I'm getting inconsistent results with Linksys WAP54G over about 400m with bursts of high speed then periods of nothing.
Both Linksys are set to repeater mode.
Far end has cantenna on one of the aerials and the other aerial is removed. Ethernet to hub.
Near end have tried with both aerials on and also one aerial to cantenna with other standard aerial picking up from notebook.
(view with fixed font for clarity)
(( )) represent wireless signal
Home Notebook )) ((Near Linksys ))Cantenna or standard aerial
| | | ~300m | House Tree | | ~380m Over an office roof | Far Linksys )) Cantenna | Office Hub
I seem to get less dropouts when the near (home) linksys has the two supplied aerials fitted and placed in the window but a low speed. The fastest results are when one aerial is connected to the cantenna at home but then it will then drop/lock out.
The aerials are set to diversity mode and I wonder whether it can get confused as to whether it is setting the aerials speeds to the notebook or to the distant cantenna.
The "obstruction" of House and Tree is about 2 - 3 metres horizontally and the office roof is about 1m vertically.
It sounds like a weird config. Ordinarily when you put a WAP in repeater mode the ethernet is disabled but you say the remote site connects to a hub via ethernet. Im really surprised it works at all that way, but then again I aint seen everything. If you want wireless at the home end then I would use AP mode there and AP Client mode at the Remote but his may depend on the number of PCs at the remote site. AP Client usually only handles "A Client", not multiples unless a router is involved or you can use bridge mode on both ends but neither end will have wireless client capability unless you add an AP on that end. Unless you can get the signals out of the trees and house ( a clear Fresnel zone) then you may continue to have intermittent type problems.
How long are the bursts of high speed? How long are the times with no traffic? Just guess at approximate values. The times are very useful for determining a possible source of interference?
Have you tried using different channels?
Bad idea for initial testing. Set it to the transparent bridge mode. That will keep other wireless users and devices from associating and only allow traffic between endpoints.
I'm guessing that you're at the bitter edge of marginal performance. If I knew how much pigtail and coax cable you had between the WAP54G and antenna, I can make a better guess. Let's play with the numbers and please correct my guesswork.
I'll guess that you have the cantenna end connected with a rather long and thin (LMR-240) pigtail with about 6dB loss including connectors. The cantenna is good for about 8dBi gain, if built correctly. I have no clue how fast your burst of performance are running and therefore cannot guess at what association speed you're running. I'll assume
16Mbits/sec OFDM for a receiver sensitivity of -88dBm. The stock rubber ducky antennas have about 0dBi gain. In theory, they should be more, but that's what I've measured. Grinding the numbers.
TX power = +15dBm (Cantenna end) TX coax loss = 6dBm TX antenna gain = +8dBi Distance = 400 meters (0.25 miles) RX antenna gain = 0dBi RX coax loss = 0.5dBm RX sensitivity = -88dBm (6Mbits/sec OFDM) Desired fade margin = ????
get a fade margin of: 10.3 dB which is awful. The way fade margin works is that the rated sensitivity of -88dBm will yield about 10^5 BER (bit error rate) which is a rather flakey and useless connection. You have a fade margin of 10dB (about 10 times) more than that resulting in a substantially better BER. However, the errors will still be there and can be statistically estimated. Fade Margin Reliability 10 dB 90 20 99 30 99.9 40 99.99
However, this is under ideal conditions and does not include the attenuation of the house and tree as well as the fact that they're in the Fresnel zone and will cause some signal to be diverted via diffraction. Without a description, I can't offer a guess as to how much additional path loss these will present, but I'm fairly sure it will be more than 10dB. In other words, this link is not going to work reliably.
In order of importance.
Get line of sight. Relocate the antennas so that there's nothing in the way and that you have Fresnel zone clearance. If you can't get LOS and clearance, you will have flakey performance (guaranteed).
Get bigger antennas. You want to have at least 20dB fade margin with the above calculations. My guess is that you need at least
+15dBi gain antennas at both ends. If you have an aesthetics problem, you might want to try add on reflectors:
Turn off the repeater mode and use the bridge mode for testing.
Try a different channel in case you are having intereference problems.
Position antennas as close to WAP54G to reduce coax losses.
Fix the speed to 6MBits/sec OFDM. Do not let it rate adjust itself. AP's spend an awful long time adjusting their speeds for noisy connections.
Run the numbers. Aim for 20dB fade margin. It's not that difficult. If it doesn't work on paper, it's not going to work on the air. Actually, I'm rather amazed that it works at all with your existing configuration.
How much coax in meters? What type of coax? I like numbers, not vague descriptions.
Think about mounting the WAP54G in an outdoor box to reduce the coax length. See:
(see tiny "next" in lower right of page) for the general idea. One correction. The WAP54G and WRT54G will run on anything between about 4VDC and 15VDC. No need for PoE (power over Ethernet) converter. Just run two extra wires from the CAT5 cable to power and ground from the 12VDC wall wart. There will be cable loss but the WRT54G power supply regulator will not care.
That's about what you will get at the very slowest speed of 1Mbit/sec. The access point has decided that the path is so bad, that it has slowed down to it's slowest data rate in a heroic effort to maintain communcations.
Yep. That's usually what happens.
Actually, you're down to about 450kbits/sec nominal thruput with a huge number of lost packets and resends. I'm amazed that it's "sustainable" as it usually hangs if the packet loss is excessive. Try playing some 64kbit/sec streaming audio through the link and you can actually hear the stalls and resends.
Oh-oh. Trouble. Try to identify (with Netstumbler) the channel that they're using and find another. Cantennas are NOT very directional:
90 degress, you're approximately either -8dB or -15dB down depending on orientation.
I goofed. I thought you had two WAP54G radios and didn't see the laptop as one end. That cannot be used in the bridge mode. The WAP54G has to be in the access point mode. Try a different channel (1, 6 or 11) but I don't think that will help. You need a better RF path.
A lot to digest there Jeff and it'll be a few days before I will be able to test out and report back.
A couple of quick comments:
1) Have tried to keep coax cable lengths down so that there is a theoretical net gain
2) I have no latitude to change the horizontal path. It is the only clear line of sight I have. I could go higher but cable lengths will go up (I'm already on a shed roof the office end to get a line of sight).
3) I've hit around 400kbits/sec (with cantennas both ends) which would be acceptable if sustained. Try to download a 3mb file for testing (from the LAN not the net) and after a couple of attempts the systems seem to lock and do not recover. I then have to revert to the "rubber duckies" in the window. Typically then I'm down to 100kbit/sec but sustainable.
4) There is another cantenna arrangement about 30 metres to one side at the office end. The office end cantenna is pointing at 90 deg to it but I'm sure it is able to pick up some signal as I can with my laptop if standing under the aerial.
5) Bridge mode seems my next test with probable channel change.
Yes. You usually need two identical radios to play transparent bridge (there are exceptions). With a transparent bridge, you can connect up to 253 computahs at the remote end. Each will get it's own IP address from the DHCP server.
That will work.
It doesn't matter. The radio will select its favorite and antenna and stay with it until it loses the connection. I still suggest you get a "real" antenna or two. It's the lack of antenna gain that's causing the crappy connectivity.
This is not an easy question but I'll try to keep it simple.
WDS (wireless distribution service) is a feature where an infrastructure access point can also act simultaneously as a transparent bridge. Wireless clients can connect to a WAP54G at the office normally. Wireless clients can connect to a WAP54G at the house normally. The house WAP54G can also transparent bridge to the office. Therefore, wireless (and wired) clients at the house, can connect through the WDS bridge (repeater) maze to the office WAP54G, to a router, and finally to the internet. This may actually work for you.
However, all is not perfect. There are problems.
Your connection between the house and office is so bad, that you're going to have very poor thruput. Half of that will make it even worse. You should make an effort to improve the path, antenna gain, avoid the other users on ch 11, and thereby improve the thruput. Once you get it at about twice your DSL or cable modem speed, then half of that will not be detrimental.
Another problem are the antennas at the house WAP54G. You could just mount the WAP54G on the roof, run power, solar, battery, AC, whatever, on the roof, and run everything from the house via wireless. Hopefully, the path between the house WAP54G and the office WAP54G will be good enough to fix the performance problem. However, the directional antenna necessary to do that at the house may not have any coverage in the downward direction into the house. Installing a 2nd antenna pointed downward on the "other" antenna port seems an obvious fix, but there's a problem. It takes time for the radio to decide to switch between antennas and that introduces some rather large delays. You can do the repeater thing with one antenna, but don't try it with diversity enabled (i.e. 2 antennas).
I know of some tricks using a Wilkinson splitter/combiner that will sorta solve the antenna problem, but you need to have at least 4dB of surplus fade margin available. You don't have that luxury so I won't suggest it. Anyway, if you can't get a decent direct line of sight, then try the repeater.
Argh. It's Sveasoft Freya v2.06-1sv for the WAP54G that adds WDS to the WAP54G. The current stock v2.08 firmware for the WAP54Gv2 doesn't have WDS. I think the other firmware releases (OpenWRT, HyperWAP, MustDie, Neo, etc) also support WDS mode but I'm too lazy to check. The big problem with the WAP54G is that it only has 2MBytes of flash as compared to the WRT54G with 4MB and the GS with 8MB. Kinda hard to cram all that code in 2MBytes.
If he uses the stock firmware, the only thing that will work for bridging is to use two WAP54G boxes setup in transparent bridge mode with CAT5 cables to each end. Since the office end needs an access point for it's local clients, these WAP54G bridge boxes would be in addition to the existing WAP54G. This could get messy and expensive.
No, repeater mode is mutually exclusive with access point mode. I'm not sure if it's mutually exclusive with client mode in the Sveasoft Freya.
Yep. Same problem in this newsgroups. No manufacturer names, no model numbers, no hardware revisions, no firmware revisions. Sigh. I guess I complain to much and expect too much.
One more correction. My testing with using repeater mode between the two antennas was done with a DWL-900AP+. It's possible that the repeater mode in the WAP54G has fixed the performance problem as it's a completely different product. I guess I should try it (later).
I was planning to initially use an ethernet cable to connect one laptop to the WAP54G at home to test. I presume that will allow the WAP54G to be in bridge mode. If all works out ok I'll put in another access point in normal access point mode then connect that to the WAP54G so that both laptops can either talk to eachother or to the office lan/broadband.
The WAP54Gs have dual aerials currently set to diversity mode. With bridge mode I plan to set the aerial to only the one connected to the cantenna.
I've tried to find an answer to a relative straightforward question but there is too much noise - what is repeater mode for if it is not for what I am doing?
I can only choose ONE of the following four modes:
Access point (default) AP Client Wireless Repeater Wireless Bridge
The last 3 require a MAC address.
There is a note on the selection screen:
Note: When set to "AP Client" and "Wireless Bridge" mode, this device will only communicate with another Linksys Access Point (WAP54G). When set to "Wireless Repeater" mode, this device will only communicate with another Linksys Access Point (WAP54G) and Linksys Wireless-G Router (WRT54G).
Linksys, not third party, so the mode that Jeff referred to is unavailable.
Each of these requires the MAC address of the Access Point/Bridge that you want to connect to. That is to be expected. I think repeater and bridge also require the partner to have your MAC address entered in its table, but that isn't clearly documented.
I wonder about the "client" statement. That should not be true. A client should be a standardized operation.
A repeater or bridge might operate in Linksys proprietary fashion.
I've set the transmission rates to the lowest (1mbps)
Is there a simple chart showing the maximum reliable speed for a given theoretical operating margin (sorry if I've misused the terminology). I will be satisfied with solid broadband performance (512k) over the wireless link so 11mbps is in itself not essential.
I assume that WEP is not a significant factor in all of this.
Bottom line, you need either a clear path or LOTS more antenna gain.
"At most"? Does that mean that it also occurs more frequently than a "few" minutes? Let me try a different approach. Wireless traffic tends to be erratic. Click the mouse on a wireless web connection and you get a burst of traffic. Therefore, if you get dropouts at about
15-30 second intervals, followed by long pauses where they read the mail or web page, it's interference. Microwave oven interference follows the TV dinner pattern. It mostly happens around meal times and tends to be 3-6 minutes of continuous outage with long intervals in between where traffic is normal. Cafeterias are different in that they run almost continuously during meal times. Cordless phones tend to cause a total loss of signal for rather long periods (10-30 mins). Do any of these fit your pattern?
Same as LMR200. Including connectors with no pigtail, I would guess
2.5dB loss at one end, and 3.5dB at the other.
I beg to differ. See: |
gain assuming everything is perfect. I've only measured one can antenna using the above dimensions and measured 7.8dB gain on channel
6 and about 6.5dB gain on channel 1 and 11. That's using a conical feed element instead of just a 1/4 wave wire, which has better bandwidth (less VSWR). It is possible to build a 12dBi coffee can antenna, but it will be more like a horn antenna. My guess is an aperature diameter of no less than 250mm will yield about 12dB gain. So, what type of coffee can are you using?
Way too optimistic for a coffee can antenna gain. Revised numbers including corrected coax cables and the slowest possible speed: TX power = +15dBm (Cantenna end) TX coax loss = 2.5dBm TX antenna gain = +8dBi Distance = 400 meters (0.25 miles) RX antenna gain = 0dBi RX coax loss = 3.5dBm RX sensitivity = -93dBm (1Mbits/sec) Desired fade margin = ????
Now, I get 17.8dB which methinks is barely adequate. If you hang a can antenna on the other end, you'll pickup a huge 8dB increase and this thing will work. However, not that I'm using receive sensitivity at the slowest speed. If you crank up the speed somewhat (recommended), then you'll loose some sensitivity.
Then, add about 10-20dB of loss into your fade margin calculations. This isn't going to work unless you get line of sight or crank up the power dramatically.
Won't work. I goofed. I thought you had two WAP54G radios. Bridge mode only works between matching pairs of wireless (transparent) bridge radios.
Orthogonal Frequency Division Multiplexing. Think of it as a modulation method. The 802.11g box supports several others. I usually tag my connection speeds with the modulation method. Not an issue.
The settings look just fine.
That will get you the most receiver sensitivity but not necessarily the best thruput or most reliable connection. I suggest you try
6Mbits/sec OFDM, which is the slowest OFDM, because OFDM deals with reflections (off your intervening building) much better than CCK etc.
Actually yes. I can create a simple table if I knew the exact chipset demodulator characteristics and level of reliability (99.999%) you were expecting. Actually, I have one for older Orinoco Silver hanging on the wall in front of me. It's in the form of S/N ratio (Eb/No) versus BER (bit error rate). It's going to be tricky for marginal links because the BER curve jumps abruptly as soon as the demodulator starts working properly and chewing on data instead of noise.
That still doesn't account for it being tolerably usable with the Linksys WAP54G in the Window and intolerable (see next para) when I connect that Linksys to the cantenna. Even using your figures below I should be picking up 5.5dBi
Without the cantenna and with speed limited to 1mbps the throughput is fairly consistent. Adding the cantenna it freezes to a standstill. When I first tested the cantenna (before fastening it to the wall - just holding it towards the office cantenna) and settings at 11mpbs I got 4mpbs for short time then it dropped to nothing. Now whatever I force the speed setting to it runs for a few minutes then drops to nothing.
Reverting back to the rubber duck aerials and I'm running again.
I thought maybe the reason was that the higher gain cantenna might have been picking up the other distant wifi system. However changing the channel to 1 hasn't helped.
As above, just a few minutes then hang (cantenna).
No and none of these delays with the Linksys in the window.
There is just the one cable from each Linksys to the cantenna, so just one extra connection at the cantenna (the linksys already has/had one connection to the rubber duck aerial).
I looked at these earlier but it is not clear what I'm looking at. Is there a drawing of this cantenna design somewhere?
My colleague built it and I should have measured it before putting it up :) Approx 10" long and 4" diam to a design off the web. Added a conical horn to the home end so flares to about 8" diam in an attempt to improve its performance.
I think I'm going to have to wander around and take some measurements with Netstumbler and post them somewhere.
That's the problem. It doesn't but it tolerably does without the cantenna.
I infer that 20dB is minimum for good results.
I **do** have WAP54G at both ends. I'm bringing both Linksys WAP54G's into the office, set to Bridge and connect one end with an ethernet cable to my spare D-Link DWL-700AP and the other WAP54G to a hub on my network. Though I must say that doing the 400m and back dash to reset something that has hung is good for my health.
Not simple then :)
Don't mind trying this if I can rule out any fresnel effects killing the system. As mentioned earlier I have the only clear line of sight path available without going ridiculously high but not the clear path you would prefer I had. Much further expense and it will be cheaper to put in broadband at home and add a vpn (though I'd miss out on all the fun eh?)
Nope. There are too many assumptions. You haven't tested the antenna for gain. There's also the not so minor problem of the obstruction in the line of sight. I arbitrarily gave it a 10 to 20dB loss. Either extreme will create a useless path.
As for it working with an omni and not a coffee can antenna, that can also mean that your not getting the signal through the line of sight. Instead, it's bouncing off of something, and you're getting the reflection. If you had a really high gain antenna (i.e. 24dBi dish) with a 5-7 degree beamwidth, you could play direction finder and see where the signal is really coming from. I've been supervised more than I care to admit with such sneaky paths.
Either the can antenna has a problem, or your signal is not coming from the line of sight. Hard to tell from here.
That does happen, but changing channels should fix that (assuming it's
802.11b DSSS interference). Not this time.
Unclear. Never mind. It's probably not interference.
Sure. It's a model of Trevor Marshall's can antenna design. Note the gain plot on the above URL. Some notes on the original design at the bottom of:
The flared horn antennas will have more gain than 8dBi. There was a web site with construction details, but I can't seem to find it.
to 14dBi are possible.
Watch out for reflections off the ground and from nearby buildings.
Getting a signal and keeping it online are two different animals. You can acquire lock at 0dB fade margin. You'll get a connection but not much else. Everything you do will reduce the fade margin. Unexpected losses, line of sight obstructions, and interference all tend to ruin a nice neat fad margin calculation.
No. 20dB is a good target for a RELIABLE link that stays up fairly well. 20dB is 99.0% reliability. That means your link will be down for 3.65 days per year. You can run a link with much less fade margin, but the reliability gets rapidly worse. I use 20dB as a target value and settle for slightly less. You can sorta test the link fade margin by adding fixed attenuators in line with the radios
Looks like a worthy test. Methinks results will be similar to the current situation.
Well, no. It's not simple. I'm not sure it's even useful. However, it can be done. Broadcom is not too liberal with passing out their data sheets. That last time I needed one, I had to go through an ordeal process and was asked to sign an NDA (non-disclosure agreement) in order to get spec sheets and reference design info. I have some Broadcom data, but I think it's from before the WRT54G and WAP54G vintage chips.
You can't. Even with high power and big antennas, objects in the path will create knife edged diffraction and block the signal.
You're decision. VPN over ADSL will run at the rate of the outgoing speed. That's anywhere between 128kbit/sec and 768Mbits/sec. That's better than a modem, but not as fast as a decent wireless link can run.
ooo, I like that one. I used an Orinoco and a coffee can to locate some unauthorized WAPs in the building. At 10 feet, it's a pretty good pointer. Down the hallway, around the corner, maybe not so good.
(Trying the John Navas method of checking signal strength and walking around wasn't worth diddly.)
Also, a reminder about your earlier comments about the cans not being bore sight aligned. I haven't noticed with a normal cantenna, but with a USB cantenna, and lacking a line of sight shot makes for a best signal that doesn't point the barrel of the can toward the wap.
My first thought was that his cantenna was less than good in its construction. A test with some nearby WAP might be in order. For signal testing, the target WAP could be moved to someplace else. It shouldn't require a WAN connection to get a good wireless link for test purposes.