Interferences and MCS selection

Products ePMP
1

Hello,

I installed an ePMP 1000 GPS Synced with a 90° sector in a very bad environment.

The lower noise I can find is about -65 in the frequency range you can see in this screenshot:

screen_freq.jpg
screen_freq.jpg784×369 70 KB

I have very very close clients (300 meters) and the noise is only on AP side (so, only upload low speed for clients), almost no noise on SM side.

Other frequencies has unusable spectrum (up to -35 average noise).

Using MikroTik wireless equipment, on the same antenna and frequency, I can reach a maximum speed of 10Mbps aggregate on 20MHz channel, not very stable, and 5Mbps upload max.

Using Cambium, with 2.5ms frame size, 20MHz channel and 75/25 I can reach 70Mbps in download and 5-6Mbps in upload.

The problem is these speeds were true only in the "Wireless Link Test" (UDP).

Using speedtest.net (TCP) showed much lower results, and latency was jumping around.


I analized the Performances screen both on AP and SM (only one at the moment) and I discovered I had about 40% retransmission on uplink from SM. Traffic was jumping from MCS15 (1-2%) to MCS9 (40%).

I noticed the "big numbers" were on MCS10 (40%) and MCS9 (50%), so I decided to lock uplink to MCS10.

The TCP speed improved a lot, but I was seeing some retransmission also on downlink, so I decided to lock downlink on MCS14.

Now I have only 3% downlink retransmission rate and 5% uplink retransmission rate.

The speed is very close to UDP test (70 downlink, 6 uplink) and I'm satisfied with results.

I also tested 40MHz and I had 151Mbps downlink and 10Mbps uplink, so great results even if the spectrum is very noisy.

But... My question for Cambium is: 

Why I have to manually analize these numbers for each SM and manually set both to the values that give me the best?

I mean... The ePMP already knows MCS15 is unreachable for upload for this client: why trying to transmit and keep retransmission rate high? It should lock itself on MCS10 max because is the highest usable modulation, and maybe check hours later if it can transmit on higher modulation.

The same should happen for downlink. I was forced to use MCS14 as max modulation, but what if I have another client, closer than this one, that could reach MCS15 without lots of retransmissions? It would be locked at MCS14 because I set this number on the ePMP to keep the other client at its best, wasting airtime that it could use at MCS15... 

Here Cambium talks about eFortify as a technology to react to external interferences, but I think you can do more to optimize that, even if the performances I reached on this link were very good considering the environment.

2 Likes
2

What do you have the  Subscriber Module Target Receive Level set to on the AP? What model client radio are you using?

I believe that Cambium will spend some time this summer improving the rate selection algorithms when sporadic interference is encountered and there will be less circumstances (ideally none) where you will have to manually set the maximum uplink modulation rate.

In the past we've only had a couple of clients out of the hundreds we have on ePMP that we've had to manually set the maximum uplink modulation rate. We've fixed some of those by changing channels and/or using higher gain client equipment (Force 180 or Force 200).

Lastly ePMP 2000 is around the corner and with its beam-forming capabilities, it will definitely help to increase subscriber uplink SnR.

2 Likes
3

I had to set -50 as target, it’s a Force 180. I also tried -55 but the speed was unstable. At -60 I had less than 1Mbps.
The client is very close to the AP (and also others will be), so I can reach these signals with low power.
I would like to keep -55 or -60 to enable frequency reuse in the opposite direction, but at the moment it’s impossible… Maybe ePMP 2000 will help on this!

I can confirm it’s my second case on locking MCS on about one hundred customers (for now), but I wanted to report my experience since I believe Cambium is listening its customers to make better products!

1 Like
4
@giuseppe4 wrote:
I had to set -50 as target, it's a Force 180. I also tried -55 but the speed was unstable. At -60 I had less than 1Mbps.
The client is very close to the AP (and also others will be), so I can reach these signals with low power.
I would like to keep -55 or -60 to enable frequency reuse in the opposite direction, but at the moment it's impossible... Maybe ePMP 2000 will help on this!

I can confirm it's my second case on locking MCS on about one hundred customers (for now), but I wanted to report my experience since I believe Cambium is listening its customers to make better products!

have you tilted your sector to only reach as far as it has to go?    does it have really good front to back?  is the beam only as big as it needs to be?    what do your DFS channels look like?    

the 2000 im sure is going to help a lot, but telling your subs to wait a few months might not go over well.    

http://www.proxim.com/products/knowledge-center/calculations/calculations-downtilt-coverage-radius

there is always the simple dirty method,   keep tilting your sector forward until your furthest out CPE has LOST 3db.   check your noise level.    9/10 you'll gain far more than 3db SNR you've given up in RSSI.    if you see a big improvement, get greedy and go further, take 6 points off your furthest sub RSSI and check your SNR again.     

as far as why does the gear try and go higher, the noise probably isn't constant, its constant enough to show up on the graph but it doesn't consume all of the air time, so it tried to go up.      the counter idea of the MCS never going back up after being pushed down would leave the sector slow. 

3 Likes
5

Thank you for your comments!

As I wrote on my first post, the subs are very close: about 300 meters.

At this time, I'm not pushing that zone too much cause of interferences: I would like to keep only 4-5 subs there, and maybe I'll change my mind after ePMP 2000.

Unfortunately all the noise is coming from other non-synced equipment from my competitors from all directions... There are tons of wireless devices on the same tower, so changing sector direction/elevation isn't going to change the noise floor (I already tried).

Maybe a narrower beam antenna (it's 90° now) and a better FBR (it's 30dB now) can change things. I'll try :)

I wanted to share my experience because the results are not bad considering the noise, and because I think Cambium can improve MCS selection even on intermittent noise.

You're right saying it would leave the sector slow, but I imagine something that can "track" MCS and retransmission during time, not "a lock" on low MCS, and take smart decision.

6

http://www.rfelements.com/products/wireless-broadband/symmetrical-horn-antennas/overview/

I bet those little fellas could do you some good here.    some of them are 40db FTB.   

 if at all possible, try and mount it so the same tower antennas are behind its face so you can get all the possible FTB you can.     the tighter the beam, the better the FTB and gain with those, so if i were in your shoes, get the tightest, and you'll end up almost pointing it straight down, hopefully dodging your noise. 

 there patterns are perfect circles though, kinda strange to aim, but for hitting a small PMP target, they are pretty handy little guys.

1 Like
7

I was thinking to Horn antennas :-D

I use them a lot in similar situations, and they're very good. I've some 30, 50, 60 in my stock. The 50° has a stunning 53dB FBR.

I use them a lot when I want to cover close clients in small areas with differents heights that a narrow vertical beam can't cover.

I think I'll try the 50 degrees on this tower!

The bad thing is the gain, only 14dBi on 50 degrees...

8

14dBi on 50 degrees, first thing that comes to my mind is higher power output for DFS.

9

I shouldn’t have any issue on 300 meters

10

Correct, I'm wondering how the noise floor in 5500-5700 looks and if you could avoid that interference or if they all look that bad. 

11

Unfortunately it doesn't look so good :-(

5500.jpg
5500.jpg792×183 42.8 KB