are appearing. They are expensive because they are proprietary. Industry standard N connectors range around $10. You will need to ensure that you purchase the correct connectors to suit your type of coaxial cable.
The type of coaxial cable used for this component is generally fine grade (usually 4.9mm OD) and therefore quite lossy so best to keep the length as short as you can (say under 60cm).
Overall, this piece of equipment will cost between $30 and $150 and depends on the type of proprietary connectors needed, the grade of the cable and quality of construction.
Often antennas (particularly indoor consumer ones) come with a built-in RF cable fly-lead. These can also be called a pigtail. Should you have one of these read on: The connector on this fly-lead may already be compatible with your network adaptor's RF connector and if so you may wish to extend the length of the cable. This is best approached by evaluating what you can do with this existing length. Should it be necessary to extend the cable, install an industry-standard connector (such as N type) in place of your proprietary connector on the fly-lead and use the fly-lead's original proprietary connector (keeping intact and attached its 20cm of cable) to make your own pigtail suitable for the above application.
 |
You need to consider RF loss over the length of your cable assembly. A lengthy run would be considered anything in excess of 10m if using RG213 (or similar 4dB/m loss cable) and 18m if using LMR400 (or similar 0.5dB/m loss cable).
 |
Once you are happy with (and possibly tested) all the cabling and connectors, you should look at fastening the cable to form a secure installation. Cable clips fit over the cable and allow you to mount it against a wall for a professional-looking job. These cost around 20c each and should be spaced evenly to provide sufficient support (say every 1.0m). Trusty cable ties (zip ties) will secure the cable to the antenna mast. You can use an existing mast or install your own ("hockey sticks" which mount underneath the guttering start from around $15). Right angle "N" connectors can alleviate stress on your coax cable if the mounting orientation is awkward. Lightening arrestors, signal splitters / combiners and amplifiers may also form part of your RF connection hardware rig.
Yes, you can use any operating system on the computer that contains your wireless network card, as long as it supports that card and can do IPv6. If you want to act as a backbone node, then you should be able to get IPv6 routing software for your OS. All major operating systems can do this.
See below for more information about IPv6.
Assuming that you are using a PC or similar for a router (instead of purpose-built hardware like an access point etc), and like Unix, you can install a Unix operating system on it, plug in the wireless network cards, configure them with IPv6 addresses, configure the card with the omni to act in BSS mode (if possible), enable packet forwarding/routing in the operating system, and run your routing daemon.
[examples?]For more information:
A note on omnis: There is concern that an omni consumes a lot more spectrum-volume than you would with a directional. In plain talk, that means you are more likely to affect more area around you, cause interference and suffer interference yourself. Narrow-beam, directional antenna do not really have this problem, but if you're running a hub with a significant number of links coming in from strange directions, there isn't really a cheaper solution. [hmm, what about beam splitting?]
My advice is to simply be aware of the large footprint of omnis and plan to be flexible with your setup; find out if there is any competition for the channel frequency in the area your omni is using and talk to those people - try and work something out. They will probably be equally affected and willing to compromise, since it is shared spectrum after all. If you can't change channels, a bit of shielding around your omni in their direction might do the trick.
Don't let threats about being reported to the ACA throw you off; as long as you keep below 4W of power (EIRP) you are fine. (NB channels 10 and above in 2.4GHz are limited to 200mW by the ACA.)
 |
However, cards that bridge from PCMCIA to PCI involve a lot more complicated circuitry and are generally more problematic. Avoid them. Also, some of the newer PCI Cardbus bridges aren't properly supported by some operating systems (check for updates).
 |
If all BIOS hackery fails, you will probably be able to take the card back to your supplier for a refund. The supplier cannot decline a refund if you legitimately believed that the card would work on your system (e.g. meets the minimum system requirements written on the side of the box.)
Finally, a simpler alternative is to acquire an old laptop with built-in PCMCIA or Cardbus sockets that are much more likely to work.
If it's routing and relaying other people's traffic, then yes. Most computers (with their monitors turned off) consume very little power; certainly less than a 60W light globe.
As the mesh becomes denser, redundant links and alternate paths will appear. This lowers the importance of individual node reliability and makes the network more robust (at the expense of some extra routing complexity.)
You're looking for Ryan Armanasco's "PWNPs Apple Airport Cosmetic Surgery Clinic".
 |
Many of the free Unixes (Linux, *BSD) will run on these old computers and don't require things like mice, disks or displays. They drive the hardware at almost optimal performance with a very small memory footprint and extremely high reliability.
You don't have to care too much about memory or CPU speed, as routing packets is not a particularly taxing activity. The main constraint is the bus speed. The 16 bit 8MHz AT (ISA) bus from 1984 has a theoretical bandwidth of 128Mb/s, but, due to the bus transfer protocol used, it is practically limited to 8Mb/s. (Compare this with PCMCIA at 20Mb/s, and PCI and Cardbus at 132Mb/s.) In addition, the AT bus bandwidth is shared across all the network interfaces, so if an AT bus is servicing an ethernet card and a wireless card (possibly in an adaptor) it will theoretically max out routing at 4Mb/s. Similarly, a laptop with network interfaces in both of its two PCMCIA slots will max out routing at 10Mb/s.
So what does this all mean? It means that an el-cheapo AT bus may be enough for you, since 11Mb/s 802.11b cards realistically operate at less than half of their theoretical maximum namely 6.5Mb/s. In long-distance or high-noise circumstances, the air bandwidth becomes even more limited.
Finally, if your junky computer is a particularly robust laptop, you should install it as close to the antenna as possible - possibly in the roof of your house (but mind the summer heat!) In this way you will minimise the length of the cable to the antenna and the signal loss associated with that length.
Note: Mb/s and MB/s as used above and throughout this FAQ and in the context of bandwidth have special meaning. Little 'b' stands for 'bit', and big 'B' stands for 'byte' (8 bits). The SI prefix 'M' (mega) means 106. So, 5Mb/s is 5,000,000 bits per second. Only when talking about RAM does MB (megabyte) mean 10242 bytes (1,048,576 bytes) and even that is slowly becoming anachronistic.
Yes! Donate them to the Router Hardware Pool.
David Leonard and Adam Secombe are coordinating a pool of donated hardware that could be useful to people setting up wireless nodes. If you or your work want to donate old, but still-useful equipment, they can come and pick it up.
Please see the Router Hardware Pool page for information on donations and distribution of items. A database of what items have been donated and how they are being used is available there, too.
It's getting harder to get laptops, according to this slashdot article
However, Adam Secombe provided the following story/answer when looking through auction houses or used computer markets (check your newspaper):
"Keep this in mind when hunting for your notebook for our purposes, a notebook for wireless routing does not require a working keyboard or screen. During setup both of these items can be plugged in externally and removed once the machine is setup.All the same, don't spend extrananous amounts of money on non-working hardware, it might be completely ratshit. $20 for a "condition unknown" notebook is fine.
The bare minimum I would go looking for is a 486DX of some sort, 8meg should be enough to boot Linux & any modules you require. For our purposes, we can also boot these routers off a single floppy disc - hence your notebook won't require a hard disk either. So if you see something around the place with 2 PCMCIA slots, grab it.
Most people won't want to frig around with their hardware as much as others, so getting a completly working notebook would be preferable..
From experience, cheaper notebooks can be found: "right place at right time" sorta thing.
Go to the Sunday Computer Market . Go to some auctions (non-online), check Saturday's Courier Mail for the auction and inspection time. Again keep in mind that you cannot return an item from an auction, so don't spend too much on something if you don't know what condition it is in.
If you get a chance to inspect a notebook, and power it up - do so. Make sure the unit powers up and you get a screen. I bought my 486 notebook like this, only problem was it has a shit battery.
The P100 I bought the other day had a bad screen and bad battery, but other then that it (was) fine, I traced the source of the screen's problems to a ribbon cable which had a cracked trace. I forgot to put the CPU back in the PCB & when i went to fire it up, I must have underloaded the power supply and it blew a 220 uF tantalum (because of the smell) capacitor, I'm hoping that's all that blew so a little surface mount soldering (oh I'm looking forward to that
:() and it will fire up again.This one I will probably use for wireless, it is all on the one PCB which is extremely small.
That is of course if I get it going again, kicking myself.
-adam"
Ben Dale's email on this topic says (abbreviated):
[more information needed about ESSID responses and broadcast behaviour]The residential gateway is just an access point with a modem. The software that comes with it is a dumbed-down home-user package which simplifies the set-up of the device.
Access points are typically bridges, and not routers.
The difference between bridges and routers can be explained after first mentioning the protocol stack model.
| OSI layers | example | |
|---|---|---|
| 7 | Application | FTP (client or server) |
| 6 | Presentation | n/a |
| 5 | Session | n/a |
| 4 | Transport | TCP |
| 3 | Network | IP |
| 2 | Data link | 802.11b |
| 1 | Physical | air |
Each layer communicates only with the layers directly above and below it. Packets (usually) work their way down to the physical layer, then across the physical layer and then work their way back up, but in a different host (with the same stack). Typically a hardware device (such as an access point) which operates at a given OSI layer also operates at all of the layers below that level.
 |
Devices that 'copy' packets from one side to another do so at a particular 'layer'. They have different names, but they can be explained in terms of the model:
This is best described in the About page for the database. The map generating code was hand-written in PHP and C. The elevation maps use the Brisbane and Gold Coast tiles of DEM (elevation data) from AUSLIG was converted into a quadtree of floats for fast access.
Conversion from different grid and geodetic coordinate systems was assisted by using basic geodesy equations and the NTv2 GDAy shift data files from Qld Dept of Natural Resources and Mines.
Maybe. Send the request to the list, after checking the archive to see if it's already been discussed.
You don't. The database is open so that other people may correct your entry if they notice an error. This does rely on a significant amount of honesty and trust. It is good etiquette to send mail to the node's contact before editing it.
The database is backed up nightly, and modifications to entries are logged if you have a serious gripe. You can configure your entry so that you are notified by email when the entry is changed.
If you are showing up in the black zone of the elevation map, then sorry, we only bought two tiles of altitude data. Each tile costs about $100; if there are enough people willing, we can split the cost to buy the missing tile. Again, mail the list.
Decibels (dB) is a logarithmic measure of a ratio, usually
the signal-to-noise ratio.
Decibels are convenient because you only need to add and subtract gain
and loss factors instead of multiplying them together.
If a device makes something (like voltage or power) stronger by a
factor of x, then with logarithms we can calculate that it has a gain of
dBm is the imagined gain from a 1mW input. For example, if your card says it transmits at 100mW, then that means it has an imaginary gain of 100, or in dB that is 10log(100) = 20dB, and therefore we say it transmits at 20dBm. (100mW=20dBm, 1mW=0dBm) Another card may transmit with 50mW of power, which is 17dBm.
dBi is the imagined gain of an 'isotropic radiator'. For example 0dBi is the gain of the hypothetical antenna that radiates all its input power into a perfectly spherical distribution. Antennas with such a radiation pattern don't actually exist. Instead, real antenna are built to concentrate the signal power into one particular direction. To a distant receiver (which doesn't know what kind of antenna you have) it only "sees" the concentrated signal sent its way, and it looks to it like you could have a spherical radiator that has somehow increased the input power and radiated it spherically... and so we call say antennas have 'gain'. Remember, the 'i' in dBi tells us its assuming an isotropic radiator model.
Typical low-cost antennas have gains between 10dBi and 20dBi. These gains are measured in the primary direction in which the antenna is focusing the power, called a 'lobe'. If it was measured from the side or behind, it would look less, even go negative. Besides, we usually don't care about those directions where no signal comes out (called nulls).
Next, all the stuff that the radio signal passes through have loss (or gain). Cables and connectors can contribute significantly to signal loss. Cable attenuation (loss) is measured in dB/m. For example, RG213 cable is rated at -0.4dB/m for 2.4GHz signals. Even air and the vacuum of space have loss, but amplifiers have a gain (positive dB) of course!
In general, you add up all the gains and subtract all the losses and see where you can and can't improve things. This is called your link budget.
For example, if you have a +17dBm card, 5m of -0.4dB/m cable and a +19dBi antenna, and no connectors, than the effective signal power is 17-(5*0.4)+19 = 34dBm at the antenna (we haven't counted the air or the other side's antenna yet).
By law, we're limited to 4W of EIRP (effective isotropic radiated power). That means the magic number we can't exceed coming out of the antenna is 10*log(4000) = 36dBm. You really want to get your system as close to 36dBm as possible, without exceeding it. A bit like driving at the speed limit.
The next thing you have to worry about is the space that the signal goes through. The loss due to this is called path loss. Clear path loss is the loss you get when the transmitting and receiving antennas are separated by a vacuum and there is nothing remotely in the way (like rooves or trees or people). The mesh database can compute clear path loss for your site, and tell you what it is for different frequencies beneath the elevation plot. The equation is basically the inverse square of the distance. For example, a 2.64km link using 2.4GHz has a clear path loss of -108dB.
And, finally, when the signal gets to the receiving card, (through an antenna and some lossy cable) it has to be strong enough to be sensed. For example, the older 2Mbps Lucent cards have a listed sensitivity of -90dBm.
Again, in our running example, you just add up the gains, and subtract the losses and it should be more than the listed sensitivity. Given that our example sender station has been boosted to transmit at +36dBm after the antenna stage, and that the path loss is -108dB and the receiver is a +17dBi antenna with -2dB of cable loss back to the card... add all those together to get -57dBm. This is plenty more than the minimum requisite of -90dBm for the card to operate.
That means you could get away with longer cables, weaker antennas or crappier cards. Remember, this is all theory: practical losses and gains can vary by the wind blowing your antenna around, water in the cable, low-flying planes.. etc.
Happy link budgeting!
(This answer tells you how antennas work so that you can understand what makes antennas different from each other.)
There are heaps of different kinds of antennas. An antenna is basically a shaped bit of metal. Antennas can be classified by their design, and their design determines their 'radiation pattern'.
To understand radiation patterns, you need to know some simple facts about about "radiators".
Theorists sometimes talk about a hypothetical "spherical radiator", which is an imaginary antenna whose signal power is uniformly distributed in a sphere about it. Although nice to think about, perfect spherical radiators don't exist.
Instead, real antennas emit power that is different from different angles; it is bunched up in some areas (lobes) and absent in other areas (nulls). The main lobe has a "beamwidth", which is the angle containing power within 3dB of the maximum. Usually there are lobes hanging off the side, and these are (surprisingly) called "side lobes".
 |
The radiation diagram shown here is of a slice through space; the actual radiation is in three dimensions, not just two. (c.f. the 3-D dipole radiation donut pattern below). Sometimes, two beamwidth angles are given; they are the azimuth (side-side) and elevation (up-down) beamwidth angles.
 |
 |
The most basic kind of antenna you can get is called a half-wavelength dipole. This is basically two bits of wire of a special length, and is the type of antenna most commonly found built-in to those PCMCIA wireless cards. Its radiation pattern looks like a fat doughnut; you get no signal when you look through the "hole" (i.e. down the wire).
At 2.4GHz, the half-wavelength of microwave radiation is about 6.25cm.
 |
 |
Intuitively, yagis work by having a "driver" dipole, a "reflector" element (5% longer than the driver) and some "director" elements (5% shorter than the driver). Distances and lengths between these "parasitic" elements are chosen carefully. Putting a yagi in a PVC pipe ("radome") doesn't inhibit its reception.
 |
Narrow helixes produce an omnidirectional radiation pattern. You find these in "helical whip" antennas.
There is also something called a cavity antenna. In microwave applications, this is sometimes called a feed horn. A monopole or dipole radiator element is carefully located in a half-open cavity (e.g. tin can) and using some resonance properties, a beam pops out the open end. [Is this right? It looks like a feed horn to me.]
Cheap antennas will work, but the more you spend, the better the quality of the gear is. What you should consider first is what kind of range or area you want to cover, and how much you're willing to spend on the gear.
Typically, you'll either want a highly directional (yagi, helical or dish) for a single-person link, or an omni-directional (whip) or sector antenna for multi-person link.
The mesh idea is to get most houses to have more than one antenna: two, or hopefully three. Because only when you have two or more anteannas can you start to relay traffic further than anyone else!
Putting up omnis is not such a great idea, because you can have only so many omnis in one area (before the channels start colliding), so it is better to have more directionals than omnis. Also, the bandwidth provided by an omni/hub is decimated by the number of clients using it. Links using directionals have a smaller footprint, can reuse the spectrum more efficiently, need less power to go further and have dedicated capcity.
If you're starting off new with this, I'd suggest you go for a directional antenna first. Even cantennas would be a good start - they're cheap to build, and you can reuse the cable and so on when/if you get a proper dish. You learn a lot about what kind of antenna would be right for your spot too by going through the exercise.
You want to read these:
Also, if you can get hold of PayTV antennas you can 'mod' (modify) them for use in 2.4GHz applications. The ex-Galaxy MMDS grid antennas cost US$128.50 new but you can get them much cheaper (around AUD$10).
And, no, a TV antennas won't work.
At 2.4GHz, the water in and on the leaves of trees is a particularly good signal absorber. (So too are people and concrete as they also contain significant quantities of water.)
If you set up your link on a dry summer day and there are nice dry trees in the middle, it might work then but you'll probably find your signal completely disappears as soon as it rains. We don't have too much of a fog problem in subtropical Brisbane but that would also have an effect.
The only real way to know if trees affect you is to test the signal strength during, or just after it rains. One rough guide is -0.3dB/m for dry forests.
It's not just trees that affect signal, either. Predicting radio propagation is a very tricky problem.
No. (I can't believe people worry about this, and then light up a smoke.)
First, The signal power output of the cards is typically 100mW. Compare this with a microwave oven that puts out about 1kW of power into a concentrated area. Microwave ovens work because 2.4GHz signals magnetically vibrate water molecules so fast that they can't keep up, and lag. The end result is that the molecules absorb signal energy and release it as heat. Actually signals from about 2GHz and up will heat water; 2.4GHz was chosen by manufacturers for statistical and practical reasons (i.e. it penetrates food a bit better).
So, even if you wrap your mouth around the small antenna of a particularly active wireless card, the teeny amount of heat that the water in your skin receives is almost immediately dissipated by your blood circulation.
Second, the radiation emitted by microwave antennas (and ovens) is non-ionizing radiation. This is unlike the radiation from X-rays and plutonium, so your DNA won't be altered and you won't get cancer.
Some mobile phone haters remain unconvinced. High voltage power lines have been shown to be mildly unhealthy by the tiny amounts of ozone and free radicals created by buzzing little sparks at junctions. So, it is possible that wireless network card energy could be the cause of an as-yet unknown chemical reaction in your body. You could always wrap some alfoil around your head and breathe through a gas mask.
Here is some FUD on the subject:
Fresnel (pronounced Freh-nel) noticed that objects placed on some ellipsoids with the transmitter and reciever at their foci, would strongly affect the signal strength received. Briefly, this is due to the slightly longer paths of reflected signals causing near-exact cancellation or doubling of the signal at the receiver. Therefore, objects in or near the signal path can improve or degrade the signal, depending on where it is. In general we want to keep the signal path clear.
The Fresnel radius is the width of the 1st Fresnel zone, found half way between the sender and receiver. You should make sure that this radius is clear of obstacles.
[links on how to compute fresnel radii]Jason Hecker says
They are compatible. You just have to subtract 3dB from the link budget. So if the helical is 18dB and the omni 5dB, for this sort of connection the helical's gain in the calculations should be reduced by 3dB to 15dB. Not too shabby. This assumes the omni is linearly polarised. Helicals don't distinguish between vertically or horizontally polarised antennas. They are seen as one and the same. 3dB is nothing to fret about most of the time. You will loose 3dB in a mere 5m of RG-213.
Helicals are great at talking to each other. Yagis can talk to omnis.
Yes, but the losses are large. You can overcome this with inline amplifiers (and the 4W EIRP limit is now 8W because you have two antennas.) [check this]
It's called antenna diversity. From the Cisco Aironet FAQ:
Q. What is the purpose of antenna diversity?
A.The use of two antennas on the Access Point is referred to as diversity. Diversity allows us to move away from reflections that cause us to lose or retransmit data.
Antenna diversity provides the receiver with a choice between antennas. Typically there is ciruitry inside the receiver that automatically switches to the antenna with the strongest signal level. i.e. the separate antennas are not agregated in their power, but rather provide the receiver with a choice.
The maximum height of antenna mast structures depends on your council, how close you are to an airport, if you are in an area zoned commercial or residential, and possibly how your neighbours feel about it.
In most municipalities and shires, towers on the ground are allowed up to a height of 10m; while masts on a roof are allowed a height of up to 3m above the top of the roof. You can usually get approval to go higher.
| Council | Free-standing | On-roof | Notes |
|---|---|---|---|
| Redland | 10m | 3.3m | DL phoned council |
| Logan | 10m | 3m | Adrian's email |
Also
We are going to use a 'backbone' topology. Read on.
A backbone in the mesh is a tree-structured connection of particular nodes that have nearly triple the amount of wireless hardware as normal nodes. The term was introduced by Brendan Hemmings.
Basically, most nodes would operate in infrastructure mode, with a backbone node acting as an access point and using an omni to talk to non-backbone nodes. The backbone nodes themselves connect to each other using point-to-point (ad hoc) links, probably with highly directional links.
This puts extra cost on those people who wish to be backbone nodes (as they need multiple cards and antennas) but it vastly improves the scalability and usability of the mesh.
It has been noted that infrastructure mode has only been reported reliably working with vendor supplied access points, i.e. home-made routers out of PC parts may not have driver support for infrastructure mode operation of the wireless network cards.
See also:
Peering, in the context of routing, is where two parallel but independent networks connect to each other for some benefit. For example, Telstra and Optus might both run fibre links up and down the east coast at their own expense. To improve communications between the users of the two networks and avoid a single bottleneck they "peer" at multiple points.
Peering adds complexity to routing tables while improving network utilisation. It's a symmetric agreement to route the other's traffic.
If there is sufficient external interest in services provided in the mesh, external networks may ask to peer with us.
Redundant links help in two ways:
IPv6 is the next version of the Internet Protocol (IP, as in TCP/IP). To avoid confusion, the traditional IP adrressing scheme that we all know and love is now referred to as IPv4.
IPv4 uses 32 bit addresses written like this: 130.102.176.108,
but IPv6 uses 128 bit addresses written like this: fe80:0000:0000:0000:0210:5aff:fe69:9744
(abbreviated as fe80::210:5aff:fe69:9744.)
IPv6 has been developed over the last decade basically because forecasters have predicted that the world will run out of IPv4 addresses. This is not the only reason why we are using IPv6. In the context of the Brisbane Mesh project especially, IPv6 has some other very nice properties. Let's look at some of them.
IPv6 has priority bits.
That is to say, there is a more sophisticated and more believable
packet priority scheme available for IPv6 than what came with IPv4.
Since congestion is an important problem
in wireless networks like the mesh,
IPv6-enabled software that appropriately sets packet priority
will generally improve network usability.
Here are some of the standard priorities and what they are intended for:
| Priority | Application category |
|---|---|
| 0 | uncharacterized (default) traffic |
| 1 | "filler" traffic (e.g., netnews) |
| 2 | unattended data transfer (e.g., email) |
| 4 | attended bulk transfer (e.g., FTP, NFS, HTTP) |
| 6 | interactive traffic (e.g., telnet, X, games) |
| 7 | internet control traffic (e.g., routing protocols, SNMP) |
IPv6 has a huge address space. While much of the 128 bit address space has been carved up already, what's left is still huge. (It's worth remembering that when IPv4 came in, 32 bits looked huge too!) There are automatic ways being developed in which nodes in an IPv6 network can obtain unique addresses, for example DHCPv6 is still a work in progress.
IPv6 supports classless/prefixed routing. If this is exploited properly (by allocating addresses in a hierarchical fashion) routing information traffic (that would otherwise have to occupy precious link bandwidth) can be kept to a minimum.
IPv6 is backward compatible with IPv4. Since IPv4 is so entrenched in so many systems, many programs simply don't (and won't ever) understand IPv6 addresses. There are a few ways that this is overcome by IPv6:
Most networking software already works with IPv6. For example, DNS servers already do IPv6. Most operating systems already have IPv6 implementations: most free unixen already support it natively, and so does Windows XP.
IPv6 is a new thing, and we are being brave [(trendy?)] in using it.
There has been some activity to obtain an block of 'real' internet-routable IPv6 addresses for the Mesh, but unfortunately it involves megabucks. (Robert Brockway is our current nominated contact with 6BONE and APNIC.) Fortunately, many other free alternatives exist, (such as 6to4 and freenet6) so we can use them for sure.
Further reading:
And even more:
Reasons against:
Its also worth noting that using IPv6 site-local addreses has the same drawbacks of RFC1918.
You could create a VPN between your friend's node and yourself.
[need someone to write a detailed description/how-to]Note that some internet providers prohibit you from sharing their service with others.
AARNet's policies prohibit general routing of non-university host data onto their network. If the Brisbane universities ever link up to the mesh they will still not route general mesh traffic onto the AARnet backbone. The best you can hope for is to be able to login to a uni host that you are authorized to use, and then connect to AARnet from there.
UQ ITS are thinking about this. [August 2001]
Yes. Most internet games use quite small amounts of bandwidth (because they expect a 56kb/s modem) so this should not be a hassle for intermediate nodes.
Some games may not understand IPv6, or may require IPX. There is software available to tunnel IPX packets over IP.
However, if the mesh gets particularly large, each intermediate node introduces more latency. Some games don't cope too well with high latency.
I don't know. Brisbane Mesh isn't about making money, anyway. As soon as money is involved we're looking at having to buy expensive licences. If you plan on participating with a backbone node, you can't charge people to carry their data through your node. Perhaps you can make money by providing useful services at your node?
Yes, as long as you route wireless mesh traffic in the same way as you'd expect the mesh to route yours (i.e. for free). Of course, that doesn't mean you have to give everyone free internet access, just that you make routes across your wireless infrastructure available to the mesh for free.
If your wireless nets are separate, you could certainly use the mesh to connect them, but reliability cannot be guaranteed.
Sure. You can host anything you want on the mesh. Only if it's X or RC rated porn or something else illegal in that context will you have problems.
This doesn't make sense in a non-profit, community-driven system. With sponsorship usually comes conditions. If anyone in the mesh did accept your sponsorship/purchase they would have to be acting independently and their obligations could not be expected to apply to the other participants.
However, condition-free "sponsorship" -- namely, non-cash donations -- are gratefully accepted via the router hardware pool.
The main mailing list is not a moderated forum. This is good because in my (David Leonard's) opinion,
It is also bad because it leaves the list open to "noise" from sources inluding: spam, viruses, morons. (Although some forms of protection are in place: hidden subscriber list, UQ's anti-virus mail gateway, a spam detector, and a stated list topic.)
Ideally, a good moderator would improve the quality of the list, but they would have to satisfy me that they
If ever anyone of these qualities steps forth,
volunteering to be a moderator, I (David Leonard)
undertake to immediately create a moderated
version of the list. That is, construct a read-only list (mesh-moderated) which carries the traffic from the original list, but filtered through the moderator.
To date (May 2002), the list traffic has remained mainly on topic and the odd spam and off-topic flames haven't been overwhelming.
The main list is a mixture of planning and technical enquiry, and it would be helpful also if everyone has read this:
Here are some reasons:
The problem of filtering out spam and viruses correctly is a difficult one, and is correctly identified as just another form of the moderating problem. (See previous question on moderators).
For now, a deliberate policy of permissive access to the mailing list is maintained. It is the list owner's belief that it is more important and more efficient to let people speak freely than to apply some ill-conceived form of censorship that could impede or inappropriately obstruct messages of valuable content.
Still, if you wish to set up a second-stage filter on the mailing list (i.e. one that automatically drops spam and so forth) a great many people would be interested in subscribing to it!
Some other mailing lists re-write the Reply-To header so that replies go back to the list. This is wrong. There is no good reason for it.
When you receive a message via the mailing list, its sender is (usually) a single person. If you wish to reply to that person privately, then you reply as normal. If you wish to reply to that person and Cc the list, then use the group-reply feature of your mailer. That is what it's there for.
If the Reply-To header had been re-written to point back at the list it
See also:
The mailing list service providers (ITEE) have installed SpamAssassin.
The filter software at ITEE is configured to bounce mailing list
messages with a spam rating of 8 or higher (as at June 2002).
You can see the filter working as it inserts a
X-Spam-Level header indicating how suspicious-looking it is:
X-Spam-Warning: This message may be SPAM X-Spam-Score: 9.9 X-Spam-Level: ********* (9.9) X-Spam-Tests: PLING,CHECK_OR_MONEY_ORDER,ONE_TIME_MAILING,CLICK_BELOW,NO_DISSAPOINTMENT,FOR_FREE,CASHCASHCASH X-Spam-Report: SPAM: 9.9 hits, 8 required;
If 8 is not high enough for you, you may be able to configure your mail reader to drop messages that reach a lower threshold of your choice.
Unix users can use a procmail filter;
Microsoft Outlook users can do the following:
Note that these are instructions for Microsoft Outlook. Outlook Express (a different program) is too crippled for this to work.
The system administrators set and monitor their suggested threshold level (8 in the above case) and you can rely on this if you wish, or insert rules that depend on the number of stars.
David Leonard first drafted this FAQ and drew upon the mailing list for many of the answers. Significant extracts are cited with links to the original messages.
Brendan Hemmings and Adam Secombe contributed answers to the Connector section.
Chris Pascoe sent in corrections for some of the technical parts.
Various other people have suggested and/or made minor changes.
Anna Gerber now tries to keep the FAQ up to date.
Mail Anna Gerber.
The information in this FAQ is not a substitute for legal advice or for your own research. Brismesh Inc. and the University of Queensland ("we") make no representation about the accuracy or authenticity of the information expressed above. We do not endorse activities implied or expressed in this document. You are advised that all such activities may involve risk, and that we accept no liability for any such actions that you may take in following them. All opinions expressed are those of the respective authors' and do not represent those of the University of Queensland.
Last updated: Mon Oct 18 09:15:07 2004
Corrections and additions to this FAQ should be sent to
Anna Gerber.
If you want to directly edit the FAQ,
download and edit this source file and then mail it
(or a "diff") back to me.