Solving 20km wireless link issues

RouterOS Wireless Networking
1

Okay guys, I’ve been on Mikrotik for years now and know it inside and out, except I don’t have a grasp on wireless stuff completely yet. I just put up my first backhaul 2 weeks ago - 20km! Woo Hoo I tell myself.

433ah w/ XR5 on each end
PacWireless GD5W wideband grid dish, 28db

Ive been struggling over low signal strength since installation. -76 is the best I could get, with -80 being the average. That’s horrible I tell myself, must be something blocking the signal, or these antennas suck. Well out of the blue last night, without touching ANYTHING for days, I see this:

Excellent signal, -60db! Nothing changed that I have control over, so what happened? We’ll some thoughts I wanted to bounce off the ‘team’:

1 - The data center side is on a huge roof thats metal, although I am 15 feet above it.

2 - I am shooting over an air conditioning unit, as shown in the picture, but I believe I am 4-5 feet higher than it still.

3 - RadioMobile shows a good view of the link, about half way there is a community that I probably am just barely shooting over.

4 - At first I thought I was just getting into the fresnel zone and was going to find another option. Now last night the signal increases by 20db and is giving excellent performance.

5 - Without even touching anything the signal is affected - but why ? My guess is that there is noise - but every single channel gives me the same results, if I can get connected on them that is.

6 - Noise floor before and after this is the same, -102db average.

So how do I go about tracking down what’s crushing the signal across the entire spectrum ?

One side is on a water tower, however moving 100 yards either way doesnt seem to help:

And the other side on the data center:

Any help troubleshooting would be appreciated.

2

I believe the Frequency Use tool is only showing you 802.11a frames received. It does not measure raw energy received at the antenna like a spectrum analzer would. You could have an interference source and not know it from that tool. You could hook up a real spectrum analyzer to see what noise you might really have; however, you said it’s consistent performance across all the channels, so I doubt you have interference across every 5.2, 5.4, and 5.7 channel (it’s possible just not probable).

Given that it suddenly changes and if you see that sudden 20db change on all channels, you should look into the equipment. I’d check the pigtails and coax. Also, I’m not a fan of those antennas for backhaul links (much prefer Radiowaves or Andrews parabolics).

3

BTW: GREAT posting. Lots of clear data, good descriptions, graphs – I wish everyone posted like that. Very professional.

4

yeah im tired of posting like ‘my signal isnt good, help me’ : )

I did order a completely separate rb433ah with an XR5, exact same setup - but portable enough for me to drag it around for testing. I got the same results using the test setup 100 -200 yards away (on the water tower side). I think that would rule out bad hardware / power at the one site, and I am going to do the same on the other side this weekend. However, if one side was bad wouldnt the signal tx / rx be way different if it was only 1 side with the problem ?

It is 20db across the entire spectrum, which makes me wonder if it’s the air conditioning on the roof that stopped working, or a massive power line thats in middle that went out or something. I am using horizontal polarity, I am going to try vertical in the next day or two as well.

5

However, if one side was bad wouldnt the signal tx / rx be way different if it was only 1 side with the problem ?

Not really. What if the pigtail at the data center side is a 20db attenuator? Wouldn’t that affect tx/rx at the same time? If you had highly asymmetric signal levels, then you would start to look at the card(s).

You didn’t mention lightning arrestors. Are you using any? If so, try swapping it out for a pass-through.

6

Currently it is the built in cable from the GD5W directly to a MMCX / N pigtail - no arrestor or LMR in there anywhere yet. The radio is directly mounted behind the antenna so the cabling is very short; I thought that might interfere as well, but why would it get better for just 8 hours and then go back to the way it has been since installation? If it was cabling would I possibly see that signal increase for 8 hours and then back to what it has been doing ? I won’t rule out the cabling / hardware until I test both sides with another unit, but I’m leaning towards something outside of my control causing this. Just trying to figure out how to pin-point it. Thanks for your input too : )

7

It’s not a problem from pigtail, or connectors; I had the same problem, there is problem regarding the propagatin; I couldn’t find a fix, maybe replacing with bigger antennas or/and changing polarization; sun also causes to loose signal. In my case also when it was cloudy the signal was excellent or during nights.

good luck

8

Great post changeip!
Looking at radiomobile it seems you have your fresnel kinda messed up but try to put the correct values of antenna height maybe it’s not that bad once u set at around 15-20 meters the antenna over the roof of that building..
And i experienced the same problem on links over the sea ( http://forum.mikrotik.com/t/day-night-problem/22470/1 ). I almost fixed it raising a little the azimuth of both antennas, losing few points of signal strenght but avoiding a lot of those problems..
Also i suggest you to do a backup link in another frequency… it should solve you all the problems i guess..

9

The signal strength is the signal strength, regardless of interference. Interference changes the SNR, and interference isn’t so easy to reliably measure.

I would suggest pigtails, cables, feedhorn changing/swapping, etc… We’ve seen feedhorns go bad on grids and dishes where the signal is usually weak but intermittently good. This has happened to us with Pac Wireless and Gabriel feedhorns. You might have started off with a feedhorn that is mostly bad. If you have a spare antenna, borrow that part or antenna to test.

It is possible the shape of the metal roof causes some destructive multipath interference at times. You would have to temporarily move the setup to a place where it didn’t shoot over the metal roof, but other aspects were the same to test this. Some metal roofs do affect signal, but it’s really hard to predict.

10

Sorry JP, but that is not necessarily true.

Reflections can be strong enough to cause the LNA to become saturated causing the low signal strengths.
Metal structures can be troublsome.
Try lowering the Output power and see if the receive strength increases.

Metal distorts magnetic fields.
Microwave can be viewed in a simular manner, in which the field can and will be distorted by nearby metals.

Not to mention any evapoartive effects from the metals heating and cooling.

(Have you ever tryed a radio shot over loadstone?)

11

I thought I’d mention… the signal has sucked for at least 3 weeks, but it has been VERY consistent. -80db almost all the time. However, that one evening it went from -80 to -60 and then back again after about 8 hours or so.

Tomorrow I will be testing an entire new setup on the data center side to see if its hardware / antenna. Will post more info w/ more pics shortly.

12

I also have 25km link with similar behavior. Night sometime signal goes to -69 usually is -80 but sometime goes to -90 and drop. But I possible have problem with fresnel zone one of antenna is not very high.
If you find what cause this please write here.

First setup I use R52 card and latter replace with R5H with R5H work better sometime drop but establish very quick.

13

It might have something to do with temperature changes within the path.

14

I had the same problem and would like to ask you if the signal was increasing when it was cloudy on both sides?
First I used for this link SR5 with dished of 85cm and the signal was -81, throughput 28Mbps and then replaced with XR5 and dished of 120cm and the signal increased to -77. Use only dishes(offset) for backhaule because it works well even when interference is present.

15

Sounds like a problem I’ve had here.

I’ve got one customer who only lost signal between 2:30 and 3:00 PM, usually for under 10 minutes, same time, every day.

The heat coming off of the roof of his house was causing the problem. Once the sun went down, the problem disappeared. No problems at all during the winter. Still need to raise his antenna a bit, but apparently the system is more resilient than it was in the past.

16

Okay, some updates.

Yesterday I changed out the data center side completely. New rb411ah, XR5, antenna, cable, pigtails, etc. I also made a point to swap over to vertical polarization (from horizontal) to see if that helped.

1 - I gained a db or two during the horizontal to vertical change, however I think that’s probably because of the extra effort during alignment. Still sitting around -78 to -80 very steadily.

2 - Antenna alignment on horizontal was 10x easier… in vertical I really had to get it within a few millimeters turn to get it locked on. In horizontal it seemed like I could move the antennas 3-4 inches either way and still be connected.

The link has been up for 3-4 weeks now. It has been right between -78 and -82 the entire time. Only 2-3 times has it magically gotten perfect, -60db for about 4-8 hours each time.

Some thoughts, tell me if I’m way off on any of these:

1 - the freznel zone has nothing to do with this (directly), otherwise I wouldn’t see it all of a sudden get better (unless some obstacle in the fresnel zone all of a sudden moved).

2 - the freznel zone is indirectly related in that when the weather has effects (temperature inversion, fog, atmosphere)it helps the overall freznel zone effect. Rather than weather affecting it in a worse way it might be making it better? I can’t confirm it was weather when it got better, but that’s my guess.

3 - I’m shooting thru a high voltage powerline about 0.5 miles away, is it possible thats affecting it? Maybe the power on that line was turned off that time it got better? I wouldn’t think so, but just a long shot.

Both sides have been replaced independently. Each side was moved 100 meters to the side with no changes.

I’m somewhat at a loss at what to try next. I do know that with the existing hardware and link path I have gotten -60db, but what options do I have to try to keep that as the normal strength? If I change out antennas for dishes is that going to make any difference? If I raise either side 10ft (on a 20km link) would that help much? Should I switch to 2.4ghz? I assume that will make the freznel zone larger…

17

changeip,

Try lowering your tx power in half at both ends.
Check rx strengths again.

18

okay, i did test that just now. changed the setting to ‘card rates’ and then stepped down a little each time (on both sides at the same time).

xr5 card rates:
20 - 86db-88db
22 - 84db
24 - 82db-83db
25 - 80db-81db
26 - 77db-78db
27 - 76db-77db

any lower than 20 and i couldn’t get a link at all.

19

Sam..

You said that the signal for better for 4 - 8 hrs at a time,

The data center ( am assuming Sandiego L3 ) follow me here…

Dont they use DC rack equipment… Isnt there a LARGE btty UPS… system… (charging system and inverters)

Also, the other antennas on the platform..

What I am getting at… is is ther a chance that you could look at the schedual that the signal improvovrd, and see of one of the other systems in the area (eather sise) went off line for maintanance , schedualed or not,

I am NO wireless expert, but the timming sounds “interresting”, 4 then 8 her usualy at 10:PM or so…

It would possably allow you to find the source by “connecting the dots” if there was a maint outage of a system, then find out when it will be down next or if “crashes” when it did and match it to your graphs… Then smite the offending pice of equipment with a few well placed hammer blows.. :smiley:

I think I would wander down that road for at least a short while,

20

Looks like propagation effects, not interference.

I did some multipath calculation and the expected signal at the receiver, for 28dBi antennas and 23dBm TX should be -56dBm. At this distance even over sea surface, same antennas, the SLA is 99,99935% (e.g. 4 minutes total outages in one year). As your RSL is -80dBm … the fresnell zone obstruction at the middle of the path can not be neglected.

Since the modulation is OFDM I think that multipath desctrutive atenuation is out of question because this kind of interference is mostly narrowband so you miss some OFDM carriers but the avereage received power must remain the same (except if there is several paths with different time delays but the terrain seems pretty mixed).

Along the day when there is not any significant changes at temp X humidity in the valley between the mountain and the obstruction the signal still stable. But after sunset or just before sunrise (e.g overnight) some ducts can show up, acting as waveguides or reflectors, making the beam change its direction or elevation. This can exist durying the day, depends on the local weather.

The obstruction at the middle can create something called knife effect too, e.g. bending the beam downwards (good if you have a probe receiver some meters down hill :smiley:).

And looking at the antena place at the building … microwaves does not like very much big surfaces in front/below the antenna (the metal roof, a mirror). Sugestion: assemble anoter rig at this metal roof edge and take some RSL readings. I had some trouble once at 2.35GHz STL from a TV ENG van which was assembled in the middle of the van roof. Depending on the parking position the signal desapear. One (1) meter movment and lots of signal show up.

Regards;