I've received first SXT AC gear for testing.
As I realised there are some strange modifications done to manual select MCS data rates.

HOW can I set MCS4 (QAM16) to be max modulation ? as i have some MCS 0-7, 0-8 or 0-9 to choose from.

HOW to limit modulation?

For the 802.11ac you can't lower the supported data-rates below the MCS7.
So you can choose the max supported data-rate.
Here are the info which 802.11ac MCS represets what:
http://en.wikipedia.org/wiki/IEEE_802.1 ... _and_speed

Also you can specify 2 or 3 fields of supported VHT data-rates. First one is for the 1 Spatial-Stream, Second is for the 2 Spatial-Stream and Third for the 3 Spatial-Streams.

Why would you lower the supported MCS to such low MCS as the RouterOS rate-selection will automatically select the best working data-rate?

So its a big disadvantage as QAM16 works most stable under Mikrotik.
I have all around 25+25mhz QAM16 nstreme ptp links they have 225mbps rate with 100%CCQ and ping stable as hell, no problem with tcp etc..

migrating thru different modulation results in 50-70%ccq ,much worse ping jitter.

In result AC as I hoped for stable 360/360 80mhz QAM16 wireless datarates is unavailable...

Can You somehow force driver to not to go higher than QAM16 ?

Can You somehow force driver to not to go higher than QAM16 ?

The 802.11ac standard and the card hardware doesn't allow such modifications so we can't do that, sorry.

ok, as I thought. it was looking nice at first glance, now I see that AC is not going to be replacement for PtP radiolink.
two SXT AC in totally quiet environment on the desk right now stability in nv2 is unacceptable :/

The first problem is that you are generating the traffic on the SXT board and processing it - that causes a lot of CPU usage and that could cause your problems.

What are the signal levels as it could be too strong signal?
Have you tested both 802.11 and Nv2?
And for Nv2 what what tdma-period size?

Uldis is right, Test like this:

[poweful router] --> SXT1 ###### SXT2 ---> [powerful router 2]

SXT itself can forward large amount of traffic, but to generate random traffic, you need more powerful machines. or actual data (many torrent users etc)

doesn't make much difference.

<rb800>----GE <sxt ac ( ~ ~ ~ ~) sxt AC>---GE---<x86 intel c2d>

data rates 650-780 mbps both sides
it's not stable. but ofcourse it may be desk test issues.
Setting nstreme and frame limit to 1580 so lower than l2 MTU of wireless gives
100% CCQ 866,6Mbps wireles data rate and.... max 200mbps HDX of udp btest.
NV2 is behaving as wicked when no upper limit varying 200-450mbps.
but there maybe some interference issues due to desk test.
I'll try to set it in real environment on monday/tue. As for now lack of modulation
manual choice is biggest dissapoitment of AC for me.

Devices are too close. Lower output power.

please try to check the throughput speed using the Traffic generator:
http://wiki.mikrotik.com/wiki/Manual:To ... _Generator

Also please show us your signal and CCQ levels. Just to rule out that devices are too close together or there is some other factor.

Too close. I tested couple days ago UBNT Nanobeam and had same problem. Lowering output power level fixed problem.

I lowered txpowerf to 0, 1 and 3 with all rates fixed. so i got about 60db SNR ccq varyies from 50 to 90%. ohhh if I could set QAM16...

i have similar problem to this one with ccq and stability in this thread http://forum.mikrotik.com/viewtopic.php ... 98#p440398
cannot anyhow make it works stable :/

ok I got some results. distance 11km, rocketdish and jirous well shielded, clean channel for testing test done over sxt (RB493G-CCR1016)

signal -61/-62 SNR 51 dB

802.11 :
CCq 50-70 / 50-60 %
wireless rate tx/rx 526/351mbps

btest (over sxt) one direction varying 150-250mbps lots of stall

ping avg 12ms max 63ms

nv2:
2 ms frame
ccq 50-70%
wireless rate 270/351
btest 50-150 mbps even more breaks
ping avg 14 max 82 10% loss...
link is UNUSABLE

nstreme:
best fit 3200:
ccq 70-80%
wireles rate 351/520mbps

btest quite stable 160mbps
ping avg 16ms max 27ms 0% loss

So :
1/nv2 is no usable protocol
2/nstreme is no faster than 802.11n 40mhz and nowhere near stability
3/802.11 is nothing we couldn't achieve on plain ac router.

this was my backup link build on AR9220 with 150mbps troughput with ping stable as hell.

either there is something really wrong with ac under mikrotik yet or there is something wrong with my setup.
Anybody got different / better results?

I think they released it too early and soft is not good yet....

Of my research you can try settings like this, for me allowed to get a bit more stability and speed:

Protocol set on 802.11
Mode: 5GHz-only-AC
Data rates uncheck all under A/G supported and basic, then I put MSC 0-8 two rows under supported VHT and the same under Basic.
I set also distance manually to 1km as it is less than 500m and sometimes it connect with 2km or more in status.
Long preamble.

Thats all so far... maybe ll help you a bit ?

I don't get it, how do you connect an SXT to rocketdish and jirous?

using soldering iron. it's quite easy

to many pigtails

There are Two pigtails to be precise. Each for one chain as usual.

there is no need to solder the SXT device as soon you will be able to get the Netbox5 which has SMA connectors like the BaseBox.
http://routerboard.com/RB911G-5HPacD-NB

yes ofcourse. I'd choose netbox if it would be available.

But when rb 11e AC be available?

not soon enough...

small update:

the same pair of sxt ac switched to 802.11 a/n only, can go with 99%CCQ link mcs12 180/180 mbps nstreme
stable 120mbps under btest with ping stable as rock

definetely it's about AC implementation and Mikrotik stuff please work on how to force MCS rates, it would give very stable links on 5ghz.
If You could tweak 802.11 standard in very hard ways up to now why couldn't You do so simple thing as selectable/avoidable datarates in 802.11ac ?

hmm need to test your suggestion thanks!
btw what is your setup for nstream ? I am using Disable CSMA, policy= best fit, limit=4000

802.11 A/N MSC12-23:
UDP Tx 240Mb - ccq 95-100% latency up to 17ms

nstreme A/N MSC12-23:
UDP Tx 180Mb - ccq 95-100% latency up to 17ms

802.11 A/N/AC MSC12-23 VHT 0-8,0-8:
UDP Tx 340Mb - ccq 80-100% latency up to 19ms

So here is our test. Done in a room, 5m distance, Btest running on the SXT itself (usually I recommend running from a different device, through the link, so that the SXT CPU is not overloaded), but as you see the result is still quite good. With a longer distance (1km) the speed could even be better, such short links are not the best example

The signal levels are really high. Should we apply a different doctrine to the new -AC units? I mean, with the "classic" -N ones it's recommended to aim for a signal level between -60 and -70 dBm in order to avoid overload.

Should we keep the signal levels higher with these new -AC units? Won't we have overload problems? In this case, which is the optimum range of signal levels?

The signal levels are really high. Should we apply a different doctrine to the new -AC units? I mean, with the "classic" -N ones it's recommended to aim for a signal level between -60 and -70 dBm in order to avoid overload.

Should we keep the signal levels higher with these new -AC units? Won't we have overload problems? In this case, which is the optimum range of signal levels?

Sensivity for a 20MHz Channel MCS-9 is -72 (SXT AC). For 40MHz -69. So I guess -60 is the sweet spot. With -70 you will not get 256-QAM 5/6. With 20MHz Channel you get 256QAM-3/4 Max so you want 40MHz Channels.

We also made the same test with signal -55, but the speed is the same.

Should we keep the signal levels higher with these new -AC units? Won't we have overload problems? In this case, which is the optimum range of signal levels?

Sensivity for a 20MHz Channel MCS-9 is -72 (SXT AC). For 40MHz -69. So I guess -60 is the sweet spot. With -70 you will not get 256-QAM 5/6. With 20MHz Channel you get 256QAM-3/4 Max so you want 40MHz Channels.

Thanks

I was wondering, as the test performed by Mikrotik shows signal levels around -30 dBm

very nice normis
How to explain then the situation on my sxts ? I can see your speeds are kinda stable while my likes to flap... also your ccq looks poor but probably due to desk test.
I am not able to get stable ccq and speeds but as I made last tests it looks better than my other tests.
I can create supout.rif so you you may take a look maybe there is smth wrong, what do you think ?

very nice normis
How to explain then the situation on my sxts ? I can see your speeds are kinda stable while my likes to flap... also your ccq looks poor but probably due to desk test.
I am not able to get stable ccq and speeds but as I made last tests it looks better than my other tests.
I can create supout.rif so you you may take a look maybe there is smth wrong, what do you think ?

hmm, interference, reflections, noise? did you try complete reset of config, only setting the IP addresses, and trying the link with regular 802.11 without nstreme or nv2? did you try to move to different frequencies, or changing location of the test ?

so far I could achieve this: hmm bridge is under omni antenas and SXTs are with shield right now (which did not bring anything better)
I tried at night to turn off this 2 omni anteas to see if there ll be better results but also nothing better.
About test it in different place, that could be possible but so far do not have time for that but I am not sure if other bridges has so good clear visibility, anyway I ll take that as an option to check.
About reset I did not make full reset, so far I made radio reset and config using quickset.

we are using sxt 2hnd and sxt lite5 for two different customer and link distance at both the location is 300 meter.however at both location when link is in ideal condition we find a heavy traffic flow and with signal strength of -39 db packet delay is almost 150ms also when we bringing device back to back at a distance of 5 meter inside the room we face a similar type..what could be the reason for packet delay and traffic flow

could you specify what do you mean by heavy traffic flow ?

as I got advised before what could eventually help:
1. reset to defoult.
2. upgrade to new stable soft - check forum if there are no issues
3. you need to care signal be not so strong -39 is too strong, lower the power, what MT said around -55 -60 should be perfect
4. check frequency and be sure there is no noise
5. if nothing help then maybe one device is bricked...
Greetings and GL

BTW my sxtgs actually are not able to transfer more than 100Mb/s tcp while competition solution is doing easly 400Mb/s out of box xD waiting for better weather to switch MT.