Page 1 of 1

About tx-ccq and rx-ccq and low bandwith with strong signal

Posted: Thu Sep 15, 2005 6:02 pm
by aviper
I'm wondering how are they calculate ? I've got link 9 km and I'm getting very strong signal -49 to -50 dB. But those variables are standing low and unpredictable about 20 to 30 percents. The links are connecting with 36 Mbps, and 20 % are about 8-9 Mbps, which is the real amont of data transfered throw the link (watched by torch).
Where could be the problem?

status: connected-to-ess
band: 5ghz
frequency: nqma zna4enie
tx-rate: 24Mbps
rx-rate: 24Mbps
ssid: nqma zna4enie
bssid: nqma zna4enie
radio-name: nqma zna4enie
signal-strength: -50
tx-signal-strength: -49
tx-ccq: 20
rx-ccq: 30
wds-link: no
nstreme: no
framing-mode: fast-frames
framing-limit: 3999
routeros-version: 2.8.26

Station:

0 R name="Senao" mtu=1500 mac-address=**:**:**:**:**:** arp=enabled disable-running-check=no interface-type=Atheros AR5213 radio-name="*******" mode=station ssid="***" frequency=*** band=5ghz scan-list=**** rate-set=configured supported-rates-b=1Mbps,2Mbps,5.5Mbps,11Mbps supported-rates-a/g=6Mbps,12Mbps,18Mbps,24Mbps basic-rates-b=1Mbps basic-rates-a/g=6Mbps max-station-count=2007 ack-timeout=83 tx-power=20 noise-floor-threshold=default periodic-calibration=default burst-time=disabled fast-frames=yes dfs-mode=none antenna-mode=ant-a wds-mode=disabled wds-default-bridge=none wds-ignore-ssid=no update-stats-interval=disabled default-authentication=yes default-forwarding=yes hide-ssid=no 802.1x-mode=none disconnect-timeout=3s on-fail-retry-time=100ms

Bridge

0 R name="Senao" mtu=1500 mac-address=*** arp=enabled disable-running-check=no interface-type=Atheros AR5213 radio-name="***" mode=bridge ssid="***" frequency=*** band=5ghz scan-list=default-ism rate-set=configured supported-rates-b=1Mbps,2Mbps,5.5Mbps,11Mbps supported-rates-a/g=6Mbps,12Mbps,18Mbps,24Mbps basic-rates-b=1Mbps basic-rates-a/g=6Mbps max-station-count=2007 ack-timeout=83 tx-power=20 noise-floor-threshold=default periodic-calibration=default burst-time=disabled fast-frames=yes dfs-mode=none antenna-mode=ant-a wds-mode=disabled wds-default-bridge=none wds-ignore-ssid=no update-stats-interval=disabled default-authentication=yes default-forwarding=yes hide-ssid=no 802.1x-mode=none disconnect-timeout=3s on-fail-retry-time=100ms

I'm starting preparations for upgrade from 2.8.x to 2.9.x, but I don't think I'll resolve the problem.

Posted: Thu Sep 15, 2005 6:48 pm
by Alex
try 5-turbo and tx-power=default

Posted: Thu Sep 15, 2005 6:54 pm
by aviper
5ghz-turbo just increase the tx-strength to -60
tx-power=default should be tx-power=20, but I'll try it right away

Posted: Fri Sep 16, 2005 2:19 pm
by aviper
status: connected-to-ess
band: 5ghz
frequency: 5595
tx-rate: 24Mbps
rx-rate: 24Mbps
ssid: ***
bssid: ***
radio-name: ***
signal-strength: -52
tx-signal-strength: -50
tx-ccq: 49
rx-ccq: 48
wds-link: no
nstreme: yes
polling: yes
framing-mode: best-fit
framing-limit: 3200
routeros-version: 2.8.26

Ok still far from what I need (I'm targeting at 36 Mbps with ~ 70 ccq), but it is getting better.
Does nstream always improve soo much wireless bandwidth or it is only with my link ?

Posted: Fri Sep 16, 2005 2:50 pm
by stephenpatrick
Nstreme improves throughput only if
- there is enough CPU power to allow the overhead
(recent MT versions are much better in this respect)
- you are above a "marginal" signal level
(sometimes you have to "lock" the radio to a rate one lower than the max it tries to negotiate to)

Locking down the radio rates solves "varying ping times" and "varying signal levels" too.

Regards

Posted: Fri Sep 16, 2005 3:25 pm
by aviper
Nstreme improves throughput only if
- there is enough CPU power to allow the overhead
(recent MT versions are much better in this respect)
- you are above a "marginal" signal level
(sometimes you have to "lock" the radio to a rate one lower than the max it tries to negotiate to)

Locking down the radio rates solves "varying ping times" and "varying signal levels" too.

Regards
There is enought Horse Power in those machines (2 GHz 512 of ram). I've lock down the radio rate one step below, on default rates it connects at 36 Mbps rate. But still the problem with the low bandwidth exists. I expect from those radios real 24 Mbps (36 Mbps with tx-ccq~rx-ccq~70), now i get about 12-14 Mbps.

I don't know the reason and I'm still looking for it. I'll try 2.9.x in Monday 'cause in weekend I don't have access all the time to the room where is the bridge and I don't what to make any mistakes and lock the internet for 2 days :).

Is there anywhere description of the formula or the way of computation of tx-ccq and rx-ccq?

Posted: Fri Sep 16, 2005 3:33 pm
by stephenpatrick
Well your CPUs look fast enough -
Interference is possible?

Try locking the radios one step lower and using Turbo mode if not already. Bizarrely, we found 2x24Mbps works better than 54Mbps (non-turbo) on a given link conditions.

During a one-way bandwidth test, on a "good" link ccq can read around 95-105 range in the test direction, and about 6 in the return direction.
John Tully (I think) gave an explanation before on the forum.

2.9.1 and 2.9.2 are good software.

Regards

Posted: Fri Sep 16, 2005 4:00 pm
by budi_a
Sptehen,

Could you sharing some trick perhaps,
how to improve troughput ?? In MT 2.8 ???


Beside :
1. Lock radio rate to one lower than the max
2. Disable connection tracking
3. Enable fast frames

I'm couldn't get the rate that i'm hoping to get..

What version of MT (2.8.x) that have the best
Atheros (AR5213) driver performance ???

In lab test, the max i could get from :

- Data Rate 48 Mbps (band = 5 GHZ)
- CPU Pentium 3 866 Mhz (Bus 133)

Troughput :
- Bandwidth test : (UDP)
25 s/d 35 Mbps

- Bandwidth test : (TCP)
20 s/d 22 Mbps

- Data Rate 48 Mbps (band = 5 GHZ-turbo)
- CPU Pentium 3 866 Mhz (Bus 133)

Troughput :
- Bandwidth test : (UDP)
35 s/d 40 Mbps

- Bandwidth test : (TCP)
25 s/d 33 Mbps

Stephen could you sharing some trick / configuration,
what's we must set in MT in order to get more
bandwidth / troughput ?

We confused, because no matter what we do,
we couldn't get even nearing troughput that you
or other forum member say they could get.
Eg : UDP (60-65 Mbps)
TCP (45-55 Mbps)

Thank you very much...

Regards,
Budi

Posted: Fri Sep 16, 2005 4:30 pm
by stephenpatrick
Hi Budi,

(Very kind, but don't consider me to be the "bandwidth king" - others have reported slightly faster using routed, not bridged ;-).

But I'm very happy to share some config ideas.
Give me a short while and I'll post a PDF with some winbox screenshots.

Regards

Posted: Thu Sep 29, 2005 5:30 pm
by budi_a
Dear Stephen,

If i may ask,
Do you have any news perhaps,
regarding sharing some config ideas ? :)

Thx a lot
//Budi

Posted: Thu Sep 29, 2005 5:48 pm
by stephenpatrick
Hi Budi,

We're going to put together a document for our customers explaining step-by-step how to set up the routers for bridged mode with WDS, with some screenshots. It'll be posted on our site, and I'll put a link here.
Sorry this is taking a while, we are really busy on some major projects.

Regards

Posted: Fri Oct 14, 2005 9:10 pm
by aviper
Okey, I've recieved a pair of 400mW beasts, and one pair of CM-9 "kitties". This weekend I'll try to get some more from this link with those 2 cards. A friend of mine told me that Senao's I've bought are very good for long distance.

Posted: Sat Oct 15, 2005 12:25 am
by jonbrewer
Bizarrely, we found 2x24Mbps works better than 54Mbps (non-turbo) on a given link conditions.
That's because with Turbo on lower speed you're using a more robust modulation (16QAM instead of 64QAM) and spreading out your data over 40MHz instead of 20MHz. You have twice the subcarriers and less complex data to get over the air.

Posted: Sat Oct 15, 2005 10:14 am
by aviper
Bizarrely, we found 2x24Mbps works better than 54Mbps (non-turbo) on a given link conditions.
That's because with Turbo on lower speed you're using a more robust modulation (16QAM instead of 64QAM) and spreading out your data over 40MHz instead of 20MHz. You have twice the subcarriers and less complex data to get over the air.
Doesn't that mean near 4xCPU time and heat from the adapters?

Edit: In theory I mean ...

Posted: Sat Oct 15, 2005 11:44 pm
by blt
Hello

The senao 5g card not good. change to other 5213 (sr5 or other) and haven't problem.

Posted: Sun Oct 16, 2005 1:18 pm
by stephenpatrick
Bizarrely, we found 2x24Mbps works better than 54Mbps (non-turbo) on a given link conditions.
That's because with Turbo on lower speed you're using a more robust modulation (16QAM instead of 64QAM) and spreading out your data over 40MHz instead of 20MHz. You have twice the subcarriers and less complex data to get over the air.
Doesn't that mean near 4xCPU time and heat from the adapters?

Edit: In theory I mean ...
Turbo mode does use more CPU power than non-turbo for a given throughput, but not as much as you would think.
A converse argument to the above, if you have a CPU-limited system, you may get better throughput at good signal levels by using non-turbo and higher order modulation.

Regards

Posted: Fri Oct 21, 2005 10:25 pm
by aviper
Another fast question. Is WDS faster than "bridge" mode ? Which license level does WDS require?

P.S. As I understand from other posts, routing is faster than EoIP ?

Posted: Thu Nov 10, 2005 2:40 pm
by aviper
Okey, I've recieved a pair of 400mW beasts, and one pair of CM-9 "kitties". This weekend I'll try to get some more from this link with those 2 cards. A friend of mine told me that Senao's I've bought are very good for long distance.
And again problems.

With SR5 I get 75 % of 48 mbit turbo mode.
But when I try to put CM9 and Senaos on a 3 km link I got -50dBm signal and 2% of 9 mbit tx-/rx-CCQ ...
I've switch to 2.9.7. After I saw such a little ccq, i made another couple of PCs and switched back to 2.8.x, but the ccq still stays low. I've change the cables with new one. Feedhorns with improved one. But the problems still stays on ... Where should I look for the answer, have you ever saw such a good signal and such a low ccqS?

Posted: Thu Nov 24, 2005 7:03 am
by hitek146
I have the same problem, with 6 towers... All 6 towers worked just fine in the past, with clients showing 15-20 CCQ before, but now, some clients on all six towers show CCQ in the 0-1 range, which doesn't work at all. I don't know if it's because I upgraded to newer 2.9.x versions of RouterOS or if there is just a whole lot more noise now, but these links were working fine before on multiple towers, and now I have nothing, no matter which frequency I choose. Now, I have some clients that still have -59 signal level in both directions, but CCQ in the zeros.... I also find it odd that there have been multiple questions here on the forum(found by searching) asking for more specifics about what the CCQ means and how it is calculated, but none of them were ever answered by MT...

Can you MT guys please fill us in on this info???????

Hitek

Edit: Intel P4 Towers running MT with RB500 clients....

Posted: Thu Nov 24, 2005 11:34 am
by uldis
the CCQ will be low if there is no data going through.

From the manual:
rx-ccq (read-only: integer: 0..100) - Client Connection Quality - a value in percent that shows how effective the receive bandwidth is used regarding the theoretically maximum available bandwidth
Try to transmit and receive the data and the monitor these values.

When you upgraded to the v2.9.x was it working good (with good ccq)?
Please send the support output files to support@mikrotik.com so we could check what could be the problem.

Posted: Thu Nov 24, 2005 1:20 pm
by aviper
Yes, the main problem was that I thought that CCQ is theoretical maximum TX and RX. After few tests I understand that this is "real transfer". Soo after all tx-ccq is tx-connection in mbit/s divide by real amont of data passing throw this wireless interface.

@hitek146:
For example:
When a client connect with 48 mbits and this client transfers 24 mbits ccq is 50 % ... If I'm wrong plz correct me.

Soo the question to MT is, when we will have noise level measure ? Soo we can understand why we can't get 100 % ccq ! :)

Posted: Fri Nov 25, 2005 6:45 pm
by hitek146
I've long known that the CCQ will be low when no data is being transmitted, but these problems keep the CCQ at 0 or 1, no matter what I am trying to do, and no matter how much data I am trying to transfer. If I try to transfer data on one of these problem links, it fails, and sometimes the link even disconnects until I stop trying to transfer data....

Posted: Sat Nov 26, 2005 12:59 am
by aviper
I've long known that the CCQ will be low when no data is being transmitted, but these problems keep the CCQ at 0 or 1, no matter what I am trying to do, and no matter how much data I am trying to transfer. If I try to transfer data on one of these problem links, it fails, and sometimes the link even disconnects until I stop trying to transfer data....
There is Log, read the cause of "death" :)

Posted: Sat Nov 26, 2005 1:48 am
by hitek146
Ahh... The log reads:
2.4GHz AP1 Phys: disconnected xx:xx:xx:xx:xx:xx, your ACK is set wrong, you moron
Just joking... :) I read MT logs every day, so I had obviously already looked at the logs(and ACK is also not the problem). :D All the logs really say is of no help:
2.4GHz AP1 Phys: disconnected xx:xx:xx:xx:xx:xx, extensive data loss
and the RB500 client reports:
Wireless 1 Phys: lost connection to xx:xx:xx:xx:xx:xx, extensive data loss
I've got full wireless logging turned on, and both wireless clients and tower all report their errors to an MT logging deamon so I don't miss anything, and as you can see, the logs give me no indication of the reason for the problem.