Help with AP config

phil@nowires.com.au · Thu Nov 23, 2006 5:21 am

Hi,

I currently am using Cisco 1230 access points and would like to upgrade them to RB532 running ROS.

The combinations I have tried are:-
RB112 / SR2
RB112 / CM9
RB112 / NMP-8602+
RB532A / SR2
RB532A / CM9
RB532A / NMP-8602+
With 3 different types of u.fl pigtails.

However, all of these combinations deliver poor performance as access points serving smartbridges airbridge, Tranzeo CPQ or Edimax EW-7206 client devices. I have field tested the different combinations at a variety of sites on a variety of antenna combinations but performance is always extremely well below that of the cisco 1230 access points that I swap out.

I have noticed that performance seems to depend on the number of associated stations, where 2 or 3 clients are connected the performance is relatively consistant, however thruput is lower than the cisco. On busier sites where 15 or more clients connect, performance is much worse than the cisco, with ping times constantly changing from 5-1500ms many dropped replies and very low thruput (<500kbit/agregate). Plugging antenna back into the cisco and thruput increases to 4mbit agregate and ping times return to stable 5-15ms with very little packet loss.

What have I done wrong?

I have one RB112 with SR2, one RB532 with CM9 and one RB532 with SR2 at various sites running in parallel with ciscos (on seperate channels) and can test out any changes in config that you suggest.

Here is my wireless config using SR2. wlan1 is hotspot (masquraded to public IP on bridge1), wlan2 is bridged to ether1. No encryption, radius auth on wlan2, default auth on wlan1. ROS ver. 2.9.35.

/ interface wireless 
set wlan1 name="wlan1" mtu=1500 mac-address=00:15:6D:53:33:22 arp=enabled \
    disable-running-check=no radio-name="00156D533321" mode=ap-bridge \
    ssid="www.nowires.com.au_hotspot" area="" frequency-mode=regulatory-domain \
    country=australia antenna-gain=10 frequency=2412 band=2.4ghz-b/g \
    scan-list=default rate-set=default \
    supported-rates-b=1Mbps,2Mbps,5.5Mbps,11Mbps \
    supported-rates-a/g=6Mbps,9Mbps,12Mbps,18Mbps,24Mbps,36Mbps,48Mbps,54Mbps \
    basic-rates-b=1Mbps basic-rates-a/g=6Mbps max-station-count=2007 \
    ack-timeout=dynamic tx-power-mode=default noise-floor-threshold=default \
    periodic-calibration=enabled periodic-calibration-interval=60 \
    burst-time=disabled dfs-mode=none antenna-mode=ant-a wds-mode=disabled \
    wds-default-bridge=bridge1 wds-default-cost=100 wds-cost-range=50-150 \
    wds-ignore-ssid=no update-stats-interval=disabled \
    default-authentication=yes default-forwarding=no default-ap-tx-limit=0 \
    default-client-tx-limit=0 proprietary-extensions=post-2.9.25 hide-ssid=no \
    security-profile=public disconnect-timeout=3s on-fail-retry-time=100ms \
    preamble-mode=long compression=no allow-sharedkey=no comment="" \
    disabled=no 
add name="wlan2" mtu=1500 mac-address=00:15:6D:53:33:21 arp=enabled \
    disable-running-check=no master-interface=wlan1 ssid="nowires" area="" \
    max-station-count=2007 wds-mode=dynamic wds-default-bridge=bridge1 \
    wds-default-cost=100 wds-cost-range=50-150 wds-ignore-ssid=no \
    default-authentication=no default-forwarding=yes default-ap-tx-limit=0 \
    default-client-tx-limit=0 proprietary-extensions=post-2.9.25 hide-ssid=no \
    security-profile=radius comment="" disabled=no 
/ interface wireless nstreme 
set wlan1 enable-nstreme=no enable-polling=yes framer-policy=none \
    framer-limit=3200 
set (unknown) 
/ interface wireless manual-tx-power-table 
set wlan1 manual-tx-powers=1Mbps:26,2Mbps:26,5.5Mbps:26,11Mbps:26,6Mbps:26,9Mbp\
    s:26,12Mbps:26,18Mbps:26,24Mbps:26,36Mbps:24,48Mbps:22,54Mbps:21 
set wlan2 
/ interface wireless security-profiles 
set default name="default" mode=none authentication-types="" \
    unicast-ciphers="" group-ciphers="" wpa-pre-shared-key="" \
    wpa2-pre-shared-key="" eap-methods=passthrough tls-mode=no-certificates \
    tls-certificate=none static-algo-0=none static-key-0="" static-algo-1=none \
    static-key-1="" static-algo-2=none static-key-2="" static-algo-3=none \
    static-key-3="" static-transmit-key=key-0 static-sta-private-algo=none \
    static-sta-private-key="" radius-mac-authentication=no group-key-update=5m 
add name="radius" mode=none authentication-types=wpa-psk,wpa2-psk \
    unicast-ciphers=tkip group-ciphers=tkip wpa-pre-shared-key="" \
    wpa2-pre-shared-key="" tls-mode=no-certificates tls-certificate=none \
    static-algo-0=none static-key-0="" static-algo-1=none static-key-1="" \
    static-algo-2=none static-key-2="" static-algo-3=none static-key-3="" \
    static-transmit-key=key-0 static-sta-private-algo=none \
    static-sta-private-key="" radius-mac-authentication=yes \
    group-key-update=5m 
add name="public" mode=none authentication-types=wpa-psk,wpa2-psk \
    unicast-ciphers=tkip group-ciphers=tkip wpa-pre-shared-key="" \
    wpa2-pre-shared-key="" tls-mode=no-certificates tls-certificate=none \
    static-algo-0=none static-key-0="" static-algo-1=none static-key-1="" \
    static-algo-2=none static-key-2="" static-algo-3=none static-key-3="" \
    static-transmit-key=key-0 static-sta-private-algo=none \
    static-sta-private-key="" radius-mac-authentication=no group-key-update=5m

Thanks in advance.

Cheers...ƒil

phil@nowires.com.au · Sun Nov 26, 2006 12:53 pm

Any ideas?

mneumark · Mon Nov 27, 2006 6:04 am

Your best bet is pick a data rate that the system works at and is stable at and set the support rate to that on both sides. This will force the MT to stay at one level and reduce some of the jumping that it will do.

phil@nowires.com.au · Mon Nov 27, 2006 6:23 am

Gday Mneumark...Thanks for your reply.

That sounds more like a workaround than a solution.

for example, if i was to lock the rate at 11M, and we experienced adverse weather causing signal fades, my customers could drop off, instead of stepping back to the next lowest speed. If i was to lock the rate any lower, this would cause much lower thruput than what i currently have.

Biggs · Mon Nov 27, 2006 6:28 am

What's are the ping time if the AP is in B mode only? I have 2 sites that have an erratic ping time to the clients when I set the AP in B/G or G only, but works flawlessly in B mode. Independently of the distance of the CPE 100m to 7km. Both are running Atheros based card and have more than 15 clients connected at the same time.

I have two other sites that are working well in G mode. The first have 6 clients and the other have 5.

phil@nowires.com.au · Mon Nov 27, 2006 10:50 am

Howdy Biggs,

Thanks for your help.

I changed the mode to b only. That improved the situation, but still no where near as good as the cisco.

For example, i'm 12.5Km from the AP using a airbridge outdoor CPE.
the ABo signal strength is -59dBm with ~80% link quality. At the access point (the config posted above) the received signal is -80dBm with a noise level of -96dBm, a SNR of ~16dB. The TX CCQ is 75-90 continuously varying. ACK timeout is 320us.
I am the only client device attached to the MT access point.

My ping times to the MT accesspoint are Min=5ms, Max=936ms, Avg=42ms, 350 sent, 338 received, 12 lost, 3% loss. In that time I attempted to make a Voip call, which was terrible quality. Then FTP a 4MB file from a local server, max bit rate acheived was 339Kbit/s, avg 242kbit/s. time taken 2min 29sec.

I then attached to the cisco AP on the same tower Min=5ms, max=32ms, Avg=6ms, 350 sent, 349 received, 1 lost, 0% loss. The Voip call was glitch free, good quality, and the FTP was max 1612Kbit/s, avg 1450Kbit/s, 23 seconds.
Cisco AP stats are -79dBm, SNR=10. ABo stats are -60dBm ~90% LQ.
There are 13 other clients attached to the cisco at the testing time, distance ranges from 0.5Km to 20Km.

I hope we can get a resolution.

galaxynet · Mon Nov 27, 2006 5:23 pm

Phil -

As an RF engineer for 30+ years......and a MT user for long while.....

1) I was noticing your RF stats of approx -80 and -60. I've have several MT APs running around out here and I have to say that typically I see almost the same sed/rcv signal levels at each end. That is to say if I have -60db at the client end I have -60db +/- 3db at the AP end. The MT AP could have either an omni or a sector antenna - doesn't seem to matter the stats hold true.... (as it should - theory bears this out).

2) Many of our APs typically host 20 to 40 clients. I do see slight increases in ping times but only 1-3ms under a decent load. Even under a full load using windows BW tester from end to end only seems to add another 1 -3 ms to the average ping time under normal loading.

The earlier posts have validity - locking down rates or changing to B mode only. I have not had to do that where we are located. I do use this methodology when I set up data backhauls - just seems prudent to know what you should get all the time and if you need more then setup another radio pair.

The only thing that sticks out as an issue for me is the disparity between your snd/rcv signal levels. It is probably environmental. I had a link once that use to fade in the summer - after much time spent looking at this I 'discovered' that it only faded when the large body of water in between the AP and client was absolutley still - The signal level at the client faded to less than -90db. The AP saw the client just fine.... I ended up changing the Client antenna and aiming it slightly 'over' the AP's omni (8 miles away) that change in the antenna's focal point made the multi-path interference (read as multipath signal cancellation as that was what was happening) became negotiable - less than 1db. I don't know if this will help you but I'd certainly be looking at the disparity between snd/rcv signal levels.

Cheers
Thom

phil@nowires.com.au · Tue Nov 28, 2006 1:42 am

Hi Thom,

Thanks for you help.

I understand what you are saying about the difference between the signal levels. The only difference should be the 8.5dB in transmitter output power.

MT AP -> ABo Client
(MT AP +26dB) - (cable+connector -2dB) - (sector +10dBi) =34dBi
FSL@12.5Km = -122dB
(Client antenna +21dBi) - (-1dB cable loss) = 20dB
Total at the Client N connector= -68dBm

Abo Client -> MT AP
(ABo +17.5dBm) - (cable -1dB) - (Client Antenna +21dBi) = 37.5dBi
FSL = -122dB
(sector +10dBi) - (cable -2dB) = 8dB
Total at the U.fl connector = -76.5dBm

The only thing i can attribute the large differences in predicted to actual signal levels could be the calibration of the equipment at either end. I don't have a 2.4G spec an or power meter to verify.

To test these ideas I backed off the transmit level of the SR2 card to 17dBm now at the client ABo i get -70dBm however this only made the situation worse, 20% loss avg=240ms...avg thruput was around 120kbit/sec. This is in B-only mode with MT and client locked at 2Mbit.

I have another site where I'm using a SR2 on a 10dBi Omni. There is one attached client using ABo with 21dB antenna, his distance is only around 2Km. The ABo RSSI=-20dBm signal quality~87%....MT -74dB, SNR=25, TX CCQ 90~98. Pings are 1% loss, Min=8ms, Avg=14ms, Max=153ms
I am unable to test for thruput as i'm not on-site...but from memory when I tested a week ago, max was about 1200kbit/s avg was around 600kbit/s.

I locked both the ABo and MT to 11Mbit, and the MT in b-only mode and reduced the TX power to 17dBm, however this only made the situation worse with quite a lot more dropped packets than usual (25% loss) but the packets that did make it had a lot more consistant ping times Min=8ms, Avg=11ms, Max=37ms.

As for environmental changes in free space loss...I operate in a very dry area, no large bodies of water. The only environmental effect that I have noticed is on very still sunrise/sunsets in spring or autumn we get catastroptic fades for several minutes. I assume that this is due to the layering of different tempreture airmasses (inversion layers) causing signal ducting or signal skip. Usually they last only a few minutes, but have seen it for an hour once, in a very dense low fog. (one site is on a mountain 500M above the level of most clients who are in a river valley, the fog was mainly on the river flats...clear at the transmitter site). Thankfully we don't see many environmental problems on our 5.8G backhauls.

I have been involved in the broadcast industry for 7 years, I have seen the problems you describe on microwave hops over water. It is especially hard to combat where the level of water is constantly changing (tidal). I think the best solution in that circumstance was a frequency diversity system.

Regards...ƒil

nickb · Tue Nov 28, 2006 8:54 am

I will say that I have a very few (about 5) clients using old smartbridge outdoor radios, and I have been very unhappy with their performance on MT access points. This is not the first time I have been unsatisfied with the airbridge outdoor, which is why we purchased so few units of them.

galaxynet · Tue Nov 28, 2006 5:24 pm

Phil -
Thanks for the indepth RF analaysis - that helps eliminate some of the more obvious 'stuff'. Your figures are accurate for the setup you have.

A couple of things that might help.....

Do a freqency survey from the MT and see what's out there that it might be seeing.... Can you do a survey from a client radio to see what it is seeing as well?

Set your wireless preamble mode on the MT to long.

Perhaps set the proprietary extensions to 'pre 2.9.25' vice 'post 2.9.25'.

Take one of your MTs and set it up as a client and 'attach' in to the MT AP and see what you get.

Set power for your wireless cards to 'default' vice manual....

Is there anything in the logs (MT or your other radios) that suggests something else is going on - disconnects, security (key) problems, etc?

What ver of ROS and what is the RB532 and RB112 firmware versions you are running?

Is there a reason you are using WDS in an Ap - to - client situation?

phil@nowires.com.au · Wed Nov 29, 2006 2:24 am

Gday Thom + NickB,

Thanks for taking the time...

NickB:- I had a lot of trouble with AirBridge prior to the 0.01.14 firmware becoming available. Now they are rock solid. Point taken about AB on MT, just they work so well with other AP's and I have over 30 that are owned by customers...so I need to support them.

Thom:- Scan from AP shows two motel hotspots on channel (2412Mhz) at -90~-92, One ptp bridge cochannel at 2422Mhz at -95. My Cisco AP's on 2437Mhz at -52 and 2462Mhz at -45.

ABo client shows (MT) 2412Mhz = -61dB, (cisco)2437Mhz = -59dB, (cisco)2462Mhz = -70dB.

Preamble=long (airbridge work best on long...i use it by default)

Proprietory extensions change made no difference

The power mode has been default for the above testing. I have set it to manual a few times to check the different effects.

Nothing in the MT logs...I have a few clients that have been connected for over 9 days (where i removed the cisco's, and had no choice but to leave them on the MT) and nothing other than auth request and radius query. I'm running the MT's with no security for the moment. AB has no log.

I have tried ROS ver 2.9.26, 2.9.30, 2.9.31, 2.9.33, 2.9.34, 2.9.35 currently on 2.9.38. Initially the bios was 1.2 (from memory) from factory, but I have updated it to 2.5 before the above testing.

I initially had WDS disabled, but was keen to try anything to see if I could boost the performance....with it disabled there seems to be no change.

As for the RB based client device...i'll give that a try, I did have a RB112 setup as a client to a cisco using a CM9 card, but was disapointed with the thruput so it was reconfigured as a 5gig backhaul (which works great). I'll have to wait until some favorable weather to get one down off the tower. I'll post the results when I get it configured.

Cheers...ƒil

galaxynet · Wed Nov 29, 2006 5:34 pm

Hi Phil - GDay to you too.

Ok here are a few other things to consider....

How are you powering your MT APs? PoE I assume but at what VDC? 12/24/48?

There is a 'wireless debugging' option on the MT under logging - turn that on and see what shows up.

Are you only seeing poor performance on the AB or is it everything you have out there?

What is the physical seperation between your antennas on your tower? Should be at least a meter....

I take it the client (when using the Cisco AP) has no issues w/Internet services. But w/using the MT 'he' does?

Have you tried switching the channels around - put the MT on 2462 and that cisco on 2412..... There could be some 'broad spectrum' interference that you're not seeing w/a scan of the frequencies...a lot of 2.4 phones use the lower end of the freqency band and don't show up on a scan....

Satellite tx/rx can also cause issues on occasion - though they shouldn't because of the frequency bandpass of the radio cards and antennas....but it has happened.

I know, I know, if the Cisco can do it why not the MT...I've asked that question more than once (and vice versa - the MT can do it why not XX).... It's one of the mysteries of RF - I've seen things that should never work - work great, and others that should that don't - 30yrs as an RF engineer - I've seen a lot of 'magic'!!

Ok Phil - be safe out there.

Drop a line when you've got something new for us to consider.

Thom

nickb · Wed Nov 29, 2006 8:15 pm

I had a lot of trouble with AirBridge prior to the 0.01.14 firmware becoming available. Now they are rock solid. Point taken about AB on MT, just they work so well with other AP's and I have over 30 that are owned by customers...so I need to support them.

Well, I'm not saying they don't work. I've just not been real happy, have had a few oddities with the airbridge outdoor units. In general, they work ok though. I understand your needs though - our clients buy their own gear too, hence why there are still some out there!

It's one of the mysteries of RF - I've seen things that should never work - work great, and others that should that don't - 30yrs as an RF engineer - I've seen a lot of 'magic'!!

Thank $DIETY, I'm not the only one!!

phil@nowires.com.au · Thu Nov 30, 2006 3:16 am

Howdy Thom + NickB,

I use 48V POE for supply.

With wireless debug on, still nothing. the airbridge associates and stays accociated until I get sick of the poor performance and reassociate to the cisco.

I have a few tranzeo-CPQ devices but unfortunatly they aren't in the MT sector. I did try them when I had the MT connected to an omni and thier performance was pretty well on par with the AB. The other client adaptors we have out there are cheap Edimax EW-7206APg and Minitar MNWAPB(RaLink chipset), thier performance with MT is the worst of the lot. In my laptop I have a cisco 350 minipci, it has the same poor performance when I was the only client associated to MT and standing below the tower. When i go onsite (to retrive a RB112 for testing as a client) i'll replace the laptops 350 with a CM9 and see if the performance increases.

The two sector antennas are seperated by 4M horizontally and the sector to omni spacing is 8M vertically. As a test, I disabled the cisco driving the adjacent sector antenna, which made no change.

When we switched the ciscos over to MT at the main site my phone started running hot. The main complaints was "why does page cannot be displayed keep coming up" which I take to mean packet loss and "why are pages taking ages to load all of a sudden" which I took to mean low thruput. A week later when ciscos went back in, no issues.

Switching the channels around made little difference...the MT indicated that the noise levels were -95 on both channels. Throuput and ping time were almost equal on both the cisco and MT using either channel.

The site in question has a lot of satellite services but they all look like they are receive only, ie only have LNBF and not HPAs in the focal point. I have another site that uses a bi-directional amp on a low gain omni (with long cable run) and to maintain its stability we had to use a 3pole ceramic bandpass filter between the omni and amp to get rid of the (presumed) out of band interference...I never did locate the exact source, I assumed it was the 1000W EIRP TV or FM transmitters located on the same tower... but that site did have a 2-way satellite service on another hut around 20M away....
I'll order a filter to test on the MT...worst that can happen is a 3dB loss thru the passband.

I could understand if poor performance was a feature of one site only...but i have MT AP's on three sites and performance is pretty well the same on them all...I have one site in testing that is MT only, one 5G backhaul and one 2.4G sector, using my laptop to test the 2.4G i get the same inconsistant ping times and low thruput. However, if i connect to the RB ethernet...performance is great. I have had other people associate using a variety of hardware and hotspot 'demo' thruput has been low.

I know what you mean...there is plenty of smoke and mirrors in the RF 'black art' we dabble in. I have conjured a magic trick or two in my time!

Thanks again for your help Thom

Regards...ƒil

galaxynet · Thu Nov 30, 2006 6:12 am

Hi Phil -
Well I am a at a bit of a loss to tell you where to go to next concerning the RF angle - Support might help you though they can be slow....

You did mention using an amp - is that just one location? Amps and MT, especially w/atheros cards don't work well at all....

Anyway - I have MTs as APs with different types of client radios - however none of them are of the types you use (how odd....) - I use zinwells, teletronics, zyntek, I've used some linksys usb adapters (as client), prism laptop cards, I can't remember the old style client (they're not manufactured any more) - anyway they all seem to work just fine. Of course we've used MT to MT and that always works great.

Well there is one more thing - I did run in to an issue w/routing / nat'ing, etc. Would you post your config (mask out 'XX' any ip info you don't want to be public) here include routing, mas'q, nat'ing, IP address on interfaces, firewall rules, etc. The issue was extremely hard to locate - required trial an error to narrow it down. It came from the routing and nat'ing/mas'q. The issue showed itself in the way you are describing IRT what your clients are seeing (and of course ping times). Put it this way, a bandwidth test showed I can get 800KBps or 14MBps...I fixed the issue and now get consistant 14MBps. I 'introduced' the error when I put a new router on line coming from ROS 2.28 to 2.29... Post your config and lets have a look...

Thom

phil@nowires.com.au · Thu Nov 30, 2006 7:34 am

Hi Thom,

Yeah, we amp our omni's to overcome the loss thru filters and long cable runs...we have always used teletronics 802.11b AGC TDD BDA and they have provided great performance with the cisco's. Much more range and thruput with the amp than without. I notice that most people on this forum don't like amp's....all I can say is we have had no issues using amps with cisco's. I did notice that performance was much worse when I was using a CM9/RB532 with a BDA'd omni than on an un-amped sector.

I have searched for Zinwell and Zyntek availability in Australia but I cant find any suppliers. I am familiar with the teletronics EZ-bridge, however they were expensive compared to the AirBridge family, so we never got further than looking at sales material.

Address:-
Flags: X - disabled, I - invalid, D - dynamic 
 #   ADDRESS            NETWORK         BROADCAST       INTERFACE
 0   10.200.1.12/17     10.200.0.0      10.200.127.255  bridge1   
 1   ;;; hotspot network
     192.168.22.1/24    192.168.22.0    192.168.22.255  wlan1    
 2 D xxx.xxx.189.184/26 xxx.xxx.189.128 xxx.xxx.189.191 bridge1   

Route:-
Flags: X - disabled, A - active, D - dynamic, 
C - connect, S - static, r - rip, b - bgp, o - ospf 
 #     DST-ADDRESS        PREF-SRC        G GATEWAY         DISTANCE INTERFACE
 0 ADC 10.200.0.0/17      10.200.1.12                                bridge1  
 1 ADC 192.168.22.0/24    192.168.22.1                               wlan1    
 2 ADC xxx.xxx.189.128/26 xxx.xxx.189.184                            bridge1  
 3   S 0.0.0.0/0                          r 10.200.0.20     2        bridge1  
 4 AD  0.0.0.0/0                          r xxx.xxx.189.129 0        bridge1  

Firewall:-
Flags: X - disabled, I - invalid, D - dynamic 
 0   ;;; masquerade hotspot network
     chain=srcnat src-address=192.168.22.0/24 action=masquerade

Bridge:-
Flags: X - disabled, I - inactive, D - dynamic 
 #    INTERFACE BRIDGE  PRIORITY PATH-COST
 0    ether1    bridge1 128      10       
 1    wlan2     bridge1 128      10

I use a DHCP client on bridge1 to deliver dynamic public IP, GW, DNS, NTP settings. The 10.200.x.x address is for backup...on our "device management" network. Ether1 connects to our router, Wlan1 is hotspot, Wlan2 is VAP of Wlan1.

Thanks Thom

Cheers...ƒil

phil@nowires.com.au · Thu Nov 30, 2006 10:31 am

I configured a RB112 / SR2 with 24dB grid as a client today, 48V POE, wlan1 bridged to ether1, no nat, no firewall, no connection tracking, no routing, no mangle, no queues... I'm at a distance of 12.5Km.

AP is RB532 / SR2 with 10dB sector, config is as above.

The results are:-

With client in AP bridge + WDS mode
1000 pings, size 100B, Min 4ms, avg 7ms, max 206ms 0% loss
1000 pings, size 1500B, Min 10ms, avg 55ms, max 692ms 0% loss
FTP avg 254kbit/sec max 381kb/sec

With client in Station WDS mode
1000 pings, size 100B, Min 7ms, avg 11ms, max 258ms, 0% loss
1000 pings, size 1500B, Min 9ms, avg 15ms, max 908ms, 0% loss
FTP avg 251kb/sec max 379kb/sec

AP side signal -78dB, noise -96dB, CCQ 59/67, rate 24Mbits, CPU 6-14%

Client side Signal -79dB, noise -99dB, CCQ 60/64, rate 36Mbits, CPU 4-11%

I was the only client associated. AP in configuration listed above.

I am try from closer to the AP for the next test to see if this could be a factor of distance.

Regards...ƒil

galaxynet · Thu Nov 30, 2006 5:26 pm

Phil,
Your test results first -

Amp'ed or un-amp'ed, either or both? If you're using the amps..you need to cut the power to the amps way down like to 8 - 10db. ROS does not set the power out to the values you set it at - it's the set value + 4-7db, and if I remember right the TT amps max at 100mw or 20db - too much in power wreaks havoc w/the amps.

Why aren't you placing the radios closer to the antenna's instead of making long cable runs? (curious George here...)

Teletronics amps look like the 'B' only type - are they? I know I tried using the 'B' amps for a 'G' network (they said it would work 2 yrs ago now) it didn't work for beans - as soon as I put the radio in 'B' mode I got great performance. We were doing some experimenting, Non ROS type radios....

Your CCQ factors are not that good - at 12.5km you should see better figures than that as well as better signal levels..... I've got a couple of 12 mile (~20km) links and I have as good a signal as you do with better CCQ.... 14db client, 12 db AP antennas. Even so you should get better ping & throughput...

Looking at your earlier code - do you have the cards set to automatically figure calibration values - it looked like it was set manually?

What are you pinging when you test? 10.200.0.xx or your public router @ xx.xx.189.129?

On to your config -

Why do you use bridiging instead of routing? Bridging adds a lot of overhead traffic 10 - 50%...

Connection tracking - you pretty much need to use that all the time...funny things sometimes happen if you don't use it.

Under Route - I see;
3 S 0.0.0.0/0 r 10.200.0.20 2 bridge1
4 AD 0.0.0.0/0 r xxx.xxx.189.129 0 bridge1

Which pretty much says anything that 'it' doesn't 'know' send to 10.200.0.20 - is that where you what the default route to be? Also that route does not have an 'A' in front of it meaning it is not an active route... For the 189.129 route - no 'C' meaning it's not connected.... I know you're using bridging and it shouldn't matter but....

I did not see a section on rule routing; " ip route rule print " which you should have least this rule: " src-address=0.0.0.0/0 dst-address=0.0.0.0/0 action=lookup table=main " or lookup whatever table you want it to look up... Again - I realize you are using bridging...

Unless you are using/plan to use some type of 'cell network' meshing don't use WDS - it's an extra unecessary step.... I wouldn't think you'd need this as it looks like, except for your hotspot, all your clients are at fixed locations....

Ok Phil - that's all I have for th moment - I'll think on it a little today and wait for your reply..

Thom

thom.lawless@rapidwifi.com - if you want to drop me a line direct....

phil@nowires.com.au · Fri Dec 01, 2006 4:02 am

Howdy Thom,

We use amps for omni's only. On this particular tower...I currently have a cisco on the amp'd omni, I tried a RB532 / CM9 on it but latency and thruput were horrible, so the cisco was reconnected. The ciscos have a 100mW / 20dB max output, i've never touched their output power control. I thought that CM9's were 100mW (on 802.11b), does MT overdrive them in default power mode? I didn't try a SR2 on the amp'd omni.
The RB532 / SR2 is currently on an un-amped sector.
I also have a Cisco on an un-amped sector.

...Just for curious George... Ciscos support antenna space diversity, so we actually have 2 sectors spaced horizontally apart but facing the same direction connected to the cisco. It provides little benefit with fixed clients, but a huge benefit for laptop / mobile clients. (+3dB to +6dB extra RX, compared to the omni, at the AP)...Hopefully MT will see the sense in it and support space diversity (i have read lots of posts asking for the feature).

We have minimised cable distance as much as possible...it mainly due to the construction that we use. On a normal triangular lattice freestanding tower we mount 2x 6M pipes horizontally in parallel on one face spaced about 2M apart, we then run 2M vertical pipes between the horizontal pipes at the extreme edges. We mount the sectors to the 2M vertical pipes. We then mount the omni, 3 pole filter, amp on a 6M length and hoist it vertically to clear the top of the tower. A waterproof enclosure mounts to the tower between the horizontal pipes...so usually we have 4 x 4m runs to the sectors (2 a side) and a 8M to the omni/filter/amp up the pole. The 3 pole filter has an unavoidable 3dB loss thru the passband.

Yeah, TT 250mW, 500mW, 1000mW b only. I only have ever tried them in 802.11b mode....I assume that the much denser modulation (of 802.11g) would cause saturation of the PA module causing distortion, resulting in more noise than signal.

FYI....I keep the output power of the omni as close to max legal output as possible (4W eirp) and have the broadcast ESSID set to "www.nowires.com.au" (my website). This provides a huge amount of free advertising...as many computers pop up and say "www.nowires.com.au is available for connection" people get curious and check it out. A major reason for the upgrade to ROS based AP's was to be able to offer these sorts of people hotspot (and free internet trial, free access to my webpage).

I would have thought the CCQ would be higher...i'm staying at a different place for the weekend, i'll take a tripod, antenna, long cord and RB112 with me and do some testing there. Its only 2km from a site.

I have tried automatic calibration in default, enabled and disabled.

A variety of sources...MT-MT I ping from one MT console to the next MT, AB-MT or AB-cisco I ping from pc directly connected to AB to the AP. PC statically set to 10.200.x.x network.

We use bridging because of the ciscos....when they are able to comunicate at layer 2, clients roam very quickly (keeps most connections active), so for instance i can be in a car driving around and keep a constant connection. On the tower the omni provides some redundancy if a sector AP needs to reboot, clients roam to the omni without noticing a loss of connectivity. (must be some sort of cisco proprietory RSTP with bridge communiction). Routing breaks this ability.

I thought connection tracking was a NAT/masq./proxy type of thing. I could be wrong but i can see no logic for it being used on a simple bridge. I have it enabled everywhere i perform NAT/masq...off everywhere else.

The routes....no. 3 is a 'backup route' incase the MT does not get a DHCP lease and a way for me to manage it on my "management network".
no. 4 is a dynamic public IP that is used for hotspot functions, it becomes the default route when it is leased (as well as delivering DNS, NTP)

I think that D implies C.

Hmm....I haven't got any route rules in any of my MT's...i'll have to research this.

Really i only had WDS turned on for testing....however, i turned it back on now when using the RB112 as a client, so my PC's can be directly bridged onto the network. The only way i could get all my programs to work properly with station mode was to have an external IP's on the RB wlan1 and 1to1 nat everything to a private network between RB ether1 and 1 of my PC's. I guess this is due to ROS not being able to 'mac nat/masq.' as the conventional client devices do (i think this will be available in ROS v3).

Thanks Mate,

Regards....ƒil

galaxynet · Mon Dec 04, 2006 3:17 pm

Hi Phil -
No - MT does not overdrive the CM9's. I was only making sure that you're weren't using SR2's at full power w/an amp....

Thanks for the curious George answers - I have a MUCH better picture of what you are doing now - makes sense where it didn't before... As in you have two primary functions - roaming (car like access) and fixed access to client's home.

You can achieve some 'spatial' diversity w/two radio cards - bridged to each other and your Ethernet interface (to maintain your bridging capabilities). One to your sector antennas (w/a power divider) and one to your omni.... A little more work and expense but it will work. You could actually use three cards if you wanted to skip the power divider and bridge all the cards and ethernet interfaces together....

I have tried bridging w/o connection tracking - as I said before strange things happen when its not turned on - though they do say it's not required - turn it on anyway.

"I think that D implies C" **- not necessarily!

You have to have a default rule... Don't ask me why - I just know from experience that you do - I have tried routers without it and the results can be very very strange.

"The only way I could get all my programs to work properly with station mode was to have an external IP's on the RB wlan1 and 1 to 1 Nat everything to a private network between RB ether1 and 1 of my PC's. I guess this is due to ROS not being able to 'mac nat/masq.' as the conventional client devices do" **** - I use a variety of software via masq - they all work great. I've also bridged wlan - ethernet same results - works great (in station mode). I pretty much only use routed methodology now - saves on overhead traffic and is easier to secure network(s). In all of this I have always used connection tracking and default rule w/a default route. I think in my earlier posts I gave you a basic config.

Phil - I don't sell or get anything from MT - I have just used the heck out of them - I like the flexibility - I have RBs (112, 230, 5XX) and 'home brewed' 'plain-jane' PC based units both as 'core' routers and wireless hi-end (70Mbps) wireless router units. I just wanted you to know I have nothing to gain here except to see another entrepreneur make it....

Regards,

Thom

phil@nowires.com.au · Tue Dec 05, 2006 1:19 am

Gday Thom,

If i can get the MT based AP's going, i'll give the bridged based multi-radio diversity a go...I have been recently trying out some 802.11n draft 1 radio gear based on the atheros XSPAN chipset....very exciting...300Mbit/s air side data rates, 100Mbit+ thruput. Hopefully MT supports this chipset sooner rather than later, it supports 2 or 3 spacial diversity antennas.

I have switched connection tracking back on on all of my routers and access points. I didn't notice anything obvious change/improve....i'll leave it on.

There was no real reason for me using DHCP to deliver IP/GW/DNS/NTP, so now I have configured them statically.

I added the rule route as you suggested above.

Some of my network monitoring programs rely on being directly connected to the network being monitored...assume to intercept broadcast traffic/multicast traffic. I have had no problems with masq. when using MT as a NAT router.

I understand that you have no association with MT...I thank you very much for your help. I too have many RB / x86 based routers and backhaul links that i have been using for several years. I have been able work out most config from the manual and trawling thru the forum. Just the AP problems have me stumped. Especially when I have read about so many happy MT AP users with great success.

Results of the 2km long RB532 / SR2 / un-amped 10dBi sector access point to RB112 / SR2 / 21dB grid link. This site has 4 clients, on other radios/antennas/channels. Only my client associated to the RB532.

ROS ver 2.9.38, BIOS 2.5 both ends.

AP:- (AP bridge + WDS)
ack. time 56us
signal -70
TX signal -69
SNR 30
CCQ 77/85
TX/RX speed 54-48

Client:- (in station WDS)
ack 56us
signal -70
TX signal -70
SNR 29
CCQ 71/78
TX/RX speed 54-48

I am able to achieve 2200Kbit/sec in FTP with avg 1735Kbit/sec.
Pinging from MT client-> MT AP
1000 x Ping size 1500, min 5ms, avg 11ms, max 103ms, 0% loss
1000 x Ping size 100, min 4ms, avg 9ms, max 73ms, 0% loss

At the moment, I don't have an airbridge or tranzeo to test in this location for comparison, but i'll get one and post the results.

Do you think that atheros chipset is just more sensitive to noise than cisco, (ie causing the CSMA/CA backoff mechanism to be holding off TX)?

Kind Regards...ƒil

galaxynet · Wed Dec 06, 2006 4:18 pm

G' Mornin Phil (my time...)

I been to your neck of the woods - in the 70's when I was on R&R from 'nam. And through the 80's and 90's as a civilian RF/computer engineer.

Any way - on to your results....

Looks like your throughput went up to what you were getting w/your Ciscos and ABo's. Ping stability is much improved.

I can tell you that WDS and bridging cost time in ROS. Bridging because of all the broadcast traffic, WDS due to overhead processing power (CPU) and extra broadcast packets.

When you were doing your tests - did you happen to look at the resources page (winbox) for your client RB112 - I have only a couple of RB112's and they work great at the client end but I have never looked at the CPU utilization during a flat out speed test..... Maybe look at that - the CPU power may have limited your throughput at your end....

I see your ack time - that is dynamic right - you didn't set that manually?

You could 'lock down' you client side TX rate at 36Mbps - that would keep it from switching from 48 to 54... Rule of thumb is one rate lower than the bottom rate will give you solid a transfer rate. It costs time to switch rates...

Did you notice how close your AP & Client TX/RX power levels were - notice that thay are almost the same? That's what I see when my units are properly aligned...

The chip set could be a little more sensitive - not sure - I have looked at the specs that are available but I don't see a big difference between the two....

I just put in a 2 mile (~3km) link last Friday, getting ~-66db at both ends, throughtput (connected right to a fiber node) was 15Mbps+, would be just about right for TCP, Nstreme, routed, RB532, connected to the particular source I was connected to ( I thought it should be about 20Mbps but I couldn't turn off the server to everyone else at the same time I was testing...).

Have you had the chance to test other radios on that link yet?

Ok Phil - have a good day out there!!

Thom

phil@nowires.com.au · Mon Dec 18, 2006 5:32 am

Gday Thom,

I did look at the AP resources, which was around 30% CPU, with >3MB of free memory while doing the speed test.

ack timing is on dynamic.

Yeah, i think it must be the calibration of the smartbridge device. The tranzeo is pretty spot on compared to the RB112 / SR2 client

I have quite a few RB / ROS based backhaul links on which i can get bridged thruput in excess of 15mbit. the RB112-RB112 bridged even get about 9-10Mbits. Why can't my AP behave like this....

Before i leave the place i'm housesitting (been here a week, out today) i tried out the tranzeo TR-CPQ-Nf device on the same antenna that i had the RB112/SR2 (didn't shift it, only unplugged the RB client) so we could compare apples-apples.

Thruput was around peak 2200Kbit/sec with avg at 1916kb/s.
Ping times are min 2ms, avg 8ms, max 37ms for 1000 x 1500B 0% loss
for 1000 x 100B, min 2ms, avg 6ms, max 104ms 0% loss
signal RX at tranzeo -69, noise -101, rate 5.5 - 11.
signal at RB AP. RX -71, SNR 30, ack 30us, CCQ 80~95
The tranzeo reports that 30% of the packets are retried...even with no data thruput, only a few pings and broadcasts. about the same with ~2mbits thruput.

So tranzeo is on par with the RB112 / SR2 at this location.

I'll be back home soon, so i'll do some testing using the RB112 / SR2 as a client to the cisco at 12.5KM range and post the results.

Cheers...ƒil

galaxynet · Sun Nov 04, 2007 3:46 pm

G'day Phil -
Been a while - thought I'd drop you a quick line and see how you're doing.... Haven't seen many posts from you lately so you must have gotten this under control with the new ROS releases.

Take care there 'down under'.

Thom

phil@nowires.com.au · Sun Nov 04, 2007 11:39 pm

Hi Thom,

Good to hear from you again,

I have given up on the mikrotik AP's and went back to using Cisco 1200's. The cisco's performance is far far superior to that of the mikrotik/atheros combination using omni antennas. I never could work out exactly what the problem was.... Maybe ciscos RF management?

Mikrotik still has a firm place in my network, I use it for all routers, p2p links, hotspot controllers, usermanagement, CPE..... Just unfortunately, no good when used as an access point coupled to an omni antenna on a high RF density sites.

Maybe i'll try again when 3.x is released.

Kind Regards...ƒil

Help with AP config

Help with AP config

Re: Help with AP config

Re: Help with AP config

Who is online