Given two vanilla wifi radios (say of the wrt54g class in tx power and rx sensetivity) what sort of range can I expect for rooftop, line of sight use with a pair of high-gain 15dbi omni's (eg. Hyperlink HG2415U) ? They won't be obstructed, but the antennas won't really be much above the roof-lines either.
Does anyone have any words of advice about rooftop wifi antenna placement, anchoring, grounding etc? I'm lead to believe that 15dbi
8-degree omni's are a bit touchy on how they get mounted. The MIT roofnet folks had pretty mixed reviews about using the narrower angle, higher gain omni's saying that in practice they tended to need straightening after strong winds. Before I found that info I was planning to simply bolt the antenna to an existing steel antenna pole that was mounted with two W-shaped gable mounts. I'm now wondering if I need something more rigid.
I prefer my radios chocolate flavored. If you insist on vanilla, at least try chocolate chip flavored.
That's easy but depends on what speed you're expecting. The receive sensitivity varies with connection speed. See the FAQ at:
trick is to get 20dB of fade margin. Grinding the numbers for a pair of vanilla flavored wireless bridge radios.
TX power = +15dBm TX coax loss = 2dB (1ft LMR-240 plus a mess of connectors) TX ant gain = +15dBi Distance = unknown RX ant gain = +15dBi RX coax loss = 2dB (same at other end) RX sens = -84dBm (at 12Mbits/sec) Fade margin = 20dB
conveniently get 1 mile range. You can go somewhat farther with bigger antennas or slower speed. For example, if you go to 24dBi dish antennas as both ends, the calcs show that you can go 8 miles with everything else being the same. As a rule of thumb, 6dB gain is worth double the range. 12dB gain is 4 times the range. Etc.
That's obstructed. You need *MORE* than line of sight at RF frequencies. You need at least 0.8 times the Fresnel Zone diameter at midpoint. See:
1 mile, you need a radius of 19ft around the line of sight at midpoint. For 8 miles, you need 53ft radius. Skimming the rooftops might work if the roofs are wood or shingle. However, if there are trees, masonry construction, or steel buildings, forget it.
Yes. However, it's difficult to give general advice. The basic question is whether you can mount it on the building, on a bracket of sorts, on a tubular pipe, or on a tower. Give me some idea of the roof constuction and height involved and I can offer some hints.
Omni's suck for point to point for exactly that reason. The vertical radiation angle is far too narrow. Use a dish, sector antenna, or panel.
I think they were using the typical Antenna Specialists vertical that has a right angle aluminum bracket to mount it. That's fine for VHF and UHF antennas, but the requirements of 2.4GHz antennas to be exactly vertical makes this bracket useless. Also, the typical rooftop pole used for home mesh networks is not known for its stability or precision.
Those area actually fairly rigid for an omni. The problem is getting the mounting pipe exactly vertical. There's no easy adjustments once the bolts go into the side of the house. That leaves shims and pipe benders.
The side mounts are also a real problem for big dish antennas. They're just too heavy for such a mount as there's no vertical supports.
It looks like I could buy myself another 12db on the RX side (-96dbm vs -84dbm) and another 9db on the tx side (24db vs 15db). That 21 db should be good for 3.5 doublings in distance - call it
(* 2.0 2.0 2.0 1.414) 11.312 ; ~11x further
The houses are mostly stucco over chicken-wire and concrete roof tiles painted to look like clay.
So it looks like the lower half-cylinder of the fresnel zone will get clobbered. Is this equivalent to half the signal getting lost or would it be worse? If it is only half the signal getting lost, what's another 3 db among friends?
Concrete roof tiles, stucco building, wide eves with a 2x12 beam running along the end of the roof.
I've already got a weather station mounted on an steel antenna pole thats lag-bolted to the 2x12 near the apex using 2 of those bent steel legs. I think it was the same RCA badged tv-antenna mounting brackets as in the following kit.
Because of the way the legs are bent, there is considerable flex. I have no doubt that I could mount it within a few degrees of correct and then just gently bend it until it was perfect. Of course that is what worries me a bit too.
I was actually going to use a lightweight 15dbi yagi for the laptop antenna. Its only 18" long and weighs 3oz. I figured the weight and size were ideal for laptop use.
Sounds like I should probably just use the 4ft pipe attached to my brackets.
No, don't go to 802.11b speeds. The slowest 802.11g speed will give you better sensitivity, range, and reliability. Stay with OFDM.
Both should be possible with a proper antenna. Note that my calcs assume that the radio goes on the roof along with the antenna to minimize coax cable loses.
Reading between the lines, it appears that both these locations have existing wireless installations. However you didn't disclose the hardware and antenna types and gains. I can't do a calculation without these. Got photos? Got inside info?
If these locations (park or peak) have the usual 2ft long omni, it has a gain of about 7dBi. No clue on the radios. Many of these public access points are 802.11b only and are limited to 5.5Mbits/sec connection speed. That will limit the range somewhat due to lousy sensitivity.
If you find any errors, please either let me know or just fix them yourself. The FAQ is a user supported effort.
Yep. That's it. 1Mbit/sec is also a problem in that the packets are flying through the air for a longer time to send the same amount of data. That creates a higher probability that a noise hit or interference blast is going eat a given packet. 1Mbit/sec is not terribly reliable because of this. The problem is that with 802.11b,
*ALL* management packets are sent at 1Mbits/sec. That guarantees lots of loss even if you manage to get an 11Mbit/sec connection. With
802.11g, I think (not sure) that the management packets are sent at the connection speed.
Don't guess. That's my job.
Receiver sensitivity is measured at at a BER (bit error rate) of either 1 bit in 10^5 or 10^6 (depending on standard). One error every million bits isn't all that bad but you wouldn't want to operate there. Below the BER reference of 10^5, and drop in signal results in a rather drastic decrease in error rate. Looking at the chart on my wall, a 3dB drop in signal level will reduce the error rate to about
10^3. (It varies with different modulation types and speeds). At 1 error every 1000 bits, you're guaranteed to have an error in just about every packet. In other words, this is a threshold effect with
10^5 or 10^6 being at the knee of the threshold.
A 3dB drop in signal is really easy to produce. It can come from water incursion, rain, blockage, fresnel zone diffraction, reflections, and possibly the position of the moon. Please note that my calculations are seriously simplified and offer the *BEST* case model. It can only get worse as errors and losses are introduced. For example, I have yet to see a 24dBi dish that I actually can test with a resultant gain of 24dBi. Same with the +15 to +17dBm the typical router is suppose to belch. It's more like +13dBm on my junk test equipment. Everyone lies, but that's ok because nobody calculates.
I don't have time to tear these apart. I've only played with one for a few minutes. However, the trend for increased power output seems to be epidemic. I don't like it because it creates an alligator (big mouth, small ears) which transmits far further than it can hear.
That's only true if you have a symmetrical system, where you increase the transmit power and sensitivity on BOTH ends of the link. If only one side of the link has a big transmitter, then there's no guarantee that it can hear the other side.
Incidentally, the RX specs look a bit too good. I can usually get 2dB more sensitivity by simply tweaking the BER reference level. The numbers are very close to the noise floor of the receiver. I gotta grind some numbers to be sure. Let's just say I'm suspicious.
Concrete roof tiles are like a brick wall. Hell, they *ARE* a brick wall. Nothing goes through them.
I wish it were that easy. Fresnel Zone is a knife edge diffraction zone. The signal hits the edge of the obstruction and is diffracted away from the line of sight target. In extreme cases, it will setup reinforcement and cancellation interference patterns as in Newton's rings.
's_ringsThe important point is that it's not stable. As things move around, the destination signal strength varies radically. You might be lucky and find a location with a reinforcement peak, only to find it a null when someone moves their vehicle in a midpoint parking lot.
Concrete and chicken wire supported stucco are like a wall. Nothing goes through. You'll need clearance and altitude. Also, don't forget about trees. I'm sure there are some in the area. If they're in the way, they'll block the signal (about 10dB per "average" tree).
Retch. That's even more flimsy than the Radio Shack flavor. The problem is lack of vertical support. Also no adjustments.
That's an understatement. I used a cousin of this abomination to do a temporary 24dBi dish mount to a large pipe. Of course, during the week the wind decided to blow about 50knts. The mount literally twisted loose from the wind load on the antenna. Dish mounts have to be very rigid. You would do better with a satellite DBS dish mount.
maybe a replacement for your roof peak mount:
If you can bend the mount, then it's not going to stay put.
I don't like yagi's for 2.4GHz. Too expensive per dB of gain. Also too narrow a radiation angle and they get huge for higher gains. Double the length for only 3dB more gain. However, 15dBi for $30 is rather cheap. Still, methinks a similar gain panel would be better. How about 14dBi for $30?
Something like that. It really depends on the size, weight, and wind load of the antenna. Anything over about 10ft will require guy wires (yet another complication).
Search Google for "wireless antenna mast".
Drivel: Of course you could do something interesting like:
You're welcome! I'm glad my spare dgps radio went to good use. After dragging it up many of the local mountains along with a 4.5 AH 12v battery, I decided I needed to find a good home for it that didn't involve me carrying large hunks of lead and sulfuric acid up and down
Hmm. I hadn't considered that.
After years without a laptop (after my last one walked out my back door) I did finally get another one. One of the first field trips was to drive around to two of the close by free Fremont hotspots to see how usable they were. Simply running "iwlist wlan0 scan" in a loop as I drove to the sites showed that there were tons of wifi transmitters on every residential block. There were even a few hits on what I could have sworn came from a rent-a-cop car with a ssid of 'something-or-other-security'.
My thought at the time was: this is great, look at all the folks that might be interested in helping to form a mesh network. I see now I was looking at it from the wrong side. All this chatter is going to be a real problem.
I might need to do that. I don't mind the cost as much as the neighborly karma points I burn. I figured a small omni is going to raise a lot less ire than even a wire dish pointed at what looks like some poor person's roof.
Ah... dedication. I found some GPS boards that will belch raw data and are suitable for DGPS. I'll be installing one at the local ham radio repeater site shortly. I've been using the one at MBARI on Mt Toro, but the baseline is too far from Santa Cruz and their antenna is aimed over the bay, not the land. Yet another project.
Interference isn't fatal. However, substantial interference from multiple systems is a big problem. Your best defense is to narrow the antenna pattern to only pickup junk in one direction. That means no omnis.
Say hello to Jason if you run into him. I forgot which security company he works for in Fremont. I can provide contact info in you wanna meet.
Mesh networks only make it worse. The problem with these is that there are duplicate packets floating around. If a client is perhaps 4 hops away from the wired access point (where it hits the internet via conventional backhauls), there will be 4 packets flying through the air to deliver just one packet. That leaves 1/4th the airtime available for other users. I can rant on about the evils of mesh networks but not now. It's 11:30PM and I'm still in my palatial office.
That is a problem. The tin foil hat crowd seems to be everywhere. If you can live with less gain, a panel antenna will yield 19dBi of gain and not attract the lunatic fringe.
Huh? I guess that's me. 14 hour work day today. Dinner at midnight. Brain is gone. I guess I'm ready to answer questions.
Disclaimer: I are not an expert in GPS.
To do DGPS, you need a GPS that will belch individual satellite doppler delays (also known as raw data), and not NEMA-183 ASCII text. DGPS corrects on a per satellite basis and needs the raw data to do the corrections.
GPS accuracy is a subject of considerable contention. There are a bunch of web sites that offer test results for various technologies. One of many:
guess, opinion, and experience is: Uncorrected GPS with clear view: 5 to 15 meters (1 sigma) Uncorrected GPS with lousy view: 10 to 30 meters GPS with WAAS: 2 to 7 meters DGPS with 50 mile baseline: 2 to 5 meters DGPS with 1 mile baseline: 0.1 to 2 meters If you want cm accuracy, you gotta have a local DGPS transmitter. One big problem is that the final accuracy is totally dependent on the accuray of the DGPS antenna location. cm accuracy is worthless if the location of the correcting GPS is not located with the same precision. So, I get to do some precision surveying.
I plead ignorance. I was doing lots of commercial radio in the
1970's. I was also out of the country much of the early 1970's. To keep my sanity, I completely ignored ham radio until about 1993. I even intentionally let my ham license expire.
It's a common service. However, there are now several free and open DGPS systems on the air in the Monterey Bay area. There's also free post mortem corrections available on the internet. I have some applications in mind which I don't wanna disclose.
Those Marconi gps boards are one of the few I'd actually trust to do a good job on the corrections. The corrections from the moto oncore units seem pretty mediocre.
This whole thing is really meant to be a feasibility study for setting up a wifi community net. I just want something that I can maybe get a connection from in a few spots in town. I'm hoping that more folks might want to get in on it and it would be possible to expand the coverage as more folks entered the fold.
The reason I mentioned mesh networks is I'd like to avoid the wifi packets hitting the internet for now. That way the ISP's won't be getting pissed off as well as spammers / kiddies won't be beating down the doors to try their hand at a bit of semi-anonymous mischief. I figure folks can use their own internet connections if they want to route their own wifi packets to the net.
I do agree that mesh networks chew up quite a bit more air-time than doing it all in one hop and backhauling from there.
That sounds like a good idea.
I wonder if anyone makes patches out of aluminum to keep the weight down.
Yup. I was already planning for that. I'll probably pull the pc card out of the wrt54g and put it in a small weatherproof enclosure.
No, there is nothing at the other locations. Just spots I might want to hang out at with my laptop.
I was more interested in this card for its RX sensitivity. As you say Having a bigger tx only helps if the other side has one too otherwise the return packet isn't going to make it back.
I'm pretty sure the linux drivers allow one to set the tx power on these and other atheros-based cards. I'm not sure I'd want to pay the penalty of sucking another ~3 watts out of the laptop batteries to run this card.
Any idea what is causing all the power loss? We are radiating a claimed 300mW and chewing through 3w to do it. Thats only 10% efficient.
Elsewhere they claimed that they did their own low-noise front end. Well, I'll soon have one of these cards in my hand to play with. I'll try to do some A/B comparisons.
Sigh. I should have realized that it was essentially the same system as a diffraction pattern around an edge.
So it sounds like there is no choice but to get the Fresnel zone up and away from obstructions. That is going to be challenging.
Yea, I've been ignoring the trees issue. Figured I could always test to see how transparent they were later on. Perhaps it is time to buy that chainsaw.
The looks ok for the fixed installation. Its 2 lbs though. The yagi I found was only 3 oz. For the laptop that would be better. Although I assume there is nothing preventing someone from making a nice light aluminium patch at 14 dbi that's only a few ounces.
Now you mention this. It's always a good idea to offer some clue as to what you're trying to accomplish.
Hopefully, you'll get permission. Most hot spots don't want freeloaders.
More users mean more headaches and more intereference. Basically, you're building a wireless ISP (WISP). It has all the characteristics and requirements of a conventional wired ISP, with the added bonus of seriously unreliable connectivity and backhaul. That includes support, where you will get a midnight phone call from the neighbors asking if the internet is down. Are you sure you want to do this?
Incidentally, we've had various mustations of a neighborhood LAN since about 1975, when I started a bootleg neighbhorhood CATV system. At this time, we have wireless, coax, and CAT5 snaked through the trees and under the road. One thing nice about being in the forest. The trees hide my misakes and wires.
Ummm... wireless is bridging. There's nothing to stop any of your friend and neighbors from connecting your network to the internet. Any PC with a dialup connection can do that.
Well, yeah. However, that leave your wireless network as some type of dedicated game network. Good luck selling that idea. Everyone wants (free) internet access.
The standard construction method is a rather thick PTFE circuit board. One side has the patch etched into the board. The back is all ground plane and is covered with a vinyl sheet. The front is protected by a vacuum formed plstic sheet. I don't think aluminum will impte the weight.
If light weight is an issue, I have a design for 9dBi panel antenna made out of "foam board" and aluminum or copper sticky tape. It's not really ready for prime time, but can probably be convinced to function. Attaching the coax cable is the main problem.
That works quite well. Lots of examples on the web (some good, some rally ugly).
Sure. The RF power amplifier in any 802.11b/g have to have good linearity or they will distort the amplitude modulation component of the transmission. (1 and 2 Mbits/sec is pure FM with no AM). To gain the necessary linearity, one has to burn sme power in the amplifier.
10% is considered decent efficiency.
Incidentally did you noice the power consumption in various modes at:
One would expect the power consumption to higher with higher modulation rates. However, since the xmitter is on for a longer period, for constant data amounts, the average power consumption is higher for slower speeds.
I think it's more a matter of stating (a) theoretical performance of idealized hardware or sometimes (b) measured performance of an optimized prototype. In idealized conditions (e.g., electrical power source). Meaning that actual performance of real-world mass production hardware will always be less.
OK, lets take this one apart. It's a mesh network. 8 nodes and 20 users at this time:
which 7 are currently up. The nodes run the Quagga routing implimentation of Zebra on Debian Linux:
users are stock 802.11b/g and apparently no encryption, authorization or authentication.
I couldn't find any proceedure for joining the network other than "join the mailing list" and:
between the lines, it appears they are expanding slowly, welcome additional nodes, but would like to know who is connecting to their network, where they appear on the map, and what facilities they offer. User connections appear to quite open and do not require any form of permission.
Important information such as routing tables, IP address blocks, DHCP assignments, and Quagga configs don't seem to be publicly available.
get these, one must "check with Drew". Like I said, one should ask first before joining the network.
I'm confused as to the intent. It appeared to be open and free until I read their (MS Word) "business plan". |
Pricing Strategy section 4.5.1 which details that purchase of bulk bandwidth necessary to support a large number of users will require assessing the users. Bandwidth costs can easily be absorbed by the initial users on a small scale system. Not so once the system gets used and grows.
Incidentally, the cost of the backhaul bandwidth is what has totally stopped the growth of out local free wireless incantation: |
are only so many businesses and individuals willing to donate bandwidth for free.
8 nodes and 20 members in 3.5 years is not exactly my idea of successful. There are some built in problems to expanding wireless mesh networks. Metricom ran into a few of them. I've ranted on the topic in the past. Basically, they don't scale well at either extreme. With only a few nodes, the traffic tends to concentrate at those nodes preventing bandwidth distribution. At the opposite extreme, a large number of nodes create their own mutual inteference. It's necessary to limit the number of hops allowed (before hitting a wireline internet connection) because a store and forward network generates one additional packet for each hop, thus filling the air with duplicate packets. (Both Philadelphia and SF WLAN proposals started with 10 nodes per wired access point, and are now at 3 nodes per access point). Anyway, I can rant on the subject of why wireless mesh networks suck when I have more time to supply numbers and calcs.
Sorta. That's not exactly what I meant. The difference between theory and reality is in the minor details. My calculations ignore a substantial number of these details. Only a few of them are manufacturing tolerance accumulation and my crappy test equipment calibration. The real problems are when I ignore, terrain issues, fresnel zone effects, water vapor loss, foliage attenuation, additional connector losses, inversion layer effects, coastal zone effects, corrosion, coax cable water incursion, antenna misalignment, ground reflections, faraday rotation, polarization mismatch, radome loss, and a bunch of others. The point is that ALL of these are losses, not gains. My original calcs ignore almost all of these. Including them will always reduce the fade margin worse, and never improve it. Throw in some manufacturing variations, and it usually makes the fade margin even worse.
There are much better online link calculators available that include many of these factors:
(faster) However, these are far to complex for the typical rough guess if it's going to have a chance of working.