Last week, at inet-meeting Cyberbajt presented his new Mimo 3x3 products.
Pretty impressive results, I think. With new Atheros AR93x0 chipsets, MikroTik becomes fastest WiFi solution, in this price segment.
Nice alternative for high throughput p2p links.

Does anyone had any experience with such equipment?

Do you have any more info on this. Looks very interesting.
What was the link distance, signal levels, ccq, antennas used, etc.?

Do you have any more info on this. Looks very interesting.
What was the link distance, signal levels, ccq, antennas used, etc.?

Hi everyone
Above test was made with Cyberbajt 19dB Mimo 3x3 antenas at 500m link. Signals about -60 for all 3 chains. CCQ about 90%
Today we move this link to 4km, but signals are to low. Max transfer is 60Mb with signals about -77. Maybe if we change cards for more powerfull models, link will get better.
I'll write here more information tommorow.

you'd better to change antennas!

what model type of card you use ? manufacturer, card... i´ve done some tests with UBNT SR71, but without better results than 2x2

what model type of card you use ? manufacturer, card... i´ve done some tests with UBNT SR71, but without better results than 2x2

SR71 is only 2 chain chipset card. Check AR9160 specification.
My cards are AR9380 from Sparklan. It's new generation chip, with 3-chain support

you'd better to change antennas!

Antennas are quite good, but only 19dB. Everythings works fine for shorts distances. Now I'm looking for high gain dish antennas with 3x3 support. I think it's to new technology, because I can't find anything more than 2x2.

I have a few problems with MT software and AR9390 chipset support:
- NV2 protocol not working If I change protocol to NV2 - there is no link. I've tried 1-3 chains combination, no positive results
- MCS negotiation works veeery sloow. Best results I've got with manual MCS config.

I have a few problems with MT software and AR9390 chipset support:
- NV2 protocol not working If I change protocol to NV2 - there is no link. I've tried 1-3 chains combination, no positive results
- MCS negotiation works veeery sloow. Best results I've got with manual MCS config.

Next problem is with tx-power. When i change power values, I see no changes in "Current TX Power" tab.

currently there is a limitation for the tx-power and the only way how to change the power is by using tx-power-mode=all-rates-fixed. Please be careful and not to use higher tx-power than the highest data-rate has as that might damage the card. Use this setting if you want to do some in lab testing where you need low tx-power values.
For better MCS selection, use the rate-selection=advanced

currently there is a limitation for the tx-power and the only way how to change the power is by using tx-power-mode=all-rates-fixed. Please be careful and not to use higher tx-power than the highest data-rate has as that might damage the card. Use this setting if you want to do some in lab testing where you need low tx-power values.
For better MCS selection, use the rate-selection=advanced

Thx for reply
All-rate-fixed works as You says. MCS selection in "advenced" works really faster.

Hi all
After few tests - I can write some conclusion. Cards are awsome With good antennas can work with more than 200Mb real network throughput. In bad envoironment, where much noise is, cards work better than old MIMO 2x2. They can work with QAM16 modulation with higher MCS.

you'd better to change antennas!

Antennas are quite good, but only 19dB. Everythings works fine for shorts distances. Now I'm looking for high gain dish antennas with 3x3 support. I think it's to new technology, because I can't find anything more than 2x2.

Imho if used for distant links you can use 3 independent antenna's, just put them close together on the tower and make sure the physical distance of the path from the radio connector to the antenna is exactly the same (use same coax cable lengths and same pigtails)
For remote unit it looks like one big 3 chain antenna.
I even remember to have red somewhere that 3 slightly separated antenna's might even perform better than all in one unit. Special since at least two of the three have to work on same polarization.

I haven't find any docs so far that actually decides on best practise on polarization setup for 2x2 or 3x3 (2x3 or even 3x4 seems to be a completely unexploited area yet, apart maybe from some proprietary solution like Ruckus uses?)
So I presume you have to play to find best results. And these are than not conclusive since the environment also has a role...

you'd better to change antennas!

Antennas are quite good, but only 19dB. Everythings works fine for shorts distances. Now I'm looking for high gain dish antennas with 3x3 support. I think it's to new technology, because I can't find anything more than 2x2.

Imho if used for distant links you can use 3 independent antenna's, just put them close together on the tower and make sure the physical distance of the path from the radio connector to the antenna is exactly the same (use same coax cable lengths and same pigtails)
For remote unit it looks like one big 3 chain antenna.
I even remember to have red somewhere that 3 slightly separated antenna's might even perform better than all in one unit. Special since at least two of the three have to work on same polarization.

I haven't find any docs so far that actually decides on best practise on polarization setup for 2x2 or 3x3 (2x3 or even 3x4 seems to be a completely unexploited area yet, apart maybe from some proprietary solution like Ruckus uses?)
So I presume you have to play to find best results. And these are than not conclusive since the environment also has a role...

Hi, thanks for advise. I will try it.
I've just tested new antennas, and I see big improvements. New project is also 19dB, but diferent design. Test link at 850m can work with 150/30Mb (43k pps/12kpps) real network troughput. (TX-Power is set for 3 for both sides, NV2 is enabled, signals about -55) Ping is about 3..4 ms. I think MT software need some improvements, because link default sets to MIMO 2x2 MCS, and I have to manualy set only 3 chains MCS. There is a problem with QAM64 MCS for 3 chain's, and now I don't know is it problem with software or with antennas... Maybe new nv2 package will work better, but I can't try it, because MT developers didn't respond for my mail.
On the same link I have NB22 and max througput in test is only about 60/20Mb (with terrible pings). I try to check Rocket on 3x3 antennas, I'm curious how it can work in the same environment.

I found a formula somewhere that discusses recommended distances for antenna in MIMO systems, can't remember or find it again.

I am seriously considering replacing a 433 w/3 R52H's running 3 sectors with one of these and see how it works.

Almost all of my customers can connect to at least two sectors, might just take care of a lot of problems I have there.

I found a formula somewhere that discusses recommended distances for antenna in MIMO systems, can't remember or find it again.

I am seriously considering replacing a 433 w/3 R52H's running 3 sectors with one of these and see how it works.

Almost all of my customers can connect to at least two sectors, might just take care of a lot of problems I have there.

Looks to my you are a bit confused here...

You can not replace a 3 sector AP with 3 radio's/SSID's by one 3 chain 802.11n unit unless you suddenly want to start using a omni antenna (which will degrease signal levels some 30% on all units... depending on type etc.)

Or you want to replace the 3 present sector/R52H setup by one 3 chain mimo card working in 802.11a mode and connect each of the chains to one of the sector antenna's.
I think this is what you think off, but it is not mimo. (802.11n) It us just plain 802.11 with 3 fully synchronised sector antennas working in same frequency and with same SSID.

I already do it with 2 2 chain units and recently received a 3 chain card (ubnt) for testing purposes... if it works you'll read about it..

Can't remember the guy's name on here, smart fellow tho, he did a PTP link with an 802.11n card in the middle, one chain facing N, one chain facing S, worked fine for him.

I know beam forming would not work with sectors, but I need to redo that AP anyway and I want to try this. Considering every client on the AP can hit at least two sectors, it seems that the signal reconstruction of 802.11n would come into play.

I am not looking for 450Mbps bandwidth, just thinking of a decent test that might pay off before I have to completely change up a tower's configuration. Single chain performance would be enough for this AP for the foreseeable future.

edit: for clarity, yes, just 3 sectors synchronized on a single 3x3 card.

Can't remember the guy's name on here, smart fellow tho, he did a PTP link with an 802.11n card in the middle, one chain facing N, one chain facing S, worked fine for him.

I know beam forming would not work with sectors, but I need to redo that AP anyway and I want to try this. Considering every client on the AP can hit at least two sectors, it seems that the signal reconstruction of 802.11n would come into play.

I am not looking for 450Mbps bandwidth, just thinking of a decent test that might pay off before I have to completely change up a tower's configuration. Single chain performance would be enough for this AP for the foreseeable future.

edit: for clarity, yes, just 3 sectors synchronized on a single 3x3 card.

This could work. You can set HT-chains for 3x3 on AP and on Clients side set only 3x1, 2x1 or simply 1x1 chain. I try to test this if only I find some free time.

Interesting will be possible 3x3 Mimo antennas with gain for about 26-34dbi. Becouse 16dbi max 2-3km

Interesting will be possible 3x3 Mimo antennas with gain for about 26-34dbi. Becouse 16dbi max 2-3km

It's only matter of time, when such antennas will be in wifi shops.

so this test is on RB800 with sparklan WPEA-127NI and 3x3 cyberbajt?

so this test is on RB800 with sparklan WPEA-127NI and 3x3 cyberbajt?

Yes.
I have tested MT 5.9 with this link. Times on link are now lower. With throughput ~150/30Mb - time on link is about ~2-4ms.

Fine thing is MCS20. It's 270/270 QAM16 (with 40Mhz). Now it's really easy to make such links, because it can work in noisy enviroment.

Easy, but not cheap... RB800s are very expensive. I asked mikrotik a year ago if are they planning some kind of RB411 with miniPCI express for wireless cards, but nothing happened.
I´ve tested 2x2 mimo on miniPCIexpress cards and it worked much better, then on r52n cards..

I havenot understand, 150/30 is real transfer on 3x3? not too good..
running at the moment 433ah+r52hn 270/270, 28dbi dishes 2x2mimo@5.8ghz~ 10,5mb/s comes on(100mbps ethernet ports)
Thinking to upgrade that link.. to 3x3 using 3x23dbi panel antennas both sides

2 links with SXT, some of the traffic is generated with btest. Ping is very low, speed is good. I hope that 3x3 will be at least like this.

2 links with SXT, some of the traffic is generated with btest. Ping is very low, speed is good. I hope that 3x3 will be at least like this.

Can you test your link with iperf and tcp protocol?

no.

C'mon! Give me some houpe! =)

I gave up Mikrotik Wireless just because when i did UDP test, the link could handle 100Mbit, but when i did TCP test, the link was cca. 30-40mbit and very unstable. Maybe i did something wrong, maybe i dont have enough knowledge. But still i would like to know how do you guys do it.

I did test with 4 Rb's(RB750,Rb450G,RB433,Groove) in TCP mode, ping is still good. A pc at the other end would help...

TCP test probably takes too much CPU. You should test it by running iperf on PC's on both ends of the link.

A pc at the other end would help...

doh

How does MIMO 3X3 handle co-location interference on a mast with multiple sectors + ptp's running both horz + Vert .

I did test with 4 Rb's(RB750,Rb450G,RB433,Groove) in TCP mode, ping is still good. A pc at the other end would help...

I think You will be satisfied with 3x3. Hera are some outdoor tests:

http://www.trzepak.pl/download/file.php?id=860

https://isp4isp.pl//storage/serve/s66l6 ... 02.6.2.JPG

More information will be presented during MUM.

Was there more information presented at the mum? Can you post it here?

Hi all
After few tests - I can write some conclusion. Cards are awsome With good antennas can work with more than 200Mb real network throughput. In bad envoironment, where much noise is, cards work better than old MIMO 2x2. They can work with QAM16 modulation with higher MCS.

just ordered mine cuz i love toys!!

okay so i got micro atx atom board with
2.2ghz
2gig ram
3x3 mimo mini pci-e

I tried with hyperthreading =4
dual = 2
and single = 1


one 1 core using btest to the router, i looked at resources>cpu and it maxed out
dualcore it seems to use mostly 1 core for the btest alone

on average i could see 260mb/s btest to router
with a freak of nature peaking of 305mb/s for a short period hard to reproduce

I didnt really notice any difference going from dualcore to hypertreading
(only that hypertreading resouces>cpu showed 1 core maxing between 25-40% i couldnt really make sense of this)

i was honestly hoping to see a clean 300mb/s + easy

most of the time i saw
UDP 200mb/s that would start to average around 240-260mb/s
TCP 70ish mb/s and wouldnt really go higher than that

When i did TX and RX one direction would only peak about 60mb/s and the other way would go down a bit

I have data rates advanced all mcs enabled and all the box's ticked in HT


i dont have the router next to me to show pictures but ive been testing it all day

I changed the t to all card rates fixed 1-5dBm because it was extreemly close i have 3 x dualband 5db antenna's little swivels.
I saw signal levels of around 51-60 and SNR of around 60
and the ccq would change alot i dont understand that, the best data rates i got were when the ccq was 100%

Anyone suggest why i might not be getting better throughput

i did a bit of reading and learned that different numbers in the MCS act differently some are 2 spatial beams and i assume the higher ones are 3 spatial beams, is this correct?
im going to try this next to change to the high MCS only


any feedback?

Uncheck anything except MCS20-23. Check at what speed your radios are connected durning test, and write it here.

Hello,
i have the following problem:
http://forum.mikrotik.com/viewtopic.php ... 99#p374686
MIMO 3x3 works only as MIMO 2x2
Has anybody good advice?