Wireless shot across the parking lot.

[Nathaniel Bezanson]

As you may know, i3 Detroit is working on expanding into a nonadjacent
building. It's a stone's throw away, literally, but there's no
practical cable path to extend the network. Wireless seems like it
should be reliable over such a short distance. Here are some
considerations in mind:

2.4GHz is predicatably smashed, even in our little industrial cul de
sac. Our own client access sits on 1 and 11, and the neighbor's loud
AP is on 6.

We have just-inside-the-wall mounting locations for the equipment, and
could do genuine outdoor line-of-sight if the gear's small and
unobtrusive.

The network link would ideally have enough bandwidth for bidirectional
video between the main space and the annex. You decide what that means
in practical terms.

Antennas for 5.8GHz are no harder to build than for 2.4GHz, but we
don't have any on-hand.

So, I'm mostly torn between two options: Either buy a pair of CF cards
and refurb a pair of Routerboards with some scrap 802.11a cards in
'em, build antennas, build mounts, build PoE injectors, and beg
someone to help with the software config. Or, just pick up a pair of
LocoM5 units for $65/ea and be done with it.

On the one hand, the DIY route would be a good exercise for someone
looking to gain experience with all those facets, and there'd be
abundant help available for each phase. On the other hand, the premade
units are faster in terms of becoming operational, faster in terms of
throughput, and already weatherproof. And it's just a few extra bucks
to keep a spare unit on hand, whereas building three of the DIY units
for redundancy's sake might never get done.

Thoughts?
-Nathaniel-

[Joseph C. Bender]

On 10/23/2012 11:39, Nathaniel Bezanson wrote:
>
> So, I'm mostly torn between two options: Either buy a pair of CF cards
> and refurb a pair of Routerboards with some scrap 802.11a cards in
> 'em, build antennas, build mounts, build PoE injectors, and beg
> someone to help with the software config. Or, just pick up a pair of
> LocoM5 units for $65/ea and be done with it.
>
> On the one hand, the DIY route would be a good exercise for someone
> looking to gain experience with all those facets, and there'd be
> abundant help available for each phase. On the other hand, the premade
> units are faster in terms of becoming operational, faster in terms of
> throughput, and already weatherproof. And it's just a few extra bucks
> to keep a spare unit on hand, whereas building three of the DIY units
> for redundancy's sake might never get done.
>

While DIY will probably work, getting the LocoM5 units will allow
faster implementation, be less maintenance intensive, and be easy to
repair (with a spare on the shelf) when one of the outdoor units fail.

The overall goal that I've had for the i3 network has always been to
keep it as maintenance free as possible, and I see getting the LocoM5
units fitting that goal the best.

--
Joseph C. Bender
jcbender at bendorius dot com

[Drew]

If you use two Cantennae, one on each end and pointed at each other,
it should not matter if you're using a frequency (2.4 GHz) that is
heavilly stepped on.

(Btw, I thought the space to be acquired was adjacent?)

[Drew]

[Nathan Warnick]

the adjacent space is being taken over by someone else. We are acquiring the upstairs space to the office next to that building.

We could just mount a bi-quad on the front of the building and point it at that building. We should have 2 or 3 WRT's in the wifi bin.

Though having a solid "omg the wifi is down, where'd it break" connection should be a priority.

--
You received this message because you are subscribed to the Google Groups "i3 Detroit Public" group.
To post to this group, send email to i3detroi...@googlegroups.com.
To unsubscribe from this group, send email to i3detroit-publ...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/i3detroit-public?hl=en.

[Joseph C. Bender]

On 10/23/2012 14:48, Drew wrote:
> If you use two Cantennae, one on each end and pointed at each other,
> it should not matter if you're using a frequency (2.4 GHz) that is
> heavilly stepped on.
>

It actually does have an effect. I've had to put links in place on 2.4
that were using far better antennas than a cantenna that we absolutely
couldn't reduce interference with. The shot was about the same
distance. Increased forward gain doesn't always mean reduced interference.

I'd mentioned doing this link on 5GHz a little while ago to Nate, and
honestly this is something that I have no intention of debating.

[Nathaniel Bezanson]

On Oct 23, 3:16 pm, "Joseph C. Bender" <jcben...@bendorius.com> wrote:
>  Increased forward gain doesn't always mean reduced interference.

Yeah, like when the interferer is sitting right behind your other
endpoint. Drew's point is well taken, it does help in most cases, but
not always to a useful degree.

On Oct 23, 2:54 pm, Nathan Warnick <digital...@gmail.com> wrote:
> We could just mount a bi-quad on the front of the building and point it at

And just paint the other building with signal, putting no hardware
over there at all? That has strong KISS principle bonus, and it's not
a bad idea as a fallback, but I think we need to do better for the
main link, for two reasons:

1: I failed to mention in the first post, there's a need for wired
clients in the annex. Putting a client bridge on 'em could work, but
2.4 still sucks.

2: Gain on the AP end is helpful, but it's limited because members'
laptops in the annex would have no additional gain on their end,
leading to hidden-node problems and reliability that decreases with
utilization.

3: The 20MHz channels on 2.4GHz are a limiting factor for a number of
reasons. Wide-channel modes aren't possible on 2.4 here because of the
existing spectrum use. The bandwidth available in 5.8GHz is delicious,
and completely obviates that concern.

Members' ability to make productive use of the annex is riding on this
link, and all of us potentially supporting it already have full-time
jobs elsewhere. It's worth investing money and/or effort in
reliability, and if we get a speed bonus out of it, so much the
better. I think a point-to-point shot is the way to go, and I'm quite
certain that 5.8 is worth doing. The only question, as I see it, is
whether someone's willing to cobble it together from parts-on-hand, or
whether we just buy something solid from the get-go.

I've got my finger on the Checkout button, but I think I'll hold off
for a few days. If a solution assembles itself before, say, Halloween,
win! Otherwise I'm just gonna buy the Locos and call it a day.

On Oct 23, 3:22 pm, bracestout <bracest...@gmail.com> wrote:
> http://en.wikipedia.org/wiki/RONJA

Definitely cool, but repeat after me: A URL is not hardware. (But can
you imagine a world in which it was? I'd just google for "veyron"
before driving to work and "unimog" before going camping. It'd be like
CBG's EveryMan but with physical objects instead of superpowers!)

If you have a pair gathering dust in your garage, by all means, they'd
make a great backup link!

However, I will buy a cookie for anyone who runs a pair of RONJAs or
similar, in parallel over different colors or polarizations. :)

-Nathaniel-

[Mark Lenigan]

Having tried the homebrew route for critical infrastructure several
times before, my vote is to use current production, off the shelf
hardware. I'm personally a bit leery of using 5 GHz ISM because I've
seen so many problems with it when I was working as a wireless
engineer. However, if the band is quiet locally, it shouldn't matter
over such a short hop. Just a question for my own information, but
have we considered the 900 MHz band? Ubiquiti makes Bullets which use
802.11 on 900 MHz. Just a thought.

Mark

[Toby]

I have 2 spare offset feed dishes if that would help. small
soupcantenna feed should be sufficient for off axis rejection.

[firemandave6024]

How many hams do we have as members? Is there any prohibition on using a non-US channel with a ham license, using MAC address filtering to prevent people from joining the point to point network? I know the traffic cannot be encrypted, but IIRC, we can use measures to prevent non-licensed people from using the network.

[Roger S]

Ah ... I think we only have one of the points in the point-to-point, so at the moment we don't need that link.