noise readings

[bobbyc]

Tony or Lonnie,
This isn't so much star-OS related, but I know you guys have lots of experience with orinoco gear (?), so here we go.
We were doing some signal checks on a amped sector today, because we are trying to align the beamwidth. The customers closest to the tower get a SNR of about 30station(them) and 20 partner(tower). The customers farther away get a SNR of about 20station and 10 partner.
We then went up to a hilltop that is about level with the tower and 3 miles away. We get a SNR of 30/30 station/partner. The ~10dB difference at the other locations was because of a higher noise floor at the tower side of things, as being reported by their orinoco client managers. But why is it that there was no noise level being reported, at this hilltop we were at?I watched it for about 1 minute and it held steady.
It doesn't make sense to me. Why should location of Customer play a role in the noise level of the tower? A noise floor is a noise floor, it would be a consistent figure? I can understand the signal flucuating depending on location of CPE, but noise at the tower? Was this a misreading?
Thanks for any comments you might have.
Bob C

[dkii]

I have experienced the same situation, when associated with a star-os box about 1.5 miles away on channel 9, I got a noise floor of about -80. I then associated with a different access point(dunno where, not mine) on channel 9 and the noise floor jump to around -95 to -99(as reported by orinoco client manager) I don't understand how the equipment on the other side of the link could have anything to do with the my noise floor locally. Both of the access points were on the same channel. The star-os has a lucent card, and the other access point that I associated with had a linksys mac address. Go figure?? There is, however quite a bit of 802.11b equipment in my area, on a 4 mile drive from my office to my house I picked up(with netstumbler) over 160 different access points, raising my total to over 200 access points within a 5 mile radius. I couldn't believe it, but I have the netstumbler log to prove it!. So that might have something to do with my noise figure, but why it is changing like it does eludes me.

[bobbyc]

Are we talking about the same noise reading? I'm talking about partner noise reading flucuating depending on location. Station noise readings are always -97, 98, 99....
My question is why is it when I'm at customer location A, my amped AP hears a noise floor of -86, but when I'm at customer location B, it hears a noise floor of -96?
I am definetely going back tomorrow to make sure it wasn't a fluke.
Bob C

[dkii]

we are talking about similar noise readings, you are referring to your AP having 2 very different noise readings, with the CPE from 2 different locations, when the AP has not moved, and thus should have no change in noise reading. I am referring to myself as a client to an AP having 2 very different noise readings when I am associated with 2 different AP's, when both are on the same channel, and I have not moved, therfore my noise reading should be the same, regardless of what AP I'm associated with, as long as it's on the same channel. I have pretty much concluded that the noise/signal readings are terribly inaccurate on the cards, and the only way to get a good reading is with a spectrum analyzer, or similar device, but I'm interested to get lonnie or tony's input on this. Later, daniel.

[georgew]

I have observed this many times myself. I think I have the answer...

It has to do with how energy (photons) travel. On a macro level, the added noise is probably coming from multipath reflections of the signal generating noise with itself. In other words the noise IS the signal coming in out of phase.

On a micro level there is also noise associated with interference within the fresnel zone.

For example pick a near object like the edge of a building, and a distant object you see near that edge. As you move your observatation point to where the near object starts to cover the distant object, the edge of the near object has a "aura" around it which is the distortion of the light energy caused by oclusion of the fresnel zone the photons travel. You will see the distant object distort as the near object gets close to it. This is noise.

Radio waves are photons too, but at lower frequencies. Lower frequencies mean a larger fresnel zone, and at the edge of near objects the "distortion aura" is wider. So in radio terms the noise cause by objects in the path exists in a wide "aura" around the objects in the path. In light, the distortion aura is nearly invisible, at an arms length the aura around my thumb looks like it is a half milimeter thick. I expect this is also about 50% of the diameter of the fresnel zone of the light coming in past my thumb. At microwave frequencies that "aura" is huge, probably 50% of the fresnel zone diameter... if not 100%. It is my theory that this is what the fresnel zone IS, the zone of distortion that is caused as photons pass objects near their path.

So I suggest that it is a result of two things, one being out of phase signal (noise) coming from multipath reflections, and the noise generated by an incomplete fresnel zone.

The expremental proof is in the distant hilltop showing less noise than nearby valley-floor radios. It would take much more exprementation to determin which theory is responsible for most of the noise, multipath noise or fresnel occlusion noise. But I wager both are involved, as multipath is a well-known problem, and fresnel zones are proven influences as well. I can /see/ the fresnel noise with a light path, and the distortion is clearly there...

So the noise you are seeing is actually signal that is distorted to the point that it subtracts from the S/N level. So it is not the noise floor at all, it is a measurement of signal distortion of the transmission itself.

In case you are wondering, a photon is not a microscopic point. When traveling from one place to another, photon size is a function of wave length and distance from the originating point. the wave function collapses to a point when the photon is captured in an observation (or radio reception). An electron is nothing more than a cloud of photons, and therefor can be quite large, as large as is needed for it's current mode of travel... Really I don't think electrons exist at all, it is just a convenient way of grouping photons in a way that is easier to understand their behavior. But I digress...



George

[bobbyc]

your example makes perfect sense.
Would having a amped sector with 6 degree vertical beamwidth add to this phenomenah when we should really be using a sector with 10 degree beamwidth?
Thanks, Bob C

[georgew]

The wider the antenna beam, the more noise you will pick up, in general.

It really depends on the specific vector of the noise of course. In the case of multipath noise, if you can cancel part of the noise with a narrow beam, then the noise figure will be reduced. If the multipath noise is on nearly the same vector, such as a reflection from the back side of an omni, then it will be harder to eliminate. Noise from fresnel occlusion is probably impossible to eliminate due to it being in-beam.

The more I think about it, the more I think fresnel occlusion noise is an optical quality rather than something you would see with a radio. There would be some phase noise, which is bad, but for the most part it would be minor compared to multipath noise.

Now some radio makers are building multi-antenna arrays, and analyzing each incoming signal, and using a DSP, doing a phase realignment to reconstitute the original signal. The theory is multipath signals could be added together for extra gain if you could realign the phase angle of the individual signals.... in effect each reflecting surface adds to your signal instead of taking away from it. In transmitting a signal, the unit would use the delay factors on the incoming signal to produce properly timed outbound signals aimed at the same reflecting surfaces the incoming signal used.

This all sounds fine in theory, and one hardware engineer I have talked to made claims like being able to talk to 802.11b cards built-in antenna from a mile away using an 802.11b compatable version of the technology. Frankly I question the likelyhood of this extreme claim, but if true it would end the need for towers and reduce most customer installs to plug-and play cards with no antennas... at least for close customers. In my experience you have to take these sort of claims with a grain of salt, but it IS likely that the technology will work for small cells.

My gut feeling is this technology will work if there is only one signal source, and as soon as you have a lot of noise from other sources, you would loose your ability to pick-up multipath reflections. So the system gain is probably going to be limited by the local noise floor.

I plan to sign an NDA with a local company developing this sort of hardware, so I probably won't be able to talk about it much in the future. So what I will do is before signing the NDA, I'll tell you everything I know, including the company name, then after I sign it, all I will be able to tell you is if I am happy with it or not. I'm not sure about the cost of this sort of technology, the last similar product I reviewed has a cost in the $5000-10000 range for the base and $1500 for the CPE (it was not 802.11b compatable) so it was not commercially viable in my opinion.


George

[lonnie]

Well, I have no better explanation than anything else presented here.

Something you might try is to substiture a radio at a noisy location, or any location for that matter, and see if anthing changes. You might be seeing the difference in radio signal quality.

We don't have a Spectrum Analyzer so it is real hard to verify any ideas we might come up with.

[dkii]

very good explanation, but I still don't see how one association can have more noise than another when the frequencies/antenna direction aree the same, regardless of what I'm locked onto with the client manager, if I am in a certain location, with a certain antenna, pointed in a certain direction, the local noise level should not change base on which AP in that direction I am associated with. I DO have a spectrum analyzer back in florida, so I will do some experimenting when I get back. I did find the source of my noise, however. I was using an old cable that had been laying out in the weather. I stuck an ohmmeter to it and it measures about 30 ohms, so I figure the noise was from reflected transmitted signal coming back down the coax.

I saw something that I think would be an incredibly good idea, if the hardware were decent. Linksys's WAP11 has 2 antennas as I'm sure you know. In the software, you can set one antenna to transmit, and the other to recieve. That could eliminate a lot of cable problems, also for amped situations, I think it would be better because. On the transmit, you could place an amplifier right after the radio, where the signal is the strongest and the noise is the weakest. On the recieve, you could place an amp right next to the antenna, where the recieved signal is the strongest and noise is the weakest. On a single cable, you could only do one, or the other, making a sacrifice. Also, I'm not sure on this, and I haven't seen this specified very often, but I believe different antennas have different transmit and recieve gains, ex(totally fictitious) 19db patch may have 19db of xmit gain, but only 16db of recieve gain, but then a 19db grid may have 17db transmit and 19db recieve. I'm just guessing on this, so I might be completely wrong, but it sounds good to me :)

[lonnie]

The noise reported is total non-signal level, so if you have a noisy radio at the client, due to overdriving an amplifier or bad cable, or whatever, the reported noise will be higher on that particular association.

This is also possibly showing you the overall 2.4 GHz traffic, and since that varies at any particular point in time you will see things go up and down.

Our own systems show a 10 dB variance between customers. I would like it more uniform, but it is working as we need. Some day we will have a Spectrum Analyzer and maybe we can get to the bottom of it.