RB751U Wireless Performance

RouterOS Wireless Networking
1

I have been using some RB751Us for a while now and have some observations about some surprisingly poor wireless performance and I’d like other’s views on the topic.

My test environment:
btest server running on a desktop with an Intel i3 CPU running at 3 GHz with Windows 7 x64 always wired with a 100 Mbit connection.
btest client running on a laptop with an Intel i5-2520m CPU running at 2.5GHz with Windows 7 x64
Client wireless adapter (for some of the tests (see below)) is an Intel 6205.

Baselines:
Using an 100 Mbit Ethernet connection to the laptop:
btest results: 46 Mbps both directions, 79 Mbps rcv, 80 Mbps send

If I use the laptop’s Intel wireless: card to connect to a Cisco Aironet Series 1200 AP (G only)
my btest results are: 13Mbps both directions, 22 Mbps rcv, 23 Mpbs send (wireless speeds show 56 Mbit)

If I connect RB751U to the ethernet port of the laptop, configure the RB751U as a bridged-client and test again to the Cisco AP
my btest results are: 6 Mbps both directions, 12 Mbps rcv, 11 Mbps send. (wireless speeds show 18/24 Mbit)

Note that this is about 1/2 of the performance of the Intel card. Why?

Further testing:
I use the laptop’s Intel wireless card to connect to a RB751U running as an AP with two chains using the N protocol with a 20 MHz channel width.
my btest results are: 17 Mbps both directions, 18 Mbps rcv, 33 Mbps send (wireless speeds show 120Mbit/39 Mbit (Tx/Rx))
This seems disappointing so I decide to run the following test:

Laptop Intel wireless card connected to a RB751U running as an AP with two change using the N protocol with a 40 MHz channel width. The wireless speeds negotiate to 240Mbit Tx/39 Mbit Rx.
btest results are: 19 Mbit both directions, 18 Mbit rcs, 39 Mbit send
This is also disappointing and not really much better than the 20 Mhz channel width.

So I try to run two final tests with a RB751U as an AP and another RB751U as the client (connected to my laptop with ethernet).
With a 20 MHz channel width
btest shows 7 Mbps both directions, 12 Mbps rcv, 17 Mbps send

With a 40 MHz channel width
btst shows 11 Mpbs both directions, 17 Mbps rcv, 21 Mbps send

Why are all these results so disappointing? Does anyone have better results to report using the RB751U?

2

Did you write down any signal strength’s or link-quality during your testing?

3

Yes. I did notice that the signal strength were always roughly in the -30 to -40 range (I was not very far from the the AP). Interestingly even when I was in the basement testing and a scan show no other activity on the entire band the CCQ was listed only as about 25% to 30% in either direction. This also happened when I tested at a different location with a different RB751U.

4

That signal strength is WAY too strong. It needs to be around -60, -70, -80.

For testing purposes:
Try to get the signal strength in that range around -70, and keep it line of sight to keep the CCQ as high as possible 90%+, and see what your tests show. Note that the CCQ is only accurate when traffic is flowing. When there is no traffic, the CCQ goes way down.

Let us know your new test results, with strength and CCQ numbers.

5

OK. I’ll test again with a lower signal strength further from the AP and report back. I was reporting CCQ and strengths when the test was active, not inactive, but I don’t think that I can get the signal strength that low and also keep in line of sight. These APs are indoors.

6

There is something completely wrong with RB751U wireless. Any other wireless router (even the stupid ones) is performing at least twice better than MT. Hopefully next firmware (5.12 ?) will fix the issues.

7

Here are some more results when the client is further from the AP.

Using the Intel card in the laptop, the Tx/Rx Rates showing up on the RB751U AP were hanging around 36.0Mbps/39.0Mbps, the signal strength was about -77 at the AP. The CCQ at the AP was reported as about 40 to 45%. The peak throughput using btest was about 10 Mbits/sec when running in both directions.

Using a RB751U as a client (connected to the laptop using an ethernet and wirelessly connected to the same RB751U AP) the Tx/Rx rates at the AP are generally 18Mbps/18Mbps. The signal strentth was -65Tx/-64Rx. The CCQ was rougly 25Tx/30Rx. The btest throughput was abou 6 Mbit/sec (running in both directions). Running in one directions I got about 11 Mbit for send or rcv.

The frequency usage tool in the AP showed no traffic on channels 1-6 and less than 3% traffic on the higher channels.

FYI, the CCQ was often higher when there was no traffic.

8

make sure that you are using rate-selection=advanced
What tool are you using to test the speed and what protocol?

9

I changed to rate-selection=advanced
I’m using the Mikrotik btest tool

Even with rate-selection=advanced the results are essentially the same, perhaps a very minor improvement.

Perhaps related, perhaps not, I also observe that the external antenna (not used for the test results reported here) does not appear to operate properly. See http://forum.mikrotik.com/t/rb751u-external-antenna/52826/1 for more information.

10

Hi,

I see exactly the same.
Any hints already? I feel like on a 802.11G AP, which isn’t even very good performing.
Any other N AP gives me easy 60MBit throughput.

11

@ddejager: Did you hear something from MT themselves yet regarding this issue? I’m really stuck here, tried all kind of setting changes, nothing helps.
we need lots of replacement devices for all our wrt54gl driven hotspots out there.
We were so enthousiast over the RB751U-2HnD as it looked like the perfect replacement.

But now when testing, the wireless performance ain’t better then a WRT54GL, so we need another replacement device. (all in one box, preferly not a seperate AP)
Anyone a hint?

kr,
B

12

@beone:

I did have quite a few email interactions with Mikrotik support. Basically after giving them all my data and configurations and support files, and trying a few things that they suggested, nothing changed, and Mikrotik never acknowledged there is an issue. I quote below the last few interactions with Mikrotik support, FYI…

Thanks. I believe that there are two different problems with the RB751U:

  1. External antenna not working right. (Others, including my vendor, have also observed this.) I will simply not use that antenna.
  2. The overall 802.11N performance, is poor in comparison to other N routers or clients. I think that this can be observed by simply comparing the result of testing when only the client or AP is switched to a radio other than the RB751U.

Dale


-----Original Message-----
From: MikroTik support [Uldis] [mailto:> support@mikrotik.com> ]
Sent: Friday, January 13, 2012 12:33 AM
To: Dale
Subject: Re: [Ticket#2011122966000075] [Spam] Re: External Antenna port on RB751U

Hello,

One problem could be because of the too much interference received from the external antenna (noise from other wireless sources, for example, maybe could be from the 3g radios) and that could cause the RX radio to lower the signal. It depends on the antenna specification as well how wide frequency range it can receive. One option would be to add a hardware RX filter to boards radio part to filer out the noise from non 2.4gz devices, but that requires soldering and also it might not be the cause in this case.
We usually suggest to use built-in indoor antennas and use the external connector is an option feature.

Regards,
Uldis

Come to the MUM in Poland
March 15-16, 2012
http://mum.mikrotik.com/2012/PL/info

01/11/2012 17:30 - Dale wrote:

Yes, tst2 uses ant-b. It always provides much lower throughput for the btest.

Also if I test with my laptops wireless card instead of the RB751U as
the client I
get much better throughput (about 2x better, up to 17 Mbit/sec).
However all of the test results are inferior to testing with a
non-mikrotik wireless N AP with a
non-mikrotik client in the same environment.

If I lock the data rate as you suggest, I still see the same asymmetric results
when using the B antenna. One of my main observations is that when
the AP is set to ant-B, the AP Tx chain 1 signal level get worse by
about 15 dB and the throughput of the
test also is reduced. The hardware is acting like it is not switching
the radio to antenna-B for both transmit and receive. You should be
easily able to duplicate this issue.

Dale

13

LOL, Join the club.
I also mailed MT support and got reply from uldis:

Hello,

we would need a support output file from your setup when you are doing the test so
we could analyze what could be the problem.
In our test using RouterOS Bandwidth test using TCP (connection count 20) we could
get approx 94mbps throughput in one direction and 70mbps in both directions using
40mhz channel option.

Regards,
Uldis

Well, I did send supout but I’m pretty sure there’s nothing wrong with my config. (nothing heard from them since then)
What does get my attention is that they got that throughput with 20 simultaneous tcp sessions, while I was doing a simple speedtest with a single tcp session.
Could this be a buffer problem somewhere in the MT box? (or in the used wifi driver itself)
Anyway, someone at MT really has to solve this, we are in urgent need of a solution.

Kind regards,
B

14

Just adding my voice to this one. Similar situation. Max I can get through is about 27mbits tested with a single threaded run of iperf.


I am looking forward to a solution as well…

15

Well, just my 2 cents - I did try 751 and the performance solely depends on the card that is in the pc and clean spectrum.
My HP Probook with Broadcom N card works waaaaay better with old wrt54gl than new 751 and the signal strenght of 751 antennas is something to talk about as well.
Right now I have replaced that scrap-router with RB411AH + noname chinese Atheros n card with two 6db omni antennas and now I can say, that mt works better than wrt54gl, but not much.
And the shame is - I cannot replace my wifi card with atheros, because of that darn HP policy of marrying card to bios - replace wifi card and no booting for you.
My point for now is - MT ↔ MT link is perfect, MT ↔ PC link… not at all

16

I was also having poor wifi. I changed from the default Channel Width to 20MHz. WiFi is much more stable now.

17

No, it doesnt.
When I test with a deliberant APC 2Mi I pull easily 70MBit on a simple http speedtest.
This is with the same notebook, in the same spectrum and at the same distance.

18

Cross linking two threads about same issue. http://forum.mikrotik.com/t/rb751u-2hnd-wireless-issues/53674/1
Problem still persist and quick resolution would be appreciated from MikroTik.

19

Just in case somebody needs help to set up 802.11n for regular SOHO use. - Connect to RB751U/G using Laptop or Phone etc…

After few days of tweaking and reading MT forum I found out that I’m not alone with my Mikrotik wireless issues… There are lot of confusions how to effectively set up 802.11n for home use.
Standard configuration will allow you 25-30Mbps and that’s all. Regardless of wireless card you have (Intel or Atheros…) you will need to work if you need more.

  1. Centrino certified Intel cards will work only using 20Mhz channel width. - I made up to 70 Mbps (TCP http, cifs) with such cards - which looks much better than 30.
  2. Atheros cards will use full potentials (40Mhz channel width). Due to my noisy environment I couldn’t get much more then 85Mbps (TCP http, cifs) from this card, but it’s probably possible to get better results.

My configuration:

/interface wireless
set 0 adaptive-noise-immunity=none allow-sharedkey=no antenna-gain=3 antenna-mode=ant-a area="" arp=enabled band=2ghz-b/g/n basic-rates-a/g=6Mbps basic-rates-b=1Mbps bridge-mode=\
    enabled channel-width=20/40mhz-ht-above compression=no country=RealCountry default-ap-tx-limit=0 default-authentication=yes default-client-tx-limit=0 default-forwarding=yes dfs-mode=\
    none disable-running-check=no disabled=no disconnect-timeout=3s distance=indoors frame-lifetime=0 frequency=2452 frequency-mode=regulatory-domain frequency-offset=0 hide-ssid=no \
    ht-ampdu-priorities=0 ht-amsdu-limit=256 ht-amsdu-threshold=256 ht-basic-mcs=mcs-4,mcs-5,mcs-6,mcs-7 ht-guard-interval=any ht-rxchains=0,1 ht-supported-mcs=\
    mcs-0,mcs-1,mcs-2,mcs-3,mcs-4,mcs-5,mcs-6,mcs-7,mcs-8,mcs-9,mcs-10,mcs-11,mcs-12,mcs-13,mcs-14,mcs-15,mcs-16,mcs-17,mcs-18,mcs-19,mcs-20,mcs-21,mcs-22,mcs-23 ht-txchains=0,1 \
    hw-fragmentation-threshold=disabled hw-protection-mode=cts-to-self hw-protection-threshold=2347 hw-retries=4 l2mtu=2290 mac-address=00:0C:42:F7:B6:E1 max-station-count=7 mode=\
    ap-bridge mtu=1500 name=wlan1 noise-floor-threshold=default nv2-cell-radius=30 nv2-noise-floor-offset=default nv2-preshared-key="" nv2-qos=default nv2-queue-count=2 nv2-security=\
    disabled on-fail-retry-time=100ms periodic-calibration=default periodic-calibration-interval=60 preamble-mode=short proprietary-extensions=post-2.9.25 radio-name=000C42F7B6E1 \
    rate-selection=advanced rate-set=default scan-list=default security-profile=WPA2 ssid=MySID station-bridge-clone-mac=00:00:00:00:00:00 supported-rates-a/g=\
    6Mbps,9Mbps,12Mbps,18Mbps,24Mbps,36Mbps,48Mbps,54Mbps supported-rates-b=1Mbps,2Mbps,5.5Mbps,11Mbps tdma-period-size=2 tx-power-mode=default update-stats-interval=10s \
    wds-cost-range=50-150 wds-default-bridge=none wds-default-cost=100 wds-ignore-ssid=no wds-mode=disabled wireless-protocol=802.11 wmm-support=disabled
/interface wireless manual-tx-power-table
set wlan1 manual-tx-powers="1Mbps:5,2Mbps:10,5.5Mbps:10,11Mbps:10,6Mbps:17,9Mbps:10,12Mbps:10,18Mbps:10,24Mbps:10,36Mbps:10,48Mbps:10,54Mbps:10,HT20-0:17,HT20-1:17,HT20-2:17,HT20-3:17\
    ,HT20-4:17,HT20-5:17,HT20-6:17,HT20-7:17,HT40-0:10,HT40-1:10,HT40-2:10,HT40-3:10,HT40-4:26,HT40-5:25,HT40-6:24,HT40-7:20"
/interface wireless nstreme
set wlan1 disable-csma=no enable-nstreme=yes enable-polling=yes framer-limit=3200 framer-policy=best-fit
/interface wireless align

For some reason following settings improve throughput:
enable-nstreme=yes enable-polling=yes

I don’t know why nStream should be enabled, but it was performed better with it than without it.

hw-protection-mode=cts-to-self hw-protection-threshold=2347 hw-retries=4

and most important was: ht-amsdu-limit=256 ht-amsdu-threshold=256.

Decreacing AMSDU from 8192 increases throughput! I read oposit of this on many posts, but try it and you will see.

Good luck :slight_smile:

20

Well, should I carry AP glued to my laptop? Not the reason I bought it rather slim and good-looking.