Hello. I've been experimenting with mesh nodes using Ubiquiti hardware and small EnGenius EAG-2408 8dB collinear antennas, mostly with the goal of introducing the idea to my club, the Ski Country ARC. We have several repeater sites that have good line of site but are somewhat far away from anything, but centrally located. My hope is to get enough interest from the club members to start to build a mesh network and have a few "hub" sites for big hops between towns.

We're all fairly spread out from Vail and Aspen to near Grand Junction CO and many points in-between. There's a lot of potential for building out a good network thanks to the "natural towers" all around us, but a lot of wilderness between the high locations and the actual users too, so high gain antennas will be essential if we want to take advantage of these locations. At the same time it would be preferable to keep the high locations as open as possible, and I think that would lead to omni directional antennas, at least until we know if the network will be used or not.

For example, I see a 15dBi omni that seems like it would work OK, but at the same time I'm sure the noise floor would be much higher when compared to using (for example) 3 or 4 16dB sector antennas, which might end up with an unusable node.

What has been the experience with omni antennas over sectored or highly directional antennas? Do you find that the mesh concept falls apart when long distances enter the picture?

A lot of this will need to come from your own deployment needs to some degree, however in the WISP (Wireless Internet Service Provider) world  Sectors always win out over OMNI's

1) With an omni the central hub can hear 360 but the edge nodes can not hear every other edge node.

What happens is you end up with CSMA Hidden Transmitter Syndrom where node B is your center, and node A cant hear C    A and C both transmit garbeling the packet when it gets to node B (same as in 2m packet hidden transmitter problem)

This can repeat many times depending on how big the sending packets are and the backoff time calculated.

With sectors (especially if each one is on a different set of frequencies) you can actually avoid this by one sector hears Station A and the other can only hear Station C   both transmissions can come in.

2) As you noted noise floor will likely be higher on an omni, even if it is not however, noise floor only tells you part of the equation,   If a wifi device on ANY angle of the omni (say 180 degrees away from the node like in the example above)  its transmissions while not intended for the mesh can garble a weaker node coming from another direction.

Sectors help reduce all this as you only have to deal with 60-120degres (1/6-1/3 the potential) (average depending on sector beamwidth) of the interfering stations on any node

If you do go omni make sure to look for a Horizontal or better yet a Dual Polarity Omni, they are more expensive than a vertically polarized omni but you can gain 30db of noise reduction by being in the Horizontal channel.  Just be preapred you may need to replace it in the future if you have a lot of noise around or start adding more users. 

3) Sectors often have more gain compared to an omni, this can be a good and a bad thing , its good in that stations further away can get in and bad in that it means noise can also get in easier too. The gain should generally be worth having though as a few DB may be the difference between a 1.2 and a 2.0 LQ.

Personally when I look at BBHN (due to my terrain here in California  unlike in Austin where it was first created) I don't look at it so much creating a mesh topology (everyone talking to everyone direct)  I see it  creating a star network with possible redundant paths and the mesh protocol is used to share information about who is around removing the need for a central person to manage everything. So even in a star network it still makes life easier, and if people get enough nodes out  you could connect multiple stars on 2.4 and with considering a 5.8 backbone giving you multiple access routes....

I've been starting with NanoStations for my first few field sites (with a Rocket and  Sector on my first hill)      With 2 NanoStations at each end user I can reach the main node, and can design so that we feed to another user  or  double cover an area for redundancy.   Eventualy with enough nodes  even though I only have 60 degree on each NSM2 of beamwidth I hope we will see multiplerouting paths occur.

I see us trying in our area to make the network take shape as 2.4ghz access layer (few miles or so depending upon coverage), a 5.8 ghz backbone (link the access layer sites together)  on a major hilltop.  so  LARGE HILL--5.8--- small hill --2.4 -- user ---other users meshing 2.4 ---- maybe when the network gets big enough 2.4 merges and if we loose a 5.8  we have a 2nd path --

End of the day ANY NODE is often better than NO NODE  so while an omni may not be 'ideal' for peak performance  its sure better than not having any network at all.

Attached is an example of star network we are exploring for San Diego County.  Note the use of sector antennas at the higher-ground nodes and the cross-banding that occurs at each level to off-load data from the RF channel coming from "lower" nodes with data heading toward the "higher" backbone.

Since the higher nodes need to cover wider areas (through the use of sector or vertical antennas) the lower nodes always carry the burden of path budget. Therefore lower nodes will generally always use antennas with greater gain.

Note that this network has not yet been constructed, so I offer it up only as an example in support of Conrad's earlier post... Comments are welcomed.

Andre, K6AH

Attachments
 Node Drawings.jpg [75 KB]