Question for users of 2.4GHz infrastructure with 4-sector base stations.
Any opinions from the experienced users out here, which is preferred - a single high powered MT box with 4 radio cards, or 4 separate radios with more modest CPUs?
Obvious points spring to ming,
resilience/single point of failure
mutual interference between radio cards in one chassis
cost
My question is about the architecture: are there big real-world issues in provisioning services over 4 separate radios/routers, compared to 4 cards in one router?
Which do most WISPs prefer?
Well, I have some experience with using 4 channels in close proximity (1 mile apart towers with LOS, 1 tower with 3 sectors, 1 tower with 2 sectors, 4 of them facing each other). I think one thing you can do is use a 4 channel scheme (1,4,8,11), which lowers the chance of self interference. Or, you could use 2 channels in vertical polarity, and 2 in horizontal. The 4 channel scheme works for the most part, but suffers against 1/6/11 inteference, since channel 6 ends up messing with the 2 center channels. Enough consumer routers will degrade the signal. You could also re-use the channels if you have a good F/B ratio on your antennas, they’re far enough apart and face opposite directions.
I have multiple towers with 3 sectors and a couple with 4. I have both configurations, some RB’s with 3 radios on them and some with 3 seperate RB’s with one radio each. The single board/multiple radio setup seems to work just fine until you reach about 30-40 subs on each sector. Then it is definitely worthwhile to switch to seperate boards. I stay away from using ch6 alltogether because every cable/dsl router out there is blasting away on 6 and the noise is usually very high. As for interference within the enclosure with all three radios running on one board, I haven’t noticed any real effect even with radios running on partially overlapping channels. My opinion is that you would get just as much self interference or more from 3 or 4 sectors on the same tower as you get from having the cards in the same box. After all, your antenna is a much better radiator than the connector inside the box, and typically, unless your tower is 50 feet across at the top, the near field radiation of the antennas is going to be seen by all the other antennas on the tower. Keep in mind that antenna patterns you get when you purchase your antenna are the far-field patterns. At 2.4 GHz we are talking about a distance of about 50 feet before these patterns can be accurately measured. Any closer and it is virtually impossible to tell what the radiation is actually doing. If you were to measure it at close range, my guess is you would see variations in the signal strength even while standing still. So, bottom line is, don’t worry about having them in the same box unless you can’t seperate by at least 2 channels. And, if you run them on a higher powered platform, I’d be interested to know how many people you gan get on before performance starts to degrade.
Thanks Cameron, very helpful indeed - will heed your advice.
Since I posted here we have done a lot of lab test work on inteference between adjacent antennas on adjacent and non-adjacent channels in the 5GHz spectrum - very interesting results (needs collating to a white paper, will post a link).
Regarding the “high power single box” concept we tested our 1GHz outdoor router in the lab with 3 atheros cards, 5GHz and 5GHz Turbo, 5 clients attached and spooling 130Mbps continuous test traffic split between the 3 sectors, with only 40% CPU usage (and bandwidth test terminating about 50Mbps on the box).
Now of course, that was test traffic, and 1-2 not 40 users per sector, but I have to say I was impressed …
256MB RAM (can be any up to 1GB)
1GHz AMD mobile Athlon
(we also have Pentium M, also excellent, but more expensive)
DC-DC converter for either for high-power POE or external 48V or 24V input.
The board is customised for our need, with the custom passive cooling solution integrated with the case.
We have Pentium M versions up to 2GHz if/as needed - but a 1GHz box seems to shift a lot of traffic!
send me a mail if you want any info
stephen[dot]patrick[at]cablefree[dot]co[dot]uk
A subject my co. invested a lot of time and effort on - it’s not trivial to solve.
Any high speed x86CPU (read: >500MHz) generates significant heat.
The rest of the chipset also produces some heat, but assuming “laptop chipsets” (most desktop ones run hotter, avoid them), it’s the CPU heat you need to get out. In our system, we get the heat out of the case by coupling the CPU out to a large external heatsink, which enables use even in quite harsh climates. We have them installed in several countries in the Middle East, which are some of the hottest climates in the world, with very good results.
I would post a product picture here but it’s MT’s forum, not ours, and advertising isn’t appropriate.
You can see it our our website though - an early version with the black heatsink - newer ones are more aesthetically pleasing.
I understand MT are making new routerboards that use other (powerPC) architectures that give you the MIPS without the heat, because x86 is not the most efficient architecture for embedded tasks like routing. One of those apparently has a small fan, so presumably still generates significant heat. But in the meantime, fast x86 boxes are the way to go.
Number of users - Up to 50/sector is recommended, though you can have more, and our box can run 4 or 5 sectors. User numbers limited by the airside protocols imposed by Atheros chipsets, not the processor, BTW.