Community discussions

MikroTik App
 
User avatar
tgrand
Long time Member
Long time Member
Topic Author
Posts: 667
Joined: Mon Aug 21, 2006 2:57 am
Location: Winnipeg, Manitoba, Canada

NV2 Sync

Wed Oct 14, 2020 4:42 pm

We have many sectors which use NV2 and have had good success with NV2.
With that said there are issues with Sync loss.
We are using 30deg RF elements twistport horns with the Shielded RB Enclosures.
I believe that this limits the sectors from hearing the beacons affecting Sync.

What is the feasibility of using omnitiks as NV2 masters with no clients?
Would there be broadcasts that could be seen by clients of the sectors which could impact performance?

If anybody has Ideas about this please advise.
Perhaps even MikroTiks in-house wireless engineers have suggestions?
 
server8
Long time Member
Long time Member
Posts: 592
Joined: Fri Apr 22, 2011 1:27 pm

Re: NV2 Sync

Wed Oct 14, 2020 8:18 pm

It's a good idea if the horns can hear the omntik with a good signal. You must have an omnitik for each frequency.
 
User avatar
tgrand
Long time Member
Long time Member
Topic Author
Posts: 667
Joined: Mon Aug 21, 2006 2:57 am
Location: Winnipeg, Manitoba, Canada

Re: NV2 Sync

Wed Oct 14, 2020 8:31 pm

My real concern is : will there be unwanted broadcasts outside of the sync beaconing which may interfere with data broadcast from the slave sectors?

P.S. to Mikrotik: would be nice if there was an NV2 beaconing mode as opposed to AP-Bridge. If, and only if, the OmniTik Master works in this configuration. Or perhaps it is unnecessary.

Will keep this thread posted.
 
User avatar
tgrand
Long time Member
Long time Member
Topic Author
Posts: 667
Joined: Mon Aug 21, 2006 2:57 am
Location: Winnipeg, Manitoba, Canada

Re: NV2 Sync

Fri Nov 06, 2020 8:24 pm

I have deployed an OmniTik5ac.
It acts as the NV2 Master, with default authenticate off.
All Sectors are syncing, and so far everything is playing well.
 
digicomtech
Frequent Visitor
Frequent Visitor
Posts: 77
Joined: Fri Apr 20, 2007 5:03 pm
Location: Alma, Qc, Canada
Contact:

Re: NV2 Sync

Wed Dec 16, 2020 1:31 am

Hello,
Are you saying that you use NV2 sync with success ?
If so, are notice performance leak in upload.

For example, 2 APs with 60 Mbps capacity over the air, are able to deliver each 40 Mbps in download in BTest TCP-4 with a CPE using NV2 sync at 70% downlink ratio. But when try to test the upload with the same CPE on both APs, we can hardly get a 5 Mbps...
30% of 60 Mbps = 18 Mbps
So, I was exepected something around 10 Mbps at least...

Do you have some working example ?

Regards,
Michael
P.S. Remember Chicago MUM ;)
 
marekm
Member
Member
Posts: 379
Joined: Tue Feb 01, 2011 11:27 pm

Re: NV2 Sync

Fri May 28, 2021 10:28 pm

Is nv2 sync production ready? Last time I looked, it was labeled as experimental.
One thing that makes me a bit nervous is the single point of failure - loss of sync master causes all sync slaves to stop working.

In the most common setup - 2 back-to-back sectors at the same location, on the same channel - it would be nice to automatically select one AP as a master and the other as a slave.
If the current master goes down, the slave becomes the new master.

Also, it would be nice to extend this feature to sync APs on different channels, from a common time reference.
Hardware has no GPS, not sure about 1588 PTP support, the next best thing could be sync from a common NTP server on the wired LAN.
The NTP server itself doesn't even have to be very accurate, as it is the relative timing between APs that matters.
 
User avatar
mkx
Forum Guru
Forum Guru
Posts: 11381
Joined: Thu Mar 03, 2016 10:23 pm

Re: NV2 Sync

Fri May 28, 2021 10:47 pm

The NTP server itself doesn't even have to be very accurate, as it is the relative timing between APs that matters.

You're right, absolute time is not important. However, clocks on co-located APs should be synchronized to a few ten nanoseconds ... remember, standard duration of guard period in 802.11n is 800 nano seconds and short guard period is half of that. And APs should be synchronized with precision which is fraction of guard period in order to efficiently co-operate in scheduling of transmissions.
Normal NTP allows synchronization with time error (offset from reference time) of about mili second (and jittering), which obviously is not good enough and that's where IEEE 1588(v2) comes into play (and that's why PTP is needed instead of NTP). Or GPS receivers (which were used in single-frequency networks such as IS95 before IEEE1588 standard became feasible solution of this problem).
 
marekm
Member
Member
Posts: 379
Joined: Tue Feb 01, 2011 11:27 pm

Re: NV2 Sync

Sat May 29, 2021 6:35 pm

More accurate is better of course, but I don't think nanosecond-level accuracy is really necessary.
Total frame period is a few milliseconds, for example 4 ms divided between 3 ms Tx and 1 ms Rx.
Make that 2.7 ms Tx with 0.15 ms of silence before and after for NTP jitter, this means just 10% less Tx bandwidth.
And 0.15 ms looks realistic with NTP if you connect the NTP server to the same gigabit switch (or the CRS switch itself as the NTP server - not sure how good is the ROS ntp package though, it's a separate optional package so probably not very widely tested).
It wouldn't work over a wide area like Cambium or Mimosa with GPS sync do, but should be fine for a single site with ABAB or ABCABC sectors saving half of the spectrum.
The two methods could be combined, NTP to reduce co-location interference on different channels, and listen for beacons of the other AP to refine the timing on the same channel.
NTP if properly implemented works like a very slow PLL, takes a while to sync initially but does a good job of filtering the network jitter.

Who is online

Users browsing this forum: No registered users and 38 guests