We started to use MIMO technologies, based on Mikrotik and R52N cards, but I found strange thing:

When there is just 1 card inside Mikrotik device (RB800+R52N), MIMO works correctly. Approximatelly 130Mbps (3km distance). But when I add more wireless cards inside this devices, MIMO has big troubles to work, approximatelly 10-20Mbps and so high ping latency.
I tested it on more devices in different areas on different antenas. Any settings (Data rates, HT, HT MCS, Nstreme, and so on) has minimal influence. I used the newest Router OS 4.9 .
I afraid, there is bug in mikrotik routerOS, or in R52N cards.

what about interference? make sure you use professional isolated pigtails, separate the antennas by proper distance, make sure you have an isolated outdoor box (metallic), see if you can't somehow isolate the cards between themselves (metallic sheet wrapped in plastic or something)?

what about interference? make sure you use professional isolated pigtails, separate the antennas by proper distance, make sure you have an isolated outdoor box (metallic), see if you can't somehow isolate the cards between themselves (metallic sheet wrapped in plastic or something)?

Actually we have in our network 9 lines, where we have RB800, RB600, RB433AH and R52N cards. Just 1 line works correctly - there is RB600 and just 1 wireless card inside. The rest 8 lines works very strange. We didn't know the reason of the problems, thats why we tested different antenas, different pigtails, different routerboards, but same on any hardware.
Anyway, in some of those lines there was nstreme dual before we implemented MIMO and it worked correctly before.

- Yes, we used professional pigtails, as short as possible (1m, 3m max).
- Because we didn't know, where is the problem, we tested a lot of antenas. Separate antenas, dual antenas, offsets ... at least 5 types of different antenas in different areas. Some of antenas was really very good professional antenas - I dont thing, there will be problem.
- We haven't isolated outdoor box. But when there was Nstreme dual before, everything works correctly. When there is just 1 active R52N card, everything works correctly. Just when there is more wireless cards inside, then begin problems.
- Isolate cards inside 1 routerboard? We didn't try it, I don't know how to do that. Any metalic sheet? We didn't try it!

Yesterday I did test on 1 of our lines. One side is RB800+3xR52N really good proffesional duplex antenas. 2nd side RB800+4xR52N, proffesional separated antenas (not duplex). When all cards was active (approximatelly 10Mbps data trafic on every card), we had latency around 120ms in the MIMO line. I disabled all other cards, just 1 R52N card was active in every side of the line. The line works very good, around 100-130mbps. When I enabled other cards, problems was back. I tested to change channels and change any settings of the R52N cards, but it has minimal influence.

i have same problem.
in one mikrotik rb800 (4.5) can be only one wireless card. If i put another one my link is not good !

Same problem here :

First message sent to Mikrotik support (modified to hide some private info) :

Dear support,

we encounter a real big issue on all our Mikrotik Routers (RB433 +
3×R52n or 3×R52Hn).
Typical case is :
- wlan1 and wlan2 are backhaul links interfaces (upstream and
downstream, always in 5,4GHz)
- wlan3 is a PtM interface for the wireless CPEs (2,4GHz or 5,4GHz).

The signal levels are correct on the 2 backhaul links.

If running a bw test on each link, we can reach more than 10Mbps without
any problem on each link.
If running a bw test between the 2 remote sites across the 2 backhaul
links, the bw drops down to 200kbps !!
If running a bw test on one link while the other link is trafic loaded
at 1Mbps by other trafic, the tested link bw drops also down to 200kbps.

In operational conditions, with background trafic generated by our
customers, if we disable one of backhaul the link, the remaining
backhaul link increases from few hundreds kbps to tens of Mbps.

We use layer2 switching, and R52Hn, using wds,
and a bridge. We have also tried :
- without bridge, using IP routing => same problem => not a bridge nor
WDS issue.
- with R52n => same problem => not coming from R52Hn only
- we never had this problem on CM9 nor R52 before => maybe a problem
with R52n and R52Hn
- we have tested on about 15 different motherboards (RB433), and every
time used new wireless cards. All these tests were done on different
sites where the problem was noticed (more than 10 sites) => common
problem to all the hardware, and not only to some cards, not common to
on particular site.
- we have changed the pigtails and antennas, using different
manufacturers. => no change.
- we have changed the frequencies, always watching for channels
separation (5500MHz and 5700MHz) => not an
interference problem between the channels due to misconfiguration.
Moreover, noise is stable, and low, in any cases.
- we had run many spectrum analysis using ROS, to see if interferences
where detected => nothing detected.
- we had tried : with and without nstreme, with and without 802.11n
mode, nothing changed
- we have played with a lot of parameters to see if something changed
(tx power, fixing rates, setting hw-retries to 15, ...) => no change.

At this point, we heavily suspect :
- an hardware problem on R52n and R52Hn, or a problem with RB433 to
handle these cards.
- a software bug in the driver handling these 2 new card models : IRQ
collisions, or bad IRQ handling, software collision of the wireless
frames coming from different interfaces (bad memory isolation between
cards in the driver)....

We have no more idea.. and we are at the point that we have spent
$USD3000 this week to replace hardware for nothing.

Please also note that :
[admin@HiddenName] /system resource pci> pr
# DEVICE VENDOR
NAME IRQ
0 00:14.0 Atheros Communications, Inc. unknown
device (rev: 1) 50
1 00:13.0 Atheros Communications, Inc. unknown
device (rev: 1) 49
2 00:12.0 Atheros Communications, Inc. unknown
device (rev: 1)

First reply :

Hello,

If it is possible, try to upgrade this router and the remote location to
v5.0beta2 where we have a distance setting and then check if that helps in
your situation. Could you try to lower the tx-power of the wireless cards and
then check again, maybe there are interfering with each other. How close are
the antennas located?
Have you tried the same setup with regular wireless cards?

Our 3 next replies :

Day1:
upgrading to v5.0beta2 does not help, even when playing on tx-power and
distance parameters, 802.11n or nstreme.
The antennas are at least 80cm spaced, as we have always done in the
past, with the same antennas.
After replacing the wireless cards by regular R52 card the problem seems
to be solved, so we think the problem comes from R52(H)n or from the
driver when handling multiple cards at the same time.

Moreover, our NOC just told me that there is no problem when using 2
RB433 on the same problematic location, making the trafic to flow from
one radio link to another accross the 2 routers and the wired ethernet
local link.

...

Day2 :
New test of the day :
We have tested on 2 problematic locations to use a 2,4GHz link for one
of the two backhaul links.

So on one site, we have :
- one 5,4GHz backhaul link
- one 2,4GHz backhaul link (instead of another 5,4GHz backhaul link).

In this configuration, we have no more problems on the 2 locations where
the problem occurs, so we think the think the problem occurs when
multitple cards are used in 802.11a mode at the same time.

...

Day3 :
we have today replaced cards from R52n to R52 on another site. This
replacement has solved the problem.

The official reply from Mikrotik today is :

Hello,

Try to use a ferrite tape between the wireless cards, maybe that would help.

Here is where we are. Have someone tried :
- to use only the most distant mPCI slots on a RB433 to avoid interference ?
- to use ferrite tape or any other RF isolation between the wireless cards ?

I have the same problem. I am in despair. Return the device to the seller?

bump... still no help from Mikrotik support..

did you email mikrotik support? because this is a user to user forum

For what it is worth you may be able to get away with one minipci and one minipci-e active at the same time.

did you email mikrotik support? because this is a user to user forum

As you should read above, yes there is a opened ticket (Ticket#2010052666000569) since May, without any real solution, that's why I try to ask the forum :
- to see if we are alone => It seems that we are not alone with this problem
- to see if there are better solutions coming from users => no, because Mikrotik don't want to really cope with this problem with nobody, even big customers.

For what it is worth you may be able to get away with one minipci and one minipci-e active at the same time.

Thanks, but here I need to use 3 working wireless MIMO cards on a RB433 board. That's working great with R52 cards, but not with R52n nor R52Hn cards. And according the lack of serious answers from Mikrotik, I don't have any real solution at that time, excepting throwing everything away, and use Ubiquity hardware.

but did you try the ferrite tape or not? we don't have any other similar problems in our support system
http://shop.ebay.com/i.html?_nkw=EMI+sh ... m270.l1313

We have ordered tape, and we have tried, without any change. That was only 19mm tape.
I will receive a 25mm roll on friday.. but our hopes are very low.

Mikrotik never said that will solve the problem.. but only that may helps... We don't want to be Mikrotik's beta testers anymore.. and that's what is going on now.

We have here :
- 200 assembled RB433 board with 3 R52Hn cards on each board
- all the wireless cards have dummy 50ohms loads on unused radio port
- no certitude from Mikrotik that will work trying another shielding tape

MimiFleX: did the 25mm tape solved Your problem?
This tape probably isolates the RF, but keep the warm inside, doesn't it? It will not cause the card to overheat?

Tape does not solve this issue.

hmmm
that means we have to wait for the new board (what will be published soon) what have 3+2 minipci slot, where we can safely use 2 (maybe 3) slots without interference?

any other options?

Probably a sync would sove this kind of issue, so there will be no interference when all card transmits and receives the same time?
(i'm sure it would not need any external sync device...)

I've been reading this ´old´ thread and still have some important remarks to make that seems not to have been discussed:

What is the channel separation between the radio cards in one box? I see two cards mentioned to work in 5,4Ghz and one as AP in either 2,4 or 5,4Ghz.

The ´normal´ 5Ghz band is 200Mhz wide (5500 to 5700Mhz) and if you use 20Mhz channel you can create 11 channels each 20Mhz apart. If channels inside one box are choosen smart you can keep at least 20, but better 40 or even better 60Mhz distance between radio's.
If the 3rd card is also in same 5Ghz band than you have to spread the 3 radio to the best over the band.

If you only using 5,4ghz (which band is that? In europe it is forbidden for wlan) than you have only one freq. channel to spare. Using it for more than one radio is asking for problems.

But, another factor that is not discussed: When using ´n´ with upper or lower channel the actuall radio frequency bandwith the radio uses is increased to 40Mhz! You can understand that now suddenly the 5ghz band only has room for 6 free channels! In the 2,4Ghz you use the whole band spectrum!
To keep these away with sufficiant spectrum inbetween from other radios becomes a task now if many AP's are used in same spectrum. Let alone if other ´boxes´ in the area are to use their own radio channels while they still can ´see´ eachother....

Now, on top of all, if also TDMA is used (nv2) things even become worse. Out of experiance I found that adjacent channels best would be kept 40Mhz away from eachother even on distant but reachable AP's.
(It seems that stations leaving an AP in one working channel trigger other AP's in adjacent channels to shut down some units as well at times. Although it now looks ros5.1 is doing much better in that respect.)

If the same nv2 would now have to be applied to 40Mhz wide ´n´ links it becomes almost inevitable to have interference problems.

I live in a valley where basically all 30 AP's (AP's plus backhaul links, mine and others) using the 5Ghz band. They all tend to interfere with eachother. Only one year ago I was sort of the only with half the amount of radio units in this band and all 802.11a. In these days I hardly had problems.....
Since others started to use Airmax (UBNT, tdma + ´n´!) and I nv2 and some ´n´ backhauls things became a bit hectic.
Random disconnects all over the place up to complete crashing units that have been running for months or even 2 years! (After 3 months of struggling I sort of found that UBNT tdma protocol interferes more with legacy ubnt units (I have them in my network) while MT's nv2 is more problamatic for other ROS running devices.

To battle all this I am now removing all ubnt devices because it seems I can't bring them under control (since I have no control over ´other's´ networks) and for AP's I am now putting cards in separate metal (or spray painted plastic) box and started to use 10Mhz channels on my AP's.
I am in the process of converting all my AP's into nv2 running as well and ´harden´ all links by setting the frequency in the stations ´scan´ list, make sure both mac address and SSID are mentioned of the AP in the stations ´connect to´ list (separate rules), use very limited data and basic connection rates and setup security on the nv2 links or the man. protection on the 802.11a links.

This brougt my network back into stability.(I must state here that also the use of ros5.1 brought lots of improvement to my network. Obviously the drivers must have been improved to better withstand interference issues.)

If you use ´n´ protocol link with nv2 on 2,4ghz you basically consume the whole band for that unit only. If you use it on AP's use only channel 1 and 11 (or 12, 13 or 14 depending on your region) and shield everything in your transmission/recieve line as good as possible from other radio signals. Like said, use good materials, sector antenna instead of omni, one card-one box etc.


Regarding the isolation of each radio card with metal paper wrapping I have to mention the possible risk that comes with it:
1. Shortcuts! Be very carefull with the placing. It should not contact any conductive parts of both the board or the radio cards! (Don't even think of fitting it while board is powered up!)
2. Heat dissolvement. Wrapping radio's in prevents the heat to dissipate. This is also a very good argument not to use too many cards in one box anyway. If you have one or two HP cards running in a box that is in the full sunlight all day and you don't have sufficiant ventilation expect the radios to drop in the heat of the day and it shortens the live time anyway considerably. Let alone if you now give the radio's an extra warm ´coat´!

So, the magic word is "separation"!
Separation of radio signals. (good earthed and shielde coax. Also the little pigtails that connect to the card!)
Separation of radio cards
Separation of complete units
Separation of antenna's
Separation of working channels.

Somewhere on this forum a guy did some test on two cards in one box and you'll be amazed how much interference and how far it carries (in the band) you'll find! !0Mhz for the AP already reduces the ´footprint´ of that radio, even IN the box!

What is the channel separation between the radio cards in one box? I see two cards mentioned to work in 5,4Ghz and one as AP in either 2,4 or 5,4Ghz.

The ´normal´ 5Ghz band is 200Mhz wide (5500 to 5700Mhz) and if you use 20Mhz channel you can create 11 channels each 20Mhz apart. If channels inside one box are choosen smart you can keep at least 20, but better 40 or even better 60Mhz distance between radio's.
If the 3rd card is also in same 5Ghz band than you have to spread the 3 radio to the best over the band.

Of Course, all the cards have at least a 40 or 60MHz gap between 5GHz channels.
However, this problem does not occur when one card is running at 2,4GHz and the other one at 5,4GHz.

But, another factor that is not discussed: When using ´n´ with upper or lower channel the actuall radio frequency bandwith the radio uses is increased to 40Mhz! You can understand that now suddenly the 5ghz band only has room for 6 free channels! In the 2,4Ghz you use the whole band spectrum!
To keep these away with sufficiant spectrum inbetween from other radios becomes a task now if many AP's are used in same spectrum. Let alone if other ´boxes´ in the area are to use their own radio channels while they still can ´see´ eachother....

This problem occurs even when not using N mode.

Regarding the isolation of each radio card with metal paper wrapping I have to mention the possible risk that comes with it:
1. Shortcuts! Be very carefull with the placing. It should not contact any conductive parts of both the board or the radio cards! (Don't even think of fitting it while board is powered up!)
2. Heat dissolvement. Wrapping radio's in prevents the heat to dissipate. This is also a very good argument not to use too many cards in one box anyway. If you have one or two HP cards running in a box that is in the full sunlight all day and you don't have sufficiant ventilation expect the radios to drop in the heat of the day and it shortens the live time anyway considerably. Let alone if you now give the radio's an extra warm ´coat´!

Yes, that was VERY difficult to place this tape, shortcurts have been avoided adding paper.
Moreover, when installing 3 R52Hn on a RB433 Board, there is NO space between the cards.
Even if shielding tape was a solution (this is not), it would be impossible to place on hundreds of devices without shorcuts and heat issues... and without a huge amount of time.

So, the magic word is "separation"!
Separation of radio signals. (good earthed and shielde coax. Also the little pigtails that connect to the card!)
Separation of radio cards
Separation of complete units
Separation of antenna's
Separation of working channels.

Using one enclosure per card solves this issue.
However, this implies many cables on towers, bigger switch units which increases the global price.

But, another factor that is not discussed: When using ´n´ with upper or lower channel the actuall radio frequency bandwith the radio uses is increased to 40Mhz! You can understand that now suddenly the 5ghz band only has room for 6 free channels! In the 2,4Ghz you use the whole band spectrum!
To keep these away with sufficiant spectrum inbetween from other radios becomes a task now if many AP's are used in same spectrum. Let alone if other ´boxes´ in the area are to use their own radio channels while they still can ´see´ eachother....

This problem occurs even when not using N mode.

How is that? Normal channels are only 20Mhz or less wide. Unless you use turbo mode. But in which ITU regions is this actually allowed? I mentioned this because I think many readers of this forum are not really aware of the fact that by using ´n´ mode with upper or lower channel you actually ´occupy´ an adjacent channel. In designing your network you have to be aware of that.

So, the magic word is "separation"!
Separation of radio signals. (good earthed and shielde coax. Also the little pigtails that connect to the card!)
Separation of radio cards
Separation of complete units
Separation of antenna's
Separation of working channels.

Using one enclosure per card solves this issue.
However, this implies many cables on towers, bigger switch units which increases the global price.

Yep, and here we learn that the more professional our network has to be the higher the costs.

We are experiencing exactly the same problem - RB433 with two R52Hn in a bridge with ether1 interface. We are getting packet loss on the ethernet side and client disconnections when both wlan interfaces are running. Turn off 1 wireless interface and the problem disappears.

We are using 802.11a, not N - with channels over 200MHz apart. Pigtails are commercial MMCX, coax is LMR400 and both antennas have different polarization.

We have plenty of similar setups working fine but not with R52Hn, so I am suspecting these cards of being noisy. In fact I can even hear a high pitched sound from the board when I open the (metal) enclosure.