Hi,

Can other wisps post advice and tips for getting the best out of my NV2 PTMP setups?

I have sectors/CAPs with rocketm5/airmax that reach 30-40meg every night. The other night I saw a rocket peak at 58mbit with 25 registrations.

On the other hand I have my Mtik CAPs which are usually basebox5 and I also have some mantbox19s.

I strictly use 5ghz-onlyn at 20mhz

I don't use ac even when available. Most of my CPEs don't support AC and you shouldn't need AC to get more than 20mbit out of a dual chain 20mhz link!

I use spectrum scanning software like spectral history and ubnt airview to find clean channels and test thoroughly with single client, single stream tcp bandwidth test before allowing others on. I also try to keep tx power down as much as possible.

The current ceiling in most of my nv2 ptmp setups is 12-13mbit. A couple are peaking at around 20mbit in bursts.

All cpes are either SXT-5HPND or LHG-5. Signal levels are never worse than -65 to -68 with good modulations.

Out of my Mikrotik CAPs I have one with 25 registrations, one with 16, one with 12 and the others are 10 or less.

I don't want to bash Mikrotik in any way, I accept that I must be doing something wrong, I just do not know what, and I desperately need more bandwidth to my customers at peak hours.

Regarding coding rates I've had people advise me to
A) allow only mcs0 and mcs11 in both basic and supported rates
B) allow only mcs11 in both basic and supported rates
C) don't limit or configure rates in any way (leave on default)

What do you do to maximise sector/cap throughput and what results do you get.

Any advice that could point towards the right direction would be most welcome as I've started to lose hope.

Can anyone confirm that they can regularly hit 50-60megs with ~25 registrations on 20mhz-N and NV2. Is this even possible or am I chasing a dream.

Please allow the holiday spirit of compassion towards fellow human beings to guide you in proving a helpful response to one frustrated colleague!

Oh and Happy New Year everyone!

Bump? Anyone with feedback from their own experiences?

Seems like people don't like to share their experiences on this forum.

How much are you getting out of a single Mikrotik NV2 access point at peak usage?

10meg? 20meg? 40meg?

20mhz only-N nv2 11 clients 30-40Mbps

20mhz only-N nv2 11 clients 30-40Mbps

Many thanks for your reply it is greatly appreciated!

Really hope others also join in on this thread.

40MHz Omnitik with an SXT AC client (no real advantage, just listing parts), 120 to 180Mbps air rate, around 70Mbps throughput during testing.

One of my other Omnitiks has 100Mbps + with the same config, just slightly better signal.

Edit: Actually just realized it was running Nstreme instead of NV2. I swapped to test, 170+ mbps.

Seems not a lot do this now or dont want to show their bad results.

With nv2 we see good tcp results with low cpe numbers (50-60MBit/s) as soon as there are more cpes connected tcp goes down while udp keeps higher numbers. There we get complaining customers as speedtest.net does tcp.

Imho nv2 needs a rewrite and adaption to .ac speeds. No changelog messages regarding nv2/nstreme for a long time now.

Anyone done testing comparing nv2/nstreme/plain .ac with newer ROS-Versions? In office we see plain .ac outperforming nv2 and nstreme.

Same results here on sectors. When above 10 clients trhoughput drops to much with nv2

Verstuurd vanaf mijn D5803 met Tapatalk

Same results here on sectors. When above 10 clients trhoughput drops to much with nv2

Verstuurd vanaf mijn D5803 met Tapatalk

Did you test plain 802.11 with rts/cts or nstreme on these sectors with newer ROS releases? Did some indoor testing. Plain 802.11ac outperforms both easily. Even the 720Mhz OmniTik ac cpu seem to weak to handle the protocols using 40Mhz Channels.

Nv2 tdma is best option as clients can not hear each other and need to be served on time slot base.
NV2 is very stable for outdoor. Only problem we see is, overall throughput on the Ap is very low.
We do play a lot with the settings but it looks like its limit is in hard or software.

plain 80211ac testing indoor, does not reflect an outdoor tower with clients not hearing each other, so its always performs better indoor when testing.

But if we cant see code or mechanism, nor get any input on what is causing or limiting the throughput. we only can guess and complain.
If we know its limitations we can calculate this into the network, now today i do see a little gap here in theories and real live

http://www.radio-electronics.com/info/w ... igabit.php

Here my info / results of testing ;

For the AP radio on N-40Mhz , the max total throughput for clients is around 225Mbps TCP

this amount is shared among any connected active clients.

However,

The total upload and download amount of all clients is the total use of the AP.

so for example, if all connected clients are downloading total 200Mbps and uploading total 20Mbps is the max of an AP ( 220Mbps tcp ).

or the other way around ;

if someone is uploading 200Mbps, then you have total 20Mbps available for clients downloading

this is when all client radios are connected with 270/270mbps rates.

When 1 client is connected with lower rates and maxing out that connection....well
to make a long story short. :

The lowest Tx connection rate to an active downloading client, is the total radio rate available to all clients on the AP.

This can be somewhat be prevented by setting a tight limit on the down/upload speed of the "bad" client. so you have time to fix the client.

Here my info / results of testing ;

For the AP radio on N-40Mhz , the max total throughput for clients is around 225Mbps TCP

this amount is shared among any connected active clients.

However,

The total upload and download amount of all clients is the total use of the AP.

so for example, if all connected clients are downloading total 200Mbps and uploading total 20Mbps is the max of an AP ( 220Mbps tcp ).

or the other way around ;

if someone is uploading 200Mbps, then you have total 20Mbps available for clients downloading

this is when all client radios are connected with 270/270mbps rates.

When 1 client is connected with lower rates and maxing out that connection....well
to make a long story short. :

The lowest Tx connection rate to an active downloading client, is the total radio rate available to all clients on the AP.

This can be somewhat be prevented by setting a tight limit on the down/upload speed of the "bad" client. so you have time to fix the client.

This changes when there are more CPEs connected. The UDP Speed stays but the TCP speed of every single CPE goes down.
With 20 connected CPEs you will not see anyone getting close to 100MBit/s TCP even when the other 19 do nothing.
With 20MHz Channels you end up below 20MBit/s speedtest result once your sector is loaded with 20 CPEs.
You have to argue to customers that they get their speed while watching netflix as this is UDP but they dont take this.

Somewhat off topic, but are you guys in FCC land?

What TX power does the Omnitik AC have? I noticed the SXT 5s have been nerfed pretty hard to meet the new FCC regulations.

This topic is old seeing by the date, but I believe I can help, since I suffered trying to get good settings. I'm using Google translator so some sentences will appear strangely, but let's get down to business.
For a long time I left the channels (scan-list) at 20Mhz by default, I chose one of the frequencies, I didn't touch the HT or the nv2 settings, such as tma, radius, nothing.
But as the years went by, I updated the customers' plans (competition was tightening in rural areas), and from the initial 02Mbps I'm now serving with 15Mbps... as a result, problems have arisen in the APs to hold the loads.
The busiest APs I have are an OmniTIK 5 with 27 clients hanging and a mANTBox 19s with 48 stations, so at peak times I saw a maximum of 15Mbps traffic on Tik5 and about 22Mbps on box19s and the latency was high and also a lot of lost packet.
So I guess it's time for the tweaks that gave me peace of mind on the phone on the weekends.
- Band/channel: Only-N 40/eC
- Scanlist: OmniTik (5100-5500 or 5300-5700) | mANTBox19s (5400-5900)
- obviously NV2, in superchannel with no-country
- configure the gain both on the AP and on the stations, it will help A LOT to stabilize the connection
- Data-rates left in default
And now in NV2 the settings that gave great results for me.
- TDMA: 2ms (1 and 3 were not very stable)
- Radius: if you have any antenna in particular giving packet loss, while all others don't, put a maximum of 12 instead of 10, 15 for me was not stable
- dynamic downlink with ratio 75
- most importantly: put QoS in frame-priority and increase the Queue, for me, my mantbox with 48 stations was at 8 (max).

The results for me for peak hour traffic are (borderRB>station) 8-25ms (in normal operation) and 35-95ms (when maxed out traffic) maximum 60Mbps for OmniTik5 with 27 stations and ~75Mbps for mANTBox19s with 48 stations.