Co-location.


Good day all, Ive finished the installation of (Wipop) ... Pretty simple setup ....


(3) 17dbi sectors all connected to 3 rb532a r5 boards ultimately powered by 3 xr2's running at 26dbm, also (1) 5.8Ghz 17dbi powered by an xr5 and finally 1 sr9 Connected to a 13Dbi Sector


Simplification


Routerboard 1 : Contains two xr2's

Routerboard 2 : Contains both a xr2 and xr5

Routerboard 3 : Contains the sr9

-------------------------------------------------------------------------------------

The noise floor reading from all cards seem flawless: Average -94 for 2.4Ghz and -100+ for 5.8 and surprisingly -98 for 912Mhz, although drops to -87 for 917.


As the area of concern is quite dense/small the say the least the smallest access point can be far reaching ... When a scan is conducted from any of the sectors, I on average note about 60+ Ap's with signals ranging from -60dmb down to -84. Needless to say the spectrum seems a bit clogged for the 2.4Ghz spectrum. But the point in interest here is can having the sectors spaced approx 1ft apart and the radios co located within the same box all blasting at full power have something to do with the horribly low throughput ive been receiving from bandwidth tests. on average 0.1Mbs.

However and funny enough when nstreme is enabled each independent sector appears to work "fine" 13Mb/s on a 22Mhz Channel.


Id really appreciate it if any one has ever had the same experience and if so what was done to rectify it.

Additional notes:

Each rb is powered by a 48V 60 Watt poe supply to "ensure" that they are adequately powered.

Each ap reads a signal strength of -24dbm from the other. Way to high?

Im assuming switching to hori pol. will help a great deal but my question is will the problem of co-interference between antennas still exist as much as it does vertically.

Finally im planning on placing each radio in its own enclosure with a new antenna spacing of 3Ft with the help of a side arm bracket.




[/img]http://img410.imageshack.us/my.php?imag ... z3.jpg[img]

[/img]http://img407.imageshack.us/img407/292/ ... th.jpg[img]


[/img]http://img407.imageshack.us/img407/9365 ... co4.th.jpg
[img]


[/img]http://img407.imageshack.us/my.php?image=t1800np3.jpg
[img]



Any help would be MUCH appreciated.

Best Regards.[/img]

Argh, .... correct photo postings.

Your 3 RB532a's should all be contained in individual enclosures.

You antennas should be, in my oppinion, at least 32 inches apart.
Also, keep your LMR runs, as far apart as possible.

You should be able to acheive -85 dB or lower seperation between the APs without much else.

I know I stateing the obvious, but you have got your 3 x 2.4GHz radios on different frequencies?

If they're all on the same or very close frequencies then I'm not surprised you see high power levels, after all that's what they're supposed to do!!

You need to use non-overlapping channels, e.g. 1, 6 & 11. (2412, 2437 & 2462 MHz)

I know I stateing the obvious, but you have got your 3 x 2.4GHz radios on different frequencies?

If they're all on the same or very close frequencies then I'm not surprised you see high power levels, after all that's what they're supposed to do!!

You need to use non-overlapping channels, e.g. 1, 6 & 11. (2412, 2437 & 2462 MHz)

.... They are in fact spaced on the 3 "non-overlapping" channels ... 1,6,11.

This is really beating my head, any help is appreciated.

Regards

Switching all your antennae to horizontal polarisation is unlikely to make any difference to co-location interference. It may reduce interference from other APs though.

Are you sure you aren't misleading yourself? If you are using the Scan option in the MT wireless interface, then you will see the other sectors, because you're scanning all frequencies. This doesn't mean you have -24dB of interference on the working frequencies. When the interface is in working mode, i.e. not scanning, then it's centre frequency is locked back to the configured value.

I'm not sure how the Scan mode works, either it's rapidly switching the interface's centre frequency and listening on each or the band pass filters are configurable and are just opened wide enough to detect the entire 2.4GHz band. My guess is the frequency is being cycled.

Thanks much cmacneil, What your saying is in fact true... The ap is conducting a dynamic scan when in scan mode, hence a avrg. noise reading of -94 when they are operational ... Maybe I should have been a bit more lucid in my initial post, as I was partly referring to the CSMA backoff feature of 802.11 as explained here ....


http://forum.mikrotik.com/viewtopic.php?t=13668&

However Im expecting the h-pol sectors along with seperate enclosures and will post results. The v-pol will be used for a less critical / "noise free" site.

While we are on the topic ... I was wondering how the h-pol sectors and h-pol omnis fair out with laptops? Is vertical better with laptops or the opposite ... ?

Regards

I suggest your co-location issues are due "adjacent channel interference". Even though the 3 wifi channels are not overlapping, there is still some signal leaking over into the adjacent channels when the neighbouring APs are transmitting. Because they are so close, it does not take much leakage to raise the noise floor of a neighbouring antenna when receiving.

For example:

Let's say the power output from one antenna is a legal maximum of 36dBm.

If the isolation between antennas is good, let's say 40dB (could be as low as 20dB).

If we were using the same channel on two antennas, this would equate to a noise floor on one antenna of 36-40 = -4dBm whilst the other was transmitting and obviously there would be no chance in hell of the other receiving.

Now look at the situation for co-located channels.

Let's say we are aiming for a noise floor of -100dBm.

This would require that the side-bands (signal leakage) from co-located channels to be a further 96dBm (minimum) down on the carrier. This would not be acheived in any wifi card without elaborate switchable output filters. Wifi cards are simply not designed to operate in this way with access points "on top" of each other.

Suggestion is a high-order filter to sharply filter the edges of each of the channels 1, 6 and 11.

I know that hyperlinktech (for example) have 4 and 8-pole filters for wifi channels 1, 6 and 11. See the following link on 8-pole filters: http://www.hyperlinktech.com/web/bpf24-8xx.php

These have 60db out-of-band rejection and 3dB instertion loss.

I personally have never used them because I have never tried to operate co-located channels, but if I was doing what you were trying to do I would have been using these from day 1.

1. What type of CPE hardware are you using to test?

2. The issue is very obviously the interference you are picking up from other AP's NOT your own.

3. DECREASE your power. The output of the XR2 is WAY too high for good throughput.

3. Switching to HPOL sectors will dramatically improve your performance.

4. Adding filters into your signal path will hurt far more than help.

5. Laptop users will not have good signal to HPOL antennas.

6. If you can drop to 10mhz spacing (depending on CPE hardware) this will also help with your throughput issues and increase your range.

2. The issue is very obviously the interference you are picking up from other AP's NOT your own.

3. DECREASE your power. The output of the XR2 is WAY too high for good throughput.

Have you tried switching off 2 of the 3 APs and seeing if your throughput increases on the remaining one? This would be the obviousfirst step to determining if the interference is due to co-location or external interference or a likely problem with the XR2 card itself.

Another thing ot consider - I have recently purchased the XR2 card for my RB532A and I have not been able to get it to work properly when switching to 802.11b/g mode.
But I've only had it one day and am yet to delve in further.
Pings are 1000ms + if at all. I can get the pings down to about 100ms if the power output is reduced by 10dB or so.

TCP/IP Bandwidth tests (using IPERF) vary from about nothing to 15Mbps if I play with the power output - but even if I get 15Mbps one second the next second it can drop to next to nothing. Very strange indeed.

It does seeem to work well on 802.11b mode only, however - 1ms pings if nearby to antenna and up to 6Mbps TCP/IP throughput at 11Mbps connection speed.

If it is infact a problem with the XR2 card then I suspect a power supply or a driver issue. I have noticed that the XR2 card requires up to 1.4A during transmit. I am using an 18V 1A (18W) power supply, Depending on the switched mode power supply efficiency, the RB532A may require up to a 25W power supply to deliver full current capability of 4.25A @ 3.3V. In your case you should have no problems with the 60W 48V power supply on each RB.

Tonight I am going to measure the XR2 card voltages at different bandwidth and power settings. It should remain around 3.3V, if not that might be my problem

Don't use the RB532's PoE on 48 volts, causes bad noise. Shift to 12 volts.
We are sharing many towers with two-way radio systems. The only way we were able not to disturb the other systems is by shifting to 12 volts. Though you may rather use thicker shielded power cable and rather use the jack than the POE.

The 48 volt POE setting causes noise on the ethernet cable, which can transfer to the antennas fit on the same tower.

It is just something to do with the 48 volts extra cricuitry on the RB532s.

I no longer think the power supply is an issue.

I first tried two other 802.11b/g wifi cards including the CM9 to see if they performed as bad as the Ubiquiti XR2 in my setup. To my surprise, all cards performed horridly, if at all, when in 802.11g mode. All worked OK in 802.11b. I am using a 24dBi grid and had my laptop as a client sitting on the ground in front of it, less than 3m away. Yet it was not working (only the occaisonal ping got through) when in 802.11g mode.

In fact, I the 802.11b was still working so well that I can sit 1.5km away clear line of sight, in the signal path of the 24dBi grid with my laptop in my car using the standard internal antenna and get a 11Mbps connection! TCP/IP speed was about 3Mbps.

With the XR2 on 802.11g I then swivelled the grid antenna about 15 degrees and all of a sudden the link spung back into life! The laptop 3m away started pinging sub 3ms and bandwidth tests showed about 18Mbps.

Yet with the antenna 15 degrees to the right I can't get anything?

What's going on? Smells to me like an interference issue.

Before I chose the frequency I was using (channel 1 - 2412 MHz), I scanned the area for APs and it was the clearest channel.

So is it a case of strong co-channel interference or something else?

When I scanned for other APs on other frequencies none showed up as alarmingly strong. The strongest was at about -81dBm on channel 7. So I tried changing the channels anyway but with no real improvement in performance.

And anyway, even a strong nearby wifi signal should not cause the problem of next to no connectivity - there would still be enough "gaps" for some data to get through, transmissions would not collide all the time.

So whatever the interference is, it is constant, and seems to affect 802.11g far more than 802.11b.

So I gathered up the Spectrum Analyzer from work, trying to shed some light on the issue.

Conencting the Spectrum Analyzer directly to 24dBi grid, and pointing in the desired direction where the interference was strong, I left the Analyzer scanning the entire wifi band constantly for about half an hour on MAX HOLD. After this time I did not see anything alarming, the busiest part of the spectrum was around 2.45 GHz, most likely a few APs defaulting to channel 6 and other 2.45GHz devices and maybe a few microwave ovens...

However nothing that strong, or constant. Some sweeps could pass without putting a dent in the noise floor of the spectrum (about -95dBm).

Just for the hell of it I pointed the 24dBi grid in the direction where 802.11g worked. Then I scanned the band again for about 30 miniutes on MAX HOLD. There was very little difference.

Now I was suspicious that something very strong, out-of-band and constant was causing an issue.

So I set the spectrum to scan from 800MHz to 3000MHz.

With the antenna pointing in the direction where 802.11g stopped working, there was a strong GSM signal around 900 MHz, and an even stronger CDMA signal at 1810 MHz. In fact it was reading about -60dBm on the spectrum at 1MHz bandwidth.

Was this the problem? The literature says that the XR2 has far superior rejection at 1.8GHz than other wifi cards.

So I moved the antenna 15 degrees to the left, where the 802.11g had worked, and bingo! The signal levels at 900MHz and 1.8GHz dropped about 20dB.

I think I've found the problem...

How to fix it?

Earlier I suggested some Channel-specific bandpass filters to eliminate co-channel interference. I beleive that one of these should fix my problem.

Alternatively, a full 802.11b/g 100MHZ wide bandpass filter should adequately reject these two signals.

I am going to order one, let's see how we go...

We have a very similar setup, however we used a sectorized omni from Pac. The only difference is that we did buy bandpass filters that are placed within the box.

I have a box with 4 stacked rb532's and 1 112. 3 of them are for the sectorized omni on 2.4ghz sr2's, and are not spaced more than .5inches apart (far enough to get the connectors in without touching the board above it).

We do not have any issues, however we believe it is because of the filters, reason being we have other locations where an omni and backhaul are both 2.4Ghz and when bandwith tests are done from client pc to our core they are not as good as backhaul to core, because of the generated interference that the omni causes to the backhaul (totally different freq). The other thing is we do 5mhz channels.

We do get near max performance for having 5mhz channel sizes. I would recommend trying to turn down your power as well.