I'm looking to become part of a community "mesh", mostly as a matter of interest and partly to open a free AP in my neighbourhood. I've contacted the guy who runs the mesh and have discovered I'm 1.7km from one of his nodes.
I have clear LOS from one corner of my house; checked this earlier with a pair of binoculars.
He recommends I use a 19dbi for a reliable connection. Can anyone suggest a homemade antenna that will (or may) reach 19dbi? I'm thinking along the lines of a wok/strainer antenna but I'm not sure.
Unfortunately, a commercial antenna is out of my budget for now.
You need more clearance than optical line of sight. See:
At 1.5km, you need at least 6 meters clearance around the line of sight line at midpoint. Do you have that? Note that this implies that both ends of the link are at least 6 meters off the ground. If not, the Fresnel Zone will hit the ground at midspan.
Sigh. First, the way I understand it, mesh networks kinda look like a spider web, where connections arrive from all angles of the compass. What you're building is a point to point link. That's fine, but what happens when someone wants to connect to *YOUR* node, and finds that your antenna is pointed at a distant connection. Here's an example of one mesh network layout:
Where would you fit in with a highly directional antenna?
19dBi dish antennas are about $40 plus shipping.
That's about the price of the wok, strainer, and coax connectors.
I coudn't get an absolute gain measurement because my antique test equipment just wasn't accurate enough. However, I could run a comparison with other antennas of (allegedly) known gain. My reference antenna is a simple Maxrad 24009PTNF PCB loop antenna:
I've tested this antenna with a tolerable outdoor test range and found it to have an average gain of 8.3dBi at mid band (ch6).
Using the Maxrad as a reference, I compared receive levels on a spectrum analyzer from various local sources. I installed a fairly well calibrated attenuator in line with the higher gain antenna and used my HP 140T spectrum analyzer to match signal levels.
I tested PacWireless 19 and 24dBi antennas and found the midband gain to be 18 and 22dBi respectively. Not great, but close enough for my sloppy methods. I tried an Antenna Spec 8.4dBi omni, which was about
8dBi at midband. I also tested my various home made biquads, which did not look anything like my 4NEC2 models. One loser varied in gain from 6 to 10dBi across the band. The one's that worked (after tweaking) were about 8dBi at midband. Sorry, but I didn't have a cantenna, wok, or salad bowl available.
It's not that far..similar point-to-point links of 6km have been established in this area. I am also able to pickup a signal from his node without any kind of amplifier (via a builtin laptop wireless of all things) though it doesn't stick around for long.
Actually I do, but I see your point.
A second AP with omni antennae?
Cheers, will look into it. Yeah, a professional antenna is a damn sight easier to setup than a homebrew, and as you say (in the long run) cheaper.
Picking up the signal and maintaining a full time connection are quite different. For example, if your friend with the node is using an amplifier, his transmit range will be much farther than his receive range. You could hear him, but may not be able to talk to him. If you can talk to him, you may not be able to stay connected.
The laptop test is a good one as it indicates that you have a chance. The internal antenna laptop antennas usually have fairly low gain (0 to 2dBi), so a 19dBi antenna or better will certainly be a major improvement. It's also possible that you won't need a 19dBi antenna and can get away less gain.
The unknown here is interference. A high gain antenna has a big advantage, which could turn into a big disadvantage. The narrow horizontal beamwidth means that any interference to the side and back of your antenna will be greatly reduced. The bad part is that any interference along the line of sight whil be greatly increased.
Maybe. You can't just plug in a 2nd different antenna into the 2nd port on a typical routers diversity switch. That doesn't work. The explanation is a bit messy. See:
especially the Golf Course case study, where they have an omnidirectional antenna in one port, and a highly directional yagi in the other. The switching problem is for real.
A 2nd AP might be possible, and would certainly not have the diversity switch problem, but might be an issue with whatever mesh networking protocol and firmware you're using.
You could use a Wilkinson splitter/combiner. That will divide your transmit power equally between the two antennas. However, the receive power will not be split making this a fairly useful solution. The only catch is about 0.5dB additional loss through the splitter.
You can easily build one of these yourself:
May I suggest a compromise. Build a sector antenna. These have the advantage that they are very easy to build (if you can solder well). I've built a few and they are quite good. See:
antennas have a reasonably high gain, very narrow vertical beamwidth, but also have a very wide horizontal beamwidth. Two or three of these combined can make an excellent omni antenna, with the added bonus of facilitating down-tilt. You can start with one, and add sectors as required.
I like building my own antenna, but I have some of the (antique) test equipment necessary to do the job. Building antennas is easy. Testing, tuning, and optimizing them, is not.
Yeah, I suspected that might be the case. I thought it was a little odd that I could see the network but not connect.
That's what I was hoping to hear :)
I'm going to have a play around with a few homebuilds and see what I come up with. Then if the results aren't satisfactory I'll bite the bullet and buy a proper one.
After all, this is intended as a learning experience.
I was worried about the AP issue myself, but I reasoned I'd cross that bridge when I came to it. I understand the issue a good deal more too, and I'm going to put some more thought into it. Thanks :)
Thanks for the advice Jeff. Your replies are always well detailed and easy to understand, and it's greatly appreciated. You've certainly given me a few more things to consider and plan for - things I had previously overlooked.
Please note that I have a really bad attitude about mesh networks. I generally don't like them. I'll spare you my usual rant (mostly because I'm tired and wanna get some snooze).
for some interesting test results and statistics from the MIT Roofnet project. Not the large packet loss (delivery probability) and very slow average thruput. That's typical.
Here's another mesh network status map:
If you tinker with the web page, some nifty statistics appear.
This one is a bit tricky to use:
Click on the various names on the upper right of the map so that they switch to bold face type. The mesh will eventually appear, along with path statistics. The percentages shows are the packet delivery reliability. Lots of links show really high packet loss.
Anyway, get some numbers from your mesh network admin.