Reflection problem?

We setup a 2.5 mile link using 2 Tranzeo TR-6000 AP's with integrated 15 dbi in a rural location.

AP 1 18 ft agl AP 2 24 ft agl

There is one tree partially blocking line of site 75 feet from AP 1. A couple of buildings at about the half way point that definately come in to play. Otherwise it is wide open.

So we know this setup has fresnel issues but thought we would give it a try anyway. When the radios are dialed up to full power 23db the throughput of the radios is virtually useless. If we dial the power output on AP 1 down to 18db the link works great. I'm assuming that we are getting a reflection that is causing a high error rate and reducing the power output reduces the strength of the reflection.

Is this typical of a reflection problem? Could the tree be the source of the reflection or does a tree pretty much only absorb? Can a reflection off the ground cause this type of behaviour?

This is not a critical must be up connection, so I'll leave it for now since it has been working good for a month.What are the odds, if nothing else changes, of this link continuing to function reasonably well?

Any input is much appreciated.

Ian

Reply to
Ian
Loading thread data ...

"Ian" hath wroth:

That would be a TR-6015F.

formatting link

What's an "agl"? Oh, Above Ground Level. That's not the way it's usually specified. Try HAAT (Height Above Average Terrain) instead.

All trees are obviously identical. Is this tree directly in the line of sight? Any particular type of tree? If the tree has lost all its leafs for winter, you're gonna have a BIG problem in spring.

A couple means two. Are there two (or more) building in the way? Same questions as with the tree. What type of building? How big an obstruction?

It's always worth a try. The problem with Fresnel Zone incursion is that the effects are truely weird and inconsistent. You could move a few inches one way or the other with one end of your link and go from fabulous performance to zilch. Same with moving any reflectors or obstructions. You might get a good connection today, but tommorrow might be different.

Wrong. The power in the incident (direct) path varies directly with the reflected path. There are no transition effects. If you have a reflection issue, it should not be affect by power level. The easiest way to test for a reflection is to just move one end of the link. If the signal level and particually the S/N ratio varies sharply and radically, you have a multipath or reflection problem. Find an antenna position where the effect is minimal and consistant.

Hell no. However, what might be happening is that your connection speed is varying from 11Mbits/sec down to 5.5 or 2 or 1 when you reduce the power. That shuffles the deck and might explain the magical improvement with reduced power. It might also be that the TR-6015 is defective and generating crap in the internal RF power amplifier at 23dBm but not at lower power levels.

Suggestion.... Try reducing the power at the other end and see if the effect is identical. It might point to defective radio.

My house is in the middle of a redwood forest. I can assure you that trees only absorb and do not reflect. If they did reflect, I would be in big trouble with our neighborhood WLAN.

Ground reflections are a bigger problem than what you apparently indicate. At 2.5 miles and 2.4Ghz, the Fresnel Zone is a 37 ft radius. Your 18ft and 24ft antenna heights are well inside the Fresnel Zone at midpoint. Since you're measuring from the ground, I suspect there are quite a few obstructions at midpoint within a 36ft radius. |

formatting link
mind reflections. Think obstructions instead.

That's easy. Do the numbers: |

formatting link
no Fresnel Zone incursion (perfection).

TX power = +18dBm TX coax loss = 1db (internal connectors) TX ant gain = +15dB Distance = 2.5 miles RX ant gain = +15dB RX coax loss = 1dB (internal connectors) RX sens = -85dBm (at 11Mbits/sec CCK) Fade margin = unknown

formatting link
get a fade margin of 18.8dB. That's tolerable. 20dB would be nice. No clue how much loss the tree, buildings, or ground are contributing. Are you running at 11Mbits/sec or slower? If

5.5Mbits/sec, then your fade margin is 21.8dB which is good enough. However, that again assumes perfection in the path.

I can't predict the reliability without calculating the effects of the tree, buildings, and ground blockages. My wild guess (and I do mean really wild) is these will contribute at least 8dB of attenuation. That leaves a fade margin at 5.5Mbits/sec of 13.8dB which will probably function, but not very reliably. You can convert fade margin to reliability using the table in the FAQ. 13.8dB is about 95% reliability. That means you'll have 438 hours of outage every year.

Reply to
Jeff Liebermann

Thank's

If you don't mind I would like to tap into what you have learned from experience. When you say it's always worth a try, do you tend to find that if you set it up and it works, even though on paper it could be problematic, (I'm thinking in terms of fresnel zone incursions) that it tends to continue to work or that you end up with big headaches and would be better off staying away if the numbers don't work out.

Interesting, I talked to Tranzeo support and they suggested trying to reduce power because in their words "The radios are designed to work over a long range they may just be hitting each other too hard". It didn't make much sense to me ( I always thought more was better) but I tried it and it worked so I didn't argue. In fact lowering the power seems to have stabilized things at 11Mbits, at higher power it seemed to be hunting for a speed. At these lower power settings it seems to be working flawlessly.

Thanks again for a swift and informative response.

Ian

Reply to
Ian

"Ian" hath wroth:

I wish I had a rule of thumb for such situations. What I usually find that testing for fade margin will determine the ultimate reliability of the link. I usually carry a pile of attenuators to insert in the coax line between the radio and the antenna. If I can't insert 10dB of additional attenuation and still have a functional link, then there's gonna be trouble in the future. In systems where I can't insert and attenuator, I just cover up part of the antenna. Half the dish is -3dB. Covering 3/4 of the dish is -6dB. Make sure you use some kind of RF absorbant material (i.e. black foam or wet towel) and not something reflective.

Duh, I forgot to mention that you should temporarily fix the wireless speed when doing this test. Covering up half the antenna will probably cause the link to slow down. That hides the effect of signal loss.

There are lots of other issues that commonly screw up links. The biggest is interference. One day it works, the next day it doesn't. It's very difficult to check out a link in the presence of interference. My usually method is to rely on diagnosis by substitution. If I have SNMP running, I'll have a history of what is considered "normal". I can usually deduce the cause by the pattern and timing.

Bull shit. Back to the calculations. This time, the exercise is to figure out how far the radios need to be apart before they overload. Most 802.11b hardware just doesn't have much dynamic range. Typically, they will overload at about -20dBm input to the receiver. Going back to: TX power = +23dBm TX coax loss = 1db (internal connectors) TX ant gain = +15dB Distance = Unknown RX ant gain = +15dB RX coax loss = 1dB (internal connectors) RX level = -20dBm (overload point)

formatting link
to overload is 0.022 miles or 116ft. That's quite a bit less than 2.5 miles. I'm not sure of the -20dBm overload point. However, it's close enough as it will take a HUGE change in overload point to create overload at 2.5 miles.

I think you just proved my guess. It's not suppose to do that. There's something wrong at high power with one of your radios. My guess is that there's some envelope distortion at high power.

Asking again: What wireless speed are you running?

Reply to
Jeff Liebermann

A Spectrum Analyzer would be so much easier.

JM

Reply to
JM

I'll try a little more testing (adjusting power at the other end, perhaps bench test making sure to avoid the overload point) and see if I can get an RMA

Radios are set at best automatic but would like them to run at 11Mb/s. In reality they are used primarily for Internet sharing and remote desktop operation of one computer with very little file transfer, so I could live with lower speed.

Ian

Reply to
Ian

"JM" hath wroth:

OK, I'll bite. I really detest one line answers. What would you expect to see on the spectrum analyzer? Easier that what? Easier than doing the calculations?

Hint: The signal level at the receive end will be about -60dBm. That's the decorrelated CW equivalent for a single carrier. Spread that out over 25Mhz of spread spectrum, with an IF bandwidth of

250KHz, and you will be looking for: -10 log (25Mhz / 250KHz) = -20dB an observed signal level that would be about -80dBm. Do you have a spectrum analyzer handy that can see an SS signal that's -80dBm (assuming a similar +15dBi antenna and 2.5 miles away)?

Assuming you do see something, what would you look for? Envelope distortion can't easily be seen on a spectrum analyzer. Multipath and Fresnel zone effects cannot be seen at all. You might see envelope distortion with a sampling scope by looking at the raw RF envelope, but that's rough using spread spectrum. At best, you might see some additional bandwidth consumption on the spectrum analyzer.

Here are some awful photos of the spectra coming from an WRT54G with the power cranked up. That's about as much useful info as I would expect from a spectrum analyzer trying to troubleshoot this problem assuming the transmitter power amp has a problem.

formatting link

Reply to
Jeff Liebermann

Minor presumption - you have the equipment needed to make the try, or can borrow it for as close to free as possible. If you have to buy the equipment before you can test, it may not be worth it.

In the 1970s, we were running a test program in the Central Valley, and needed a datalink from there back to Ames. The local telco would rent a wide band wire link at some outrageous monthly fee, or we could try a wireless link. A direct shot is obviously out (both sites under 200 feet above sea level with a 2500 foot range of hills between), but we had access to a spot on a tower on Mount Diablo. The Diablo to Ames shot was a piece of cake, but the test site to Diablo missed line of sight by 130 feet at mid-path. What the heck, let's give it a try. With 18 dBi dishes at both ends, it's working, but the fade margin was only about 10 dB. OK, Andrew has these nifty 33 dBi dishes for only an arm and a leg (about what the telco wanted every six months) - buy two. Now, after we install these dishes, and the link is running flawlessly, we go back and document everything. As part of the documentation, we fly a helicopter near Diablo, and using the radar at the test site, we're able to see that we really would have needed a 600 foot tower on either end to get line of sight, but it's working so who cares. About a year later, we discovered that there is a LARGE wire fence on the top of the ridge near mid-path that seems to be acting as a knife-edge.

Things are great until about 1988, then we notice dropouts in the link - and testing on Diablo shows that the received signal strength is down 30 dB from normal. Antennas, cables, transmitter and receiver are all OK... WTF? Turns out the farmer who owned the fence at mid-path had torn it down. Adding a second repeater on Mount Oso with

10 dBi Yagi antennas (salvaged from another project) between the test site and Diablo solved the problem, except that we had pay another agency for the tower space on Oso, and con the Frequency Manager to give us the second channel he had insisted we'd need back in 1970.

Old guy

Reply to
Moe Trin

I usually have enough stuff around to do a test.

That is a good lesson. If you don't have full control of all obstructions in the way, even what you think may be a permanent non moveable structure can change. When it comes to wireless the unexpected can always come back to bite you.

Ian

Reply to
Ian

Not everyone does. Still, I hate to think how many extremely fragile, but functional antennas (usually, shallow horns built out of sheet aluminum) we cobbled up to run a test.

Some of it is local knowledge. The fence in question was about 1000 feet or 300 meters long, and 6-10 feet / 2-3 meters tall and seemingly made of "chicken wire" (about 1 inch hexagons) - certainly it had very little effect on our X band (9 GHz) radars. I still have no idea why it was there - following an East/West section line about 3 miles from the nearest road. There was also a (common here in the West) 3 strand barbed wire fence normally used to divide cattle grazing areas that was still in place in 1995.

True. In this case, it worked for us for about 16 years, so we had no reason to complain. We really didn't _know_ that the fence was the the cause until it was pulled down, and even that is a 'best guess' based on chance aerial observations (we flew over the area, often at fairly low altitudes several times a week).

Old guy

Reply to
Moe Trin

Cabling-Design.com Forums website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.