[CUWiN-Dev] Rugged Node Construction DOCO:

Sat Jan 29 14:56:35 CST 2005

Hi Carl,

Thanks for all the input, some quick comments below:

On Sat, 29 Jan 2005, Quantum Scientific wrote:

> Very thorough.
>
> A few suggestions:
>
> On Saturday 29 January 2005 12:18, Sascha Meinrath wrote:
>>        We recommend the NEMA-4x enclosures sold by Metrix Communications for use with Soekris 4526 boards.
>
> Have a look at:
> http://socalfreenet.org/node/view/267
>
> ... it's actually a (secret, shhh) product of:
> http://www.alcoa.com/afl_tele/en/product.asp?cat_id=74&prod_id=222
>
> Also there's:
> http://www.demarctech.com/products/reliawave-ode/reliawave-rw-ode-8585.html
> ... but it's way too expensive.  I use Stalin J-boxen myself (~$40), which come in white.  Most boxes come in gray;  light, light gray is the best match for the sky.

We only recommend equipment that we test in-house.  That said, once the 
initial documentation is finished, we'll probably want to include 
information from other suggested vendors.

>>        You must build your antennas for the band that you are using-2.4GHz (802.11b/g) or 5.8GHz (802.11a). Use an
>>        N connector for the connection between the cable and the antenna. The N connector is most appropriate for
>>        the high frequency signals used by 802.11x.
>
> Actually 'a' is 5.2GHz.  5.8GHz is also unlicensed, and is officially 
> the only freq you're supposed to point-to-point on, but range is not 
> good for allowed powers.  Who cares about the FCC anyway, pffft...  :j

I believe 802.11a is composed of three 100MHz bands -- 5.15-5.25GHz, 
5.25-5.35GHz, and 5.725-5.825GHz -- so we will want to clarify our 
documentation on this.

> *All* N connectors can handle up to 11GHz, and all are ostensibly 
> watertight.  But the ones with a solid ring (not split) can handle up to 
> 18GHz.
>
>
>>        Mesh nodes will typically use omni-directional antennas. This is recommended for most cases.
>>        Omni-directional antennas transmit and receive equally in any direction towards the horizon. This allows
>>        the formation of arbitrary meshes without having to aim antennas (or re-aim after a high wind). It also
>>        means that there may be "wasted" radio energy that causes interference for nearby nodes.
>
> One concern I have is one radio.  Because all radios must operate 
> half-duplex...  Node 1 transmits a packet to Node 2;  then Node 1 must 
> shut up while Node 2 passes that packet on to Node 3 (as all are on the 
> same freq).  In practice this means a 50% drop in bandwidth for *each 
> hop*.  The solution is two or more radios, but of course this greatly 
> adds to costs.  See MeshDynamics' website for an in-depth analysis.

Actually, MeshDynamics so-called analyses are full of holes, wrong 
assumptions, and misinformation -- it would be extremely shortsighted to 
rely on them for accurate throughput projections.

>>        We recommend the HyperLink Technologies HGV-2409U 2.4GHz 8dbi omnidirectional antenna for most generic mesh
>>        applications. This antenna features a flared base which makes it less susceptible to windshear than similar
>>        antennas from other manufacturers.
>
> I think Hyperlink has gotten difficult to deal with;  maybe they have too many customers.  I suggest the Superpass:
> http://www.superpass.com/SPDG16O.html    ($44)
>
> ... or 10 degree downtilt:
> http://www.superpass.com/SPDG6O-D10.html    ($74.50)
> if clients are below the node.
>
>
>>        Microwave signals are extremely dependent on line of site. The higher you can place your antenna,the more
>>        likely it is that you will achieve line of site between your nodes.
>>
>>        Generally, you can use an antenna mast on a rooftop to acheieve height. If you live in a tall building you
>>        may be able to simply place the node in a window. Ideally, you will be able to place your node on a radio
>>        tower.
>
> I recommend 35' above the ground.  If mounted too high or too low, 
> neighboring nodes may be above or below the pattern, especially if 
> close.
>
> Some mention should be made of what power level to use.  This is a 
> complex subject which would need to be discussed.  In summary, it would 
> be best if the meshing software could control the radio's power, either 
> manually or dynamically.

Yes, this is something we want to work on (just as soon as we can get the 
funding for it ;).

>>        There are several ways to mount an antenna mast on a rooftop. For flat roofs, you can use an excellent
>>        non-destructive flat roof mounting platform that is held in place by cinderblocks. There are also special
>>        mounts for gables and chimneys, as well as tripod mounts. Any hardware designed for television antenna
>>        mounting will suffice for a mesh node antenna.
>
> Tripod:
> http://www.radioshack.com/product.asp?MSCSProfile=745D84CBF04D14A48AA6FF9C89D722C0BA68C1B04FE384678A5285FCD6E056B17AF21627FDABE316B90B3C038D68EBD6B7F9F3BD1712EAA9951ACB2590A05C6517EFE46941FEFDD1985D4EFD6321F5E70B4DE9B6C1D45512DCD9FB3DBCACB947D7076BC4B63499596306E58DCCE63EABBB65646AE016C354FF5FACB7431DDA458E70AD5A313B21E2&cookie%5Ftest=1&catalog%5Fname=CTLG&category%5Fname=CTLG%5F003%5F001%5F003%5F004&product%5Fid=15%2D517
>
> Big stuff:
> http://www.valmont.com/asp/communication/specialty_structures/asp/components_catalog.asp?url=2
>
>
>>        We recommend the Radio Shack ratchet style Chimney Mount (cat no 15-839) and any 5 foot antenna mast
>>        typically sold for television antennas.
>
> Channelmaster makes excellent powder-coated masts which are inexpensive (Fry's).
>
>
>>        Lightning can travel down either of the 2 conductors in the antenna cable and, due to the design of the
>>        antenna, they can't both be directly connected to ground. To protect this signal from lightning you will
>>        need to install an inline gas-discharge lightning arrestor and connect that to ground.
>
> I would note that gas-discharge fails after a few strikes, and needs to 
> be replaced.  A quarter-wave stub is ideal in all respects, but I 
> haven't found any for less than $106 (Radiall), despite their simple 
> construction.  Maybe China is making one.
>
>>    Outdoor CAT-5 Cable and PoE Injector
>>
>>        The cable travelling from the node into the user's house is a CAT-5 ethernet cable, which also carries the
>>        node's power. Although it is more expensive than standard ethernet cable, it is important to get special
>>        outdoor CAT-5 cable. The outdoor cable's jacket will not break down in UV light and is filled with a
>>        waterproof gel that prevents condensation-related corrosion of the conductors.
>
> I suggest using solid-wire only, for higher current capacity, and note that the RJ45 connectors come in solid and stranded flavors as well.  I also suggest Cat5e for higher datarates;  If 802.11g is used, one should absolutely use Cat5e, not Cat5.  Cat5e is twisted-pair, so that any EMI/EMP is induced in both wires at the same time.  Damage to electronics comes when one wire becomes significantly different from the other in amplitude.  Finally, *shielded* outdoor is highly desirable, although probably asking too much.
>
> The POE should use *two* conductors each for voltage and ground, not one, for best current capacity.  There are 4 unused conductors in Cat5e ethernet.
>
> Hope this helps.
>
> Man, I wish I had a platform to run this on, but I absolutely need 
> NoCatSplash functionality.  Right now I'm investigating upgrading the 
> Locustworld platform to Debian Sarge, although I have serious doubts 
> about AODV.  If I were sure I could add NoCatSplash in a reasonable time 
> (never done it before), and if we were sure of a final platform, I'd 
> start work now adding it.

We definitely want to add functionality like NoCatSplash to the software 
-- it's something that would make the system much more useful for many 
potential implementers.  It would be great if you wanted to work on this 
aspect of things -- just let me know,

--Sascha

-- 
Sascha Meinrath
President                 *   Project Coordinator   *   Policy Analyst
Acorn Worker Collective  ***  CU Wireless Network  ***  Free Press
www.acorncollective.com   *   www.cuwireless.net    *   www.freepress.net