DIY high gain omni antenna

Edmund <nomail@hotmail.com> hath wroth:
>After I have successfully made some wifi antenna's I now want
>to make a high gain omni for a hotspot.
>What is the highest gain I could get with what design?
>There are some commercial types sold as 12 dB gain???
>but what I found on the internet to make yourself is
>7 dB tops.

Commercial omni antennas can be found as high as 15dBi gain:
http://www.fab-corp.com/product.php?...cat=250&page=1
http://www.hyperlinktech.com/web/hg2415u_pro.php

Past about 8dBi, omni antennas become very tricky to build. The
problem is that construction is rather critical. It's also very
difficult to test the results.

Besides maximizing the gain, there's the problem of downtilt. The
above 15dBi antenna has a vertical beamwidth of 8 degress and almost
no downtilt. If you mount that on a tall building, it's useful for
talking to the horizon, but not to users below the antenna. In
general, an end fed vertical colinear antenna will have some uptilt of
the beam. Mount the antenna on a horizontal yardarm, and there's even
more uptilt. Unless the antenna construction includes some provisions
for downtilt, you may end up sending all your signal into the sky
instead of the ground (where the users are located).
| http://www.terabeam.com/support/calc...a-downtilt.php

Unless you really need all around coverage at the horizon, I suggest
you think about sector antennas.
http://www.superpass.com/2400-2483M.html
http://www.hyperlinktech.com/web/ant...out_sector.php
If you wanna build your own, see:
http://pe2er.nl/wifisector/
http://www.qsl.net/yu1aw/vhf_ant.htm
http://www.brest-wireless.net/wiki/materiel:amos

--
Jeff Liebermann jeffl@comix.santa-cruz.ca.us
150 Felker St #D http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann AE6KS 831-336-2558

Hi all,

After I have successfully made some wifi antenna's I now want
to make a high gain omni for a hotspot.
What is the highest gain I could get with what design?
There are some commercial types sold as 12 dB gain???
but what I found on the internet to make yourself is
7 dB tops.

Thanks

In message <n7oie21d5h0de9kao2atv8nq5074ndt04u@4ax.com>, Jeff Liebermann wrote:
> Edmund <nomail@hotmail.com> hath wroth:

> Commercial omni antennas can be found as high as 15dBi gain:
> http://www.fab-corp.com/product.php?...cat=250&page=1
> http://www.hyperlinktech.com/web/hg2415u_pro.php

I found these sites but what design are those?
>
> Past about 8dBi, omni antennas become very tricky to build. The
> problem is that construction is rather critical. It's also very
> difficult to test the results.
>
> Besides maximizing the gain, there's the problem of downtilt. The
> above 15dBi antenna has a vertical beamwidth of 8 degress and almost
> no downtilt.

Yes I thought of that too after I posted this message :-)

> If you mount that on a tall building, it's useful for
> talking to the horizon, but not to users below the antenna. In
> general, an end fed vertical colinear antenna will have some uptilt of
> the beam. Mount the antenna on a horizontal yardarm, and there's even
> more uptilt.

That is kind of interesting, what about mounting such an antenna
upside down?

> Unless the antenna construction includes some provisions
> for downtilt, you may end up sending all your signal into the sky
> instead of the ground (where the users are located).

What can be done to improve or create some downtilt.

> | http://www.terabeam.com/support/calc...a-downtilt.php
>
> Unless you really need all around coverage at the horizon, I suggest
> you think about sector antennas.
> http://www.superpass.com/2400-2483M.html
> http://www.hyperlinktech.com/web/ant...out_sector.php
> If you wanna build your own, see:
> http://pe2er.nl/wifisector/
> http://www.qsl.net/yu1aw/vhf_ant.htm
> http://www.brest-wireless.net/wiki/materiel:amos

Maybe I will start with making a simple one like this:
http://wireless.gumph.org/articles/homemadeomni.html
Does it help to extent such an antenna to lets say 16 sectors?
Also they mention some brass tube design but I can't find info
about that, if it can improve the sensitivity I like find some
details about it.

Thanks for the quick reply

Edmund <nomail@hotmail.com> hath wroth:

>In message <n7oie21d5h0de9kao2atv8nq5074ndt04u@4ax.com>, Jeff Liebermann wrote:
>> Edmund <nomail@hotmail.com> hath wroth:
>
>> Commercial omni antennas can be found as high as 15dBi gain:
>> http://www.fab-corp.com/product.php?...cat=250&page=1
>> http://www.hyperlinktech.com/web/hg2415u_pro.php
>
>I found these sites but what design are those?

End fed vertical collinear. Sorry, no plans for anything better than
8dBi. Please don't build the one consisting of alternating pieces of
coax cable. The antenna is twice as long as it needs to be for a
given gain. The antenna also has the same loss as the coax cable it's
made from, which costs a few dB gain. If you're thinking of building
your own 15dBi omni, you'll find that the bandwidth is sufficiently
narrow that you'll need some test equipment to get it right. The
reason can, patch, and biquad antennas are fairly easy to build is
that they're broadband devices and therefore can be built rather
sloppily and still work. Even badly built, they're still better than
the stock antennas that come with most cards and wireless routers.
However, that changes when the gain increases. Cut tolerances become
critical and details that are often ignored are no longer
insignificant. I just tried some assorted antennas on my Wiltron 610d
sweep generator and VSWR bridge. The one's I built are fairly awful
as compared to the commercial antennas. Photos later...

>> If you mount that on a tall building, it's useful for
>> talking to the horizon, but not to users below the antenna. In
>> general, an end fed vertical colinear antenna will have some uptilt of
>> the beam. Mount the antenna on a horizontal yardarm, and there's even
>> more uptilt.
>
>That is kind of interesting, what about mounting such an antenna
>upside down?

That sorta works and I've done exactly that. I've also tilted
antennas in desperation. The usual problem is that the antennas are
not quite sealed and need a place for water to go. That's usually a
small hole in the base. Mount the antenna upside down and the pipe
slowly fills with water (mostly from condensation).

>What can be done to improve or create some downtilt.

Not much with an end fed collinear antenna. Lots with a dipole array
using phasing lines or tilting the individual dipole elements. I have
lots of books on the topic, but am too busy right now to find
references on the web.

>Maybe I will start with making a simple one like this:
>http://wireless.gumph.org/articles/homemadeomni.html

Ugh. That's exactly the type of antenna methinks is a problem. Good
luck cutting the sections to the exact correct length.

>Does it help to extent such an antenna to lets say 16 sectors?
>Also they mention some brass tube design but I can't find info
>about that, if it can improve the sensitivity I like find some
>details about it.

The brass tube version is more accurate. URL returns nobody home.
Here's some copies I found
| http://www.suntend.com.cn/How%20to%2...linear_omni%22
| http://wireless.gumph.org/content/3/...r_antenna.html
Note the balun on the base.


--
Jeff Liebermann jeffl@comix.santa-cruz.ca.us
150 Felker St #D http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann AE6KS 831-336-2558

Jeff Liebermann wrote:
> Edmund <nomail@hotmail.com> hath wroth:
>
>> In message <n7oie21d5h0de9kao2atv8nq5074ndt04u@4ax.com>, Jeff Liebermann wrote:
>>> Edmund <nomail@hotmail.com> hath wroth:
>
> End fed vertical collinear. Sorry, no plans for anything better than
> 8dBi. Please don't build the one consisting of alternating pieces of
> coax cable. The antenna is twice as long as it needs to be for a
> given gain.

OK what designs are there that gives a shorter length?
( with a vertical polarization )
Like this maybe http://cheatthezone.tripod.com/wireless/colinear.html

> The antenna also has the same loss as the coax cable it's
> made from, which costs a few dB gain. If you're thinking of building
> your own 15dBi omni, you'll find that the bandwidth is sufficiently
> narrow that you'll need some test equipment to get it right. The
> reason can, patch, and biquad antennas are fairly easy to build is
> that they're broadband devices and therefore can be built rather
> sloppily and still work. Even badly built, they're still better than
> the stock antennas that come with most cards and wireless routers.
> However, that changes when the gain increases. Cut tolerances become
> critical and details that are often ignored are no longer
> insignificant. I just tried some assorted antennas on my Wiltron 610d
> sweep generator and VSWR bridge. The one's I built are fairly awful
> as compared to the commercial antennas. Photos later...

Right I will start with a lower gain antenna then.

>
>> What can be done to improve or create some downtilt.
>
> Not much with an end fed collinear antenna. Lots with a dipole array
> using phasing lines or tilting the individual dipole elements. I have
> lots of books on the topic, but am too busy right now to find
> references on the web.

In the mean time I've seen designs with tilted dipoles.
>
>> Maybe I will start with making a simple one like this:
>> http://wireless.gumph.org/articles/homemadeomni.html
>
> Ugh. That's exactly the type of antenna methinks is a problem. Good
> luck cutting the sections to the exact correct length.

You are right, it won't be easy, or at least it will be time
consuming to cut these sections.
>
>> Does it help to extent such an antenna to lets say 16 sectors?
>> Also they mention some brass tube design but I can't find info
>> about that, if it can improve the sensitivity I like find some
>> details about it.
>
> The brass tube version is more accurate. URL returns nobody home.
> Here's some copies I found
> | http://www.suntend.com.cn/How%20to%2...linear_omni%22
> | http://wireless.gumph.org/content/3/...r_antenna.html
> Note the balun on the base.

This will be a lot easier to make, still quite some work to do
and I have to find cable and tubing from the right size.
I think it is a lot easier to make the other model and add
some sections to see how that works out in vertical gain loss.
However I don't understand the math they are doing :-)
In http://cheatthezone.tripod.com/wireless/colinear.html
the first section is 1/4 they say but it is 2/4 L.
And why the last part is shorter I really don't know.
And what if I want some additional sections, is that simple
3/4 for each section except the last one?
How about the direction of the "loops"
here http://martybugs.net/wireless/collinear.cgi
they are all winded in the same direction (right handed direction)
and an 3/4 wavelength.
While here http://home.comcast.net/~ross_anderson/HOCollinear.htm
they use 1/2 a wavelength and changed winding direction, although
that makes some sense to me, why isn't the distance between the
loops 1/2 wavelength or 1?
Not sure about the diameter of the loops either, is the length
of the wire from the loop 1/4? It seems to be :-)

Edmund <nomail@hotmail.com> hath wroth:

I found the "brass tube" collinear article:
| http://www.nodomainname.co.uk/Omnico...4collinear.htm

This is more like what's inside commercial antennas:
| http://mobileaccess.de/wlan/download...ction%20kit%22

>OK what designs are there that gives a shorter length?
>( with a vertical polarization )
>Like this maybe http://cheatthezone.tripod.com/wireless/colinear.html

Yep, that's much better. Basically, it's half wave radiators, with
half wave "delay lines" in between to keep the radiators in phase. The
coils can be almost any shape as long as they're a half wave long (and
they don't overlap). However, the antenna is missing a balun at the
base, and could probably benefit from some ground radials. It will
also have considerable uptilt, but at low gains, will not be much of a
problem. The cut lengths shown are really rough and will probably
result in an out of band tuned antenna. (I did the calcs for the
required precision in a previous rant). In other words, it's a
tolerable design for a small and simple antenna, but not so great for
a high gain omni.

Can I suggest an alternative? Try a Franklin sector antenna:
| http://yu1aw.ba-karlsruhe.de/vhf_ant.htm
| http://pe2er.nl/wifisector/index.htm
| http://yu1aw.ba-karlsruhe.de/Amos.pdf
| http://www.brest-wireless.net/wiki/materiel:amos
Typical is 13dBi gain, 120 degrees horizontal beamwidth, 5 degree
vertical beamwidth, and quite easy to build and customize. The
required balun is a challenge, but not impossible if you use 0.085in
or 0.141in semi-rigid coax. The antenna is symmetrical and therefore
has no uptilt problems. If necessary, it can be mounted at a small
angle to optimize coverage area. Three of these 120 degree antennas,
with a power splitter, can simulate an omni with adjustable downtilt.
Also, note the NEC2 plots and test results which means that someone
actually went through the trouble of calculating and measuring the
antenna.

>Right I will start with a lower gain antenna then.

Actually, what I think you need to do is figure out some way of
testing the results. What I've been doing (due to lack of suitable
test equipment) is setup an antenna range that is not affected by
nearby reflections or Fresnel Zone edge effects. There's an access
point on a mountain top about 5 miles away that I can see on my
spectrum analyzer and with Kismet. I've taken the time to make signal
strength measurements with some commercial antennas of known gains.
These are my reference antennas. When I check the gain, I first
compare the signal strength with my reference (which happens to be
8.0dBi). If Kismet shows my antenna under test to be 2dB better than
my reference, then it probably has a gain of 10dBi.

I recently added a Wiltron 610D 10Mhz to 4GHz sweeper and VSWR bridge
to my collection. About $500-$800 on eBay. Now, I can test for VSWR
and tune the antenna onto frequency.

>In the mean time I've seen designs with tilted dipoles.

Yep. That's one way. Different length phasing lines to each dipole
is another. Some antennas are made from brass or copper tubing, which
can be internally tilted.

>However I don't understand the math they are doing :-)
>In http://cheatthezone.tripod.com/wireless/colinear.html

What math? The numbers are rounded off to 1 significant figure.

>the first section is 1/4 they say but it is 2/4 L.
>And why the last part is shorter I really don't know.
>And what if I want some additional sections, is that simple
>3/4 for each section except the last one?

Good observation. The design is wrong. The length of each section
should be 1/2 wavelength long except the very top which is 1/4
wavelength.

>How about the direction of the "loops"
>here http://martybugs.net/wireless/collinear.cgi
>they are all winded in the same direction (right handed direction)
>and an 3/4 wavelength.

Well, you got me there. The design looks correct, but it's upside
down. The connector should be at the other end with the 1/4 wave
section at the top. It will work with the 1/4 wave section at the
antenna connector, but then it needs a ground plane, which is absent.

It's very much like a wire version of the alternating coax cable
vertical collinear, except that every other half wave section is a
replaced by a half wave coil. If you look at the plans for the coax
version, you'll see that the coax connector end is 1/2 wave and the
top is 1/4 wave. Weird. These are both easy enough to model. I'll
do it if I have time.

>While here http://home.comcast.net/~ross_anderson/HOCollinear.htm
>they use 1/2 a wavelength and changed winding direction, although
>that makes some sense to me, why isn't the distance between the
>loops 1/2 wavelength or 1?

That's a totally different idea. Those are horizontally polarized
loop antennas with 1/2 wave phasing sections in between. The loops
radiate, while the vertical sections do not radiate (much). The
vertical sections are usually coax cables, but can be done otherwise.
In the various end fed collinear designs, it's the vertical sections
that radiate, while the coils do not.

>Not sure about the diameter of the loops either, is the length
>of the wire from the loop 1/4? It seems to be :-)

The length of the loop is 1/2 wavelength. You can build it like a
hairpin if you want. You can also overlap the turns, but then the
mutual capacitance of the turns tends to shorten the loops. If you
build it out of coils, it's very important to keep the coil stiff and
immoveable as small changes in position have a big effect on tuning.
Best to use an oversized single turn or hairpin style.
--
Jeff Liebermann jeffl@comix.santa-cruz.ca.us
150 Felker St #D http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann AE6KS 831-336-2558

Jeff Liebermann wrote:
> Edmund <nomail@hotmail.com> hath wroth:
>
> I found the "brass tube" collinear article:
> | http://www.nodomainname.co.uk/Omnico...4collinear.htm
>
> This is more like what's inside commercial antennas:
> | http://mobileaccess.de/wlan/download...ction%20kit%22
That's OK, I wouldn't mind to make this but I like to know
what on what calculations all this is based. I seems magic to me.
In the feedline they use a 1/2 lambda = 61.5 mm and in the elements
they use 1/2 lambda = 56 mm???????? Beats me.
The we have the length of the coils with the straight ends which are
1/4 lambda to the middle of the coil? ( hope you understand what
I mean ) Where is the diameter of these coils comming from (6mm) ?
Then to top it off, the total length of the element part 1/2 lambda
=56mm + the 1/4 lambda of the coil part + the remaining part of
the coil seems to be 98 mm.
Not to mention all the different diameters of all parts.
I don't ask you to explain all this :-) I only say I don't
understand all this.

>> OK what designs are there that gives a shorter length?
>> ( with a vertical polarization )
>> Like this maybe http://cheatthezone.tripod.com/wireless/colinear.html
>
> Yep, that's much better. Basically, it's half wave radiators, with
> half wave "delay lines" in between to keep the radiators in phase.

So does that mean the radiators are the straight parts, they are not
1/2 a wave here.
And the delay line is that the coil? or the other way around?
The coil is not 1/2 wave either, it is 10 mm diameter which
comes close to 1/4 wave.

> The
> coils can be almost any shape as long as they're a half wave long (and
> they don't overlap).

Not overlap? like in a spring or the above model you mean?

However, the antenna is missing a balun at the
> base, and could probably benefit from some ground radials. It will
> also have considerable uptilt, but at low gains, will not be much of a
> problem. The cut lengths shown are really rough and will probably
> result in an out of band tuned antenna. (I did the calcs for the
> required precision in a previous rant). In other words, it's a
> tolerable design for a small and simple antenna, but not so great for
> a high gain omni.
>
> Can I suggest an alternative?

I will do that some time later, for now I prefer a really omni.
>
> Actually, what I think you need to do is figure out some way of
> testing the results.

I agree but that is not so easy here, there are a lot
of buildings here and no free sight in any direction :-(.


>> However I don't understand the math they are doing :-)
>> In http://cheatthezone.tripod.com/wireless/colinear.html
>
> What math? The numbers are rounded off to 1 significant figure.
>
>> the first section is 1/4 they say but it is 2/4 L.
>> And why the last part is shorter I really don't know.
>> And what if I want some additional sections, is that simple
>> 3/4 for each section except the last one?
>
> Good observation. The design is wrong. The length of each section
> should be 1/2 wavelength long except the very top which is 1/4
> wavelength.
>
>> How about the direction of the "loops"
>> here http://martybugs.net/wireless/collinear.cgi
>> they are all winded in the same direction (right handed direction)
>> and an 3/4 wavelength.
>
> Well, you got me there. The design looks correct, but it's upside
> down. The connector should be at the other end with the 1/4 wave
> section at the top. It will work with the 1/4 wave section at the
> antenna connector, but then it needs a ground plane, which is absent.

Now you really got me confused, the design is correct? Why?
They use 3/4 wavelength for the straight part and a diameter
of 15 mm or less for the coils. That gives a straight length
of 47 mm for the loop/coil.
>
> It's very much like a wire version of the alternating coax cable
> vertical collinear, except that every other half wave section is a
> replaced by a half wave coil. If you look at the plans for the coax
> version, you'll see that the coax connector end is 1/2 wave and the
> top is 1/4 wave. Weird. These are both easy enough to model. I'll
> do it if I have time.

Let me know what direction the coil/loop has to be then,
I wonder what comes out of this simulation.
>
>> While here http://home.comcast.net/~ross_anderson/HOCollinear.htm
>> they use 1/2 a wavelength and changed winding direction, although
>> that makes some sense to me, why isn't the distance between the
>> loops 1/2 wavelength or 1?
>
> That's a totally different idea. Those are horizontally polarized
> loop antennas with 1/2 wave phasing sections in between. The loops
> radiate, while the vertical sections do not radiate (much). The
> vertical sections are usually coax cables, but can be done otherwise.
> In the various end fed collinear designs, it's the vertical sections
> that radiate, while the coils do not.

Forget I asked about that before I give it all up :-)
>
>> Not sure about the diameter of the loops either, is the length
>> of the wire from the loop 1/4? It seems to be :-)
>
> The length of the loop is 1/2 wavelength. You can build it like a
> hairpin if you want. You can also overlap the turns, but then the
> mutual capacitance of the turns tends to shorten the loops. If you
> build it out of coils, it's very important to keep the coil stiff and
> immoveable as small changes in position have a big effect on tuning.
> Best to use an oversized single turn or hairpin style.

OK here you answered my previous question except for the fact
that a diameter of 10 mm or 15mm isn't even close to 1/2 a wavelength.