If you could do it all over ?

[rbolduc]

Hello All,
I am currently installing a new 64' tower at our pop and am going to change to horizontal pol with a waveguide antenna and move all of the clients off from the 35' vert omni. The new setup is on a fresh star server and will be running DNS and Proxy 100' lmr600 with a YDI indoor 500mw amp. I have made the steps and will be using routing with seperate subnets for each ap/repeter and will be using the 172.16.x.x using NAT at the pop. Also using MAC assoiation with tarpit enabled

This one is bugging me though..

1. Using PPPoE ver Static DHCP
A.PPPoE
a. Customers complain about having to "login" every time :x
b. More secure? as I don't have to have a ip bound to the radio
c. how do I see bandwith usage as users get different IP's
d. customer's can "roam" and login to any ap

B. S DHCP
a. easy to setup clients?
b. allways "on" with no login
c. CBQ can have a parent pipe

2.anything else, using wep?

Just looking for some input and pro's/con's. I learned the hard way about bridging and would like to light up this AP with without second guessing myself..

Thanks
Reed

[tim]

The radius server generates accounting records that show :

1) How long connected
2) Total traffic during session in & out

If you use PPPoE it shows the username of the user, if you use MAC authentication it shows MAC address (actually the username)

If you authenticate against an SQL database, this info is stored in the radacct table, and is easy to query.

This info is only generated at successful end of session so if sessions end prematurely you do get some records that are incomplete.

Tim

[lonnie]

You can do PPPoE and you will not regret it. Users can actually appreciate it as it logs them off and they are secure when not using the system. The "dialer" can be set to automatically connect, so no user intervention unless they wish it to be manual. All stats are available through radius accounting and the system is quite secure.

You can do radius of the radio card and have it automatic as well. Use this with static DHCP and you can have your CBQ rules in place. We will be working towrads automatica CBQ rules from the radius, but I have to stress that we do not have this now and I cannot provide an estimate for the feature to show up. Basically I want this for my own ISP business, but we have more basic requirements before this. It will come, just not this end of September release.

[bobbyc]

CISPA has recently tested loss through the best quality N-connectors and standard 12" pigtails... EACH N-connector had 1.5 dB loss (compared to the .5 dB rumor) and the pigtails had 2.3 dB of loss. So you're looking at 5.3dB loss before you add in your 100' of LMR-600 cable loss.
I would use a mast mounted amp in your case. The only tower that we have a indoor RFLINX amp is on our 9000'+ elevation mountain and we only have about 50' of the new LDF5-50B heliax cable.
Bob C

[rbolduc]

Thanks for the great input, but I have a question for Bob, in response to mounting the amp on the tower, Would it really make a difference? here is what I caculated.

Amp at radio:
the gain on the amp (ydi 500) is no more than 16dbi so as long as i have 11dbi to the amp it will produce full output

(14.5) + (-2.3) + (27) + (-3.00 + -4.43) +15 =34.57dBm EIRP
(radio) + (piggy) (amp) (n con lmr600) (antenna)

Amp on tower

(14.5) + (-2.3) + (-3.00 + -4.43)=4.77 to amp in +(16) +(15)=31.0 dBm
(radio) (piggy) (n conn lmr600) (amp max gain) (ant)

Am I even doing this right :lol: or will I have less total eirp with a mast mounted amp ?

Thanks
Reed

[bobbyc]

Edit: all the amps we have deployed are linear and not AGC, so you need to add the radio output and subtract cable/connector losses between antenna-amp and amp-radio.
I'm not sure... all I know is that having the amp right at the antenna it is able to hear the weaker clients better than having the amp at the end of a long cable run. Having the amp at the end of a long cable run and then using a high rx amp will cause more distortion.
We chose to do indoor amps at this tower because we've had horrible luck with outdoor mast mounted amps last winter. We had to send in every amp that we had deployed to be rebuilt. Even though they were installed properly, they got wet inside and everytime we got them rebuilt they said that water somehow leaked in through the LED lights. Since this tower is at 9000' altitude, it is hard to get to in the winter... so we choose to take no chances and have everything indoors and climate controlled.
I did my calculations like this:
13.5dB sector plus 15dB radio output = 28.5dB
28.5dB -1dB(sector n-connector loss) -3dB(2 n-connectors for cable length) -1dB(cable loss) -2.3dB(pigtail loss)
=21.2dB
36dB (max eirp) -21.2dB=14.8dB
So I ordered a couple amps with 15dB tx gain. The RFLINX amps come with a rx gain of 20dB, but I have a couple RFLINX amps in the mix already and don't like the noise floors in clean RF environments when their amps have 20dB rx gain, so I had them turn it down to 16.
Hope this helps~!

[funkywizard]

by having them reduce the RX gain 4db, does your noise floor drop more than 4db? and also, is this drop in RX gain only an advantage in a reletively noise free environment? In a more crowded environment, does having the higher RX gain make up for an increase in noise floor?

[bobbyc]

yes, the noise floor dropped improved by ~5dB. I imagine the signals dropped by about 5dB as well. In the end it's all about SNR.
Bob C

[georgew]

Remember, that though db's can be added and subtracted, not all db's are the same quality.

Amplifiers amplify noise as well as signal. Because of the way they work, they amplify noise better than signal. So amplifier db's are not of the same quality as antenna db's and coax db's.

This is just a rough swag, the actual numbers vary greatly...
But a typical amplifier will reduce the s/n ratio about a db for every 2 or 3 db's of amplification. Some are better or worse, but the point is to take this into account when deciding on the best system to build... Putting a smaller amp closer to the antenna will give much better results even if the db's you add up say otherwise.

The point is properly made antennas and coax tend to treat noise and signal the same, amplifiers tend to amplify noise more than signal. So in your system design calculations, try to get most of your db's from the antenna and coax system, and less from the amp. Don't be tricked into thinking more is better, be it on the transmit side or the receive side. You want the lowest gain possible from the amp to make your system reach maximum legal output... no amp would be better... I wish I had a 200mw orinoco card...

[rbolduc]

Thanks everyone for all the input, I am going to mount the amp on the tower as the numbers don't allways tell the whole story.

Lonnie, I have a question using static DHCP on the PPPoE connection with a hard coded QBQ. this is my radius info:


rbolduc Auth-Type := Local, User-Password == "password"
Framed-IP-Address = 172.16.0.100,
MS-Primary-DNS-Server = 172.16.0.1,
MS-Secondary-DNS-Server = 69.61.13.76,
# VNC-PPPoE-CBQ-TX = 128000,
# VNC-PPPoE-CBQ-RX = 256000k

and here is the CBQ

#172.160.0.0 Subnet PPPOE Users
qshape rbolduc 100 $384k_128k 172.16.0.100 on $mwi_0

with $384k_128k defined and be used on "regular" customers with no problems and mwi_0 is "wlan1" and is receiving the PPPoE requests. If I have "Assign a default CBQ rate to users" this is the speed that they get if I uncheck this option the users have "full" bandwith of my SDSL line up and down and the CBQ show very little or no traffic 20 or so K to a 5 MB Speed test. Is there I something I need to do different using PPPoE? :roll:

Thanks
Reed

[lonnie]

The IP can be given using dynamic DHCP with a default CBQ rule, since the radius can supply an over-ride CBQ if you want the user to have more or less than the default.

Since you can control CBQ by radius and PPPoE server and all the accounting is done, it does not require a static IP to keep control and do the accounting.

It is the easiest to setup and maintain.