Wireless bridge

Fri Jun 19 12:46:19 UTC 2009

Peter, to follow up on a few of your RF questions here:

The idea behind the Fresnel zones is that objects (larger than one 
wavelength, a few centimeters at the frequencies we're dealing with 
here) within the zones will reflect the incoming radio wave from the 
transmitting radio. As seen by the receiver, there will be two signals- 
one coming directly from the transmitting antenna, and the reflection 
coming from the object in the Fresnel zone. The reflected signal, having 
a longer overall path length, will be slightly out of phase compared to 
the direct wave, and will destructively interfere with the direct wave, 
lowering the overall received power level seen by the receiving radio. 
This is called multipath interference. Therefore, unless you're using 
very high gain antennas (large parabolic dishes) with high directivity, 
you won't gain anything by pointing them at the sky or away from the 
object in the Fresnel zone. You'll lose more signal by mis-aiming the 
antennas than you will lose from the multipath interference.

Regarding antenna polarization, your flat panel antennas are certainly 
polarized and must be oriented in the same polarization at each end.

Finally, if you have a way to check the received power level at your 
existing radios, you will want to adjust the transmitter output power of 
each end so that the received power is within a reasonable range. 
Generally speaking, for a link of that distance, you should aim for 
something in the -60 dBm range. Anything hotter than a -50 and you start 
to get into front-end overload territory, and anything weaker than a -70 
and you're beginning to run on thin fade margins.

Also, I disagree that shielded Ethernet cable is unnecessary. For the 
very low additional cost of shielded outdoor cat-5, it's well worth your 
effort if you're running new cable. Of course every installation is 
different, but why risk ethernet errors due to some large air 
conditioner or something on the roof spewing EMI?

Patrick Shoemaker
Vector Data Systems LLC
shoemakerp at vectordatasystems.com
office: (301) 358-1690 x36
http://www.vectordatasystems.com

> Message: 12
> Date: Thu, 18 Jun 2009 21:46:08 -0400
> From: "Peter Boone" <NANOG at Aquillar.com>
> Subject: RE: Wireless bridge
> To: <nanog at nanog.org>
> Message-ID: <23ab01c9f07f$b7aa6480$26ff2d80$@com>
> Content-Type: text/plain;	charset="us-ascii"
> 
> OK, from reading all the excellent feedback I've got on and off list I've
> attempted to compile a "quick" summary of findings/ideas/products so far.
> 
> - RouterBoard is no good for this type of application.
> 
> - Get a unit with radio/antenna integrated, PoE from inside the building
> (outdoor rated cat5, shielded I assume), lightning suppression for the PoE
> (properly grounded), and ensure the mast is properly grounded.
> 
> - Get off the 2.4 GHz range. Move up to 5. As for licensed vs. unlicensed,
> I'm getting mixed input. I'm fairly certain that if the price is right and
> the frequency is 5GHz+, it won't be a factor. Also, I'll be very glad to
> separate the bridge from the client access points so that allows for more
> options. Every solution at this range can easily do 20+ Mbps so throughput
> is no longer a factor.
> 
> - Products that support ARQ are highly recommended.
> 
> - I'm hearing the same products mentioned over and over:
> 	- Motorola
> 	- Ubiquiti
> 	- Aironet (Cisco)
> 	- Aruba
> A number of individuals recommended products from other brands at low cost
> that meet these mentioned requirements too.
> 
> 
> I'm not going to bother with a spectrum analyzer. In the current
> implementation we tried channels 1, 6 and 11 for a few days at a time and
> found 1 to be the most reliable. Done. At this point an analyzer will tell
> me what I already suspect: there's a problem.
> 
> I've researched the Fresnel zones and calculated out a few things with rough
> numbers and worst case. For one, the Fresnel zone is disrupted most if the
> obstruction is closer to the endpoints (e.g. antennas). In this case, this
> is fine as the antenna are mounted at the outermost corner of the buildings
> as close as possible to the other buildings, approximately 3 floors in the
> air. Other buildings become a factor near the middle. Based on channel 1's
> wavelength of 0.12438 m, and assuming 1 km apart (for simplicity sake. It's
> actually less), the Fresnel zone is largest in the center at approx 5.6 m
> radius. That could definitely be obstructed by rooftops, I'll have to take
> another look though. This radius cuts in half when the frequency is doubled,
> thus more evidence in favour of the 5 GHz+ range. Cool. Or we could just go
> with a good line of sight optical solution but they look too expensive, and
> this area can have very unforgiving fog/wind to disrupt things further. What
> if we tilt each existing antenna up towards the sky 10-20 degrees? Please
> correct me if I'm wrong.
> 
> The current antennas are plates. I'm pretty sure they are polarized. I used
> to have a product sheet on these but a Google search doesn't turn up any
> useful results anymore (SmartAnt PCW24-03014-BFL). The way they are mounted
> to the poles might make it difficult to try rotating them 90 degrees, but
> worth another look. The coax between the AP and antennas are no longer than
> 30 feet. I've often wondered if a Pringle or Coffee Cantenna would work
> better than these!
> 
> 
> For right now I'll have the coax line and ends inspected for
> damage/softspots, check the grounding, and cover/re-cover the ends in large
> amounts of rubber/electric tape. I think we might try the Ubiquiti Bullet2
> for approx $100 per side (PoE supply/lightning suppression, wiring included)
> and see what happens! If that doesn't work, no major loss and we'll move up
> to something more serious (the PoE and wiring will already be ready to go).
> I will have to look into pricing on some of these suggestions and figure out
> if we should even bother getting a Bullet but instead go straight to a
> better all-in-one solution.
> 
> Thank you guys very much for the tips. Feel free to keep them coming!
> 
> Peter