I have a 4.4 mile, single chain 802.11A link that is working perfectly. -68 signal strength on both ends. 100/100% CCQ. Good clear line of site. Running 4.5. Pushing 10-15M across it at most times. RB411AH boards. Using MPLS / VPLS bridge. CPU hovers around 25%. 2-3ms ping times (from my office).

When I switch to to N (single chain), I get latency in the 100's to 1000+ ms. I've tried all kinds of combinations with similar results.

Nstreme on, nstreme off. No framer policy either way. Changed HW Retries to 15, disconnect timeout to 00:00:09. No go.

HT Extension channel on / off - a little better with it on (<500ms).

Guard interval long / any - no noticeable difference.

Anything else I'm missing?

[Edit: Changed the subject line to get more attention to the problem after having no replies on this post or from support@mikrotik.com ]

Well, no reply here, nor from support@mikrotik.com

Mikrotik is really dropping the ball on "N" and seems to be turning a blind eye to this problem. I would love to be disproved on this point.

In the past, for reliable, affordable backhaul links, Mikrotik was the only way to go. Now they are going to lose market share to UBNT who has a product that WORKS with MIMO/N. I'd MUCH rather use Mikrotik, but if they continue to ignore the lack of working "N" for backhauls (I won't even attempt PTMP), we've got to look elsewhere. What a shame.

Good luck with SwitchOS, sorry you got beat on wireless....

I have (3) Mikrotik backhaul links using the Mikrotik "N" Cards from Streakwave. All are single stream and working great! LOVE EM, much better than 802.11a 40Mhz

Can you give some details on how you made them work? Are you using nstreme? Any other special settings?

I spent the past month trying to get a PtP backhaul working reliably with R52N cards, but had to swap them out with A cards yesterday. I couldn't afford to just keep trying different tweaks to 'try' to make it stable. 200-600ms latency at times, etc. Couldn't get it to work without nstreme at all because of distance, and with nstreme it seemed flaky.

411ah on each side.
r52n cards, single and dual chain attempts.
-61 and -65 signal on each polarity.
separated antennas at one end, HDDA5W on the other.

put in f50-pros and now get 100% ccq, rock solid, low latency as expected.

I will try N again in a few months once RouterOS flushes out these "bugs". Keep up the good work Mikrotik, just please fix this soon.

Each link is using this:

RB433AH
R52N (they now have a h-pwr version for those in need)
ARC 5Ghz Flat Panels

Single-Chain, Single-polarity (you have to use HT-Ext)

The ONLY thing I didn't try was to physically unplug the second chain from the antenna... I did uncheck all of chain 1 but maybe that wasn't enough. Either way, with dual chains it's unstable for most people on longer links I believe.

Yup, tried HT-Ext, no go.

How long is your link and what is your RSSI like?

Ya link is only like 2 miles, and signal is -65db

It is a little frustrating. It would be nice to know the status of 802.11N with mikrotik. This post for instance has been up on the boards for 5 days now, you would think one of the mikrotik guys could chime in here and let us know that they are working on it, maybe give us a time frame, etc.

When you don't hear anything about it, it makes you assume the worse. Maybe it's broke and they don't know how to fix it? Who knows?

Noticing a pattern here. Short links (1 mile or so) seem to be working for people. Longer links (4 miles is the shortest "long" link I've tried) not so much. Seems like the problem may be in the timing?

I have another new link that is around a mile, dual polarity, no extension channel, not running nstreme and it's working great. I set the HW Retries to 15, disconnect timeout to 00:00:15 and adaptive noise immunity is on.

I brought this up in another thread with no response, but I wonder if Ack timeout can be tuned to help these longer links?

Anyone? Anyone?

Thats ridiculous? We have loads of Mikrotik 11N links from 30 km to 2 km all working great and alot quicker than 11a with no problems?

How are you managing to get such problems? There is a few bits of info like you must upgrade board firmware. Set HW retries to 15.

with nstreme ? max bandwidth ?

Board firmware (/system routerboard print) is upgraded to latest. HW Retries is 15.

Well...here is picture of 5km backhaul link with 2 60cm solid dish antennas and R52N cards on RB433AH

I reached better speed with 90cm dish

What are ping times like when the pipe is half full? With those ping times and drops, there is no way I could call that a usable backhaul.

To be honest unless you have some traffic shaping setup your gonna have those results when downloading with bittorent. I think if thats a single chain link that speed is pretty good. The same thing would happen with a ethernet cable if your maxing it out. Your just flooding it with the torrents. Use pcq queues on the links and set the max rate a little below your best throughput. 11n isn't some magic switch that means you don't need to understand routing? Also when I tested using ip tunnels I did have quite bad results. WDS seems to be OK although some people here say tunnels are faster I have not found that so far but only tried EOIP.

It is a little frustrating. It would be nice to know the status of 802.11N with mikrotik. This post for instance has been up on the boards for 5 days now, you would think one of the mikrotik guys could chime in here and let us know that they are working on it, maybe give us a time frame, etc.

When you don't hear anything about it, it makes you assume the worse. Maybe it's broke and they don't know how to fix it? Who knows?

Seems there is some specializing in the WISP market now.
MT does a great job in Routing/Management of the core network
which they make great progress (OSPF/BGP is very stable now).
450G gives a good tower switch/router and 1000 is a rock solid smart
Backbonerouter. But MT works low priority on wireless (if at all).
U... seems to specialize on Cheap wireless 11n AP/Gear for
the last mile but needs MT to build a network. Nevertheless
it feels not production quality right now. But has some good
understanding what kind of gear a Wisp wants to see at the
customer/ap side. I believe MT und U... should work together.
High Volume PTP Links are best done with licensed gear or
specialized license exempt gear. Using such equipment now
I feel safe. It's expensive but just works exactly as promised.
No problems with interference on Backhaul links is a great
improvement.
Baseline: I no longer wait for 11n from MT.

Well I can say that MikroTiks 11N does work, and works great for us. But we've been a WISP for 10 years so we know what were doing

Glad to hear you know what you are doing.

Are you getting decent ping times with load (say 10-15M load?). I'm not. Based on my original post, can you make any recommendations on what to change that has worked for you? If not, thanks for your time anyway.

We like MT for G links, but haven't had good experiences with N links.

We tried to do an N link, but to use the MT N card, we had to upgrade to a routerOS version that hosed the ethernet ports on the 433ah! The wireless portion didn't work any better than a 54mb G link, so we ended the test so we could get our ethernet ports back and our customers back on line. This was just maybe 2 months ago; I'll let it be a few more months before I try it again.

UBNT's M gear rocks hard for N currently. It had some initial teething issues as you'd expect from them. We've also used Solectek MIMO gear to do similar things very well. We've got a radwin MIMO link that's pretty happy too.

Thanks JP. I guess you and I are in the same boat.. we don't know what we're doing because we haven't been in this business 10 years like spire2z has

I have been at this over 10 years now and have started putting in RocketM for Mimo PtP links. It is just worked out of the box, no messing about needed.

We have thousands of subscribers and cannot experiment.

I would LOVE to have 3x3 mimo working on MT with rb600 on GB. But even the 2x2 the MIMO was unstable, Nstreme was unstable above 3.30 and MT will not give me a model number for a supported GB POE adapter from any maker.

So we have had to use our RB600 with an A based card on 3.30 and at 100mbit eth. We had planned to use a 3x3 mimo with GB ethernet. The forum and MT is full of "oh try this" or "try that", I run an enterprise class network and cannot experiment. If this was a free WRT or something I could understand. MT is pretty amazing in many ways, but the attitude taken by the company that is some sort of free community maintained project is wearing thin when I have thousands of users dependant on stable services.

The RocketM is faster and more stable at this point and if it were to have issues I have direct access to the company president who always finds the right person to help me out. It does not support OSPF, Gigabit etc. but there are devices at each end of the backhauls that do, and I have a great number of issues with MT OSPF making our network LESS stable.

We have bought hundreds of MT units and will continue to buy, but not for the same things we used to. There are just better options out there now and I cannot keep waiting and experimenting. If they ever have a working Gigabit ethernet solution with 3x3 MIMO I may come back for a few busy links but many of my towers are more than served by a couple of RocketMs. For now if 100mbit is not enough (I push over 130mbit on my main gateway in one direction) I am either bonding a couple NStreme turbos together under 3.30 or bonding a couple RocketMs.

I recognize some people have had it work, and I have often been one of the minority to get new things to work, but I simply cannot afford to to do experiments on my clients anymore.

BTW, I am over the moon happy with ROS as my core router. Although I had to put in a MikroNOC 2200 as the RB1000 was far too underpowered for my network. I use the RB1000 for large scale corporate clients to replace Watchguards and Sonicwalls and Ciscos.

I see so many complaints about N on Mikrotik.
I have on my network some N links working on production and they are rock solid.

My oldest one is with 4.0 beta3, uptime since 25 August.

Ok, we cannot use nstream right now, but with ubnt we do not have it either, so, why buying a rocket now, and then exchange to mikrotik again when they get it working?
I prefer to go to the tower only once, and then just upgrade software.

And i have a couple of N links done only with RB411 and R52N with 1 stream and they pull about 95Mb/s so, it was a good investment 2 years ago when i brought the RB411, i only changed the radio and passed from A to N. Now exchange a PS5 to a N device and you will see what i mean, think on 114 backbone links that i wave and you will see the budget i would have to expend.

PS: Disable adaptive noise immunity, only activate it if RB800 or more cpu power, and disable all the a/b/g data rates and adjust the right mcs for that link, disable periodic calibration, adjust hw retries and frame lifetime, my normal hw retries is only 8. Ensure signal between 50 and 60 with very low fluctuation. Carefully adjust antenna alignment for cck>90%. This is my cookbook.

Kindly regards,

Jorge

How long are your links?

The periodic retry disable was one I had not heard / tried before. I'll give it a try.

-- Edit:

Nope, no difference.

The shortest is 400m and the longest is with 8,5km

Most of them are between 4km and 6km

Kindly regards,

Jorge

How long are your links?

The periodic retry disable was one I had not heard / tried before. I'll give it a try.

-- Edit:

Nope, no difference.

With N enabled, what's your signal?

If -70, ht-basic select only mcs-0, and for ht-supported, select only from mcs-0 to mcs-5. For N -70 is very low signal to achieve mcs-7.

start with ht-extension disabled and ht-guard any. for ht ampdu priorities select only 0. Disable wmm support.

Does it get better?

Kindly regards,

Jorge

Nstreme work's fine uffff.

What i do:

RB 433 3.25 > 4.6
reboot
upgrade
reset
reset (after 1st reset)
reset cards
N config from wiki (optimal)

Nstreme is stable.

When I upgreaded form 4.2>4.3>4.4 etc. without reset nstreme was unstable.

I finally got a fairly stable link here. I think either the HT AMPDU priorities were a problem, or the periodic calibration. It still does not work with Nstreme (very high latency).

Hopefully they are working on nstreme so we can do some longer links that non-nstreme won't handle

Randy

Nstreme work's fine uffff.

What i do:

RB 433 3.25 > 4.6
reboot
upgrade
reset
reset (after 1st reset)
reset cards
N config from wiki (optimal)

Nstreme is stable.

When I upgreaded form 4.2>4.3>4.4 etc. without reset nstreme was unstable.

I set up all mine from scratch, no upgrades but maybe there is some problem with config like that if upgrading. Would make sense why some of us seem to have no luck when others have good working setups. And yes Nstream works great for me? The only settings I do that are not standard is to disable all the 11a rates and i'm using sr71-a ubiquity card not MT card.

Thanks JorgeAmaral - i found some of your suggestions very useful.

If it is not too much to ask i think everyone will benefit from a topic in the wiki - "how to set up 802.11n running" - step-by-step instructions, as far as i can see you are one of the most qualified persons do make it.

Then we can make a discussion topic about it in forum and slowly add all our experience to that topic, making it easy for everyone to set up N link in RouterOS