Hi, guys,

I'm trying to solve two separate problems with the uplink of my hot-spot installation, which consists of a RB433AH, 2 x 100mW, 2.4GHz Compex WLM54G23 cards - one of them for uplink trough 24dBi, 10°, 2.4GHz grid antenna, the other via 2.4GHz 15 dBi omni antenna for a hotspot. Both antennas are using horizontal polarization, operating at 2427 and 2452 MHz respectively. Everything is mounted on one and the same pole, on the beach (read it as 5m above the sea level).

The other side of the uplink is 5km away, on the other side of the bay (thus 90% of the link is over the sea) mounted at 70-80 meters above sea level. My ISPs AP is Level One's WAB-3001. According to the specs, it's all-in-one device with integrated built-in 9dBi flat panel antenna and unspecified 20dBm wireless card.

Due to lack of reporting features of WAB-3001 I can provide link information only from my Mikrotik:
Tx/Rx Signal Strength: -55 to -62 dBm, depending of the time of day
Noise Floor: between -95 and -100 dBm
Signal to Noise: 35 to 42 dB

The link is limited to 36 mbps in both directions.

Here comes the strange thing: CCQ is quite stable at 85-100% most of the time, but suddenly drops to 40-50% and in extreme cases goes down to single digit values. CCQ stays low for some time (from few seconds to few minutes), than recovers.

The first problem I discovered was that omni antenna, being so close above (even part of it behind) the grid antenna interferes with it. The problem occurs when someone, connected to the AP (omni) starts very fast and intensive (think torrents) file transfer. At this time CCQs of the both antennas drops sharply and the speed of the uplink, especially TX speed is changing in the whole range between 33 down to 1 Mbps like crazy every 3-5 seconds or so. You can imagine what happens to the effective throughput and latency over my uplink. The same transfer initiated over wired ports of RB433AH does not affect CCQ and speed at all.

Changing polarity or frequency of the uplink was not an option, so I decided to increase the distance between two antennas on the pole and change the frequency of the hot-spot AP. Increasing the distance, almost fixed this problem. Both antennas are still on the same pole, but now grid is mounted 1 meter below omni.

The second problem seems to be sea-related, because it appears daily in the same time - around 3pm and disappears 2-3 hours later. In the mean time the behavior of my uplink is very close to what was problem no.1 - CCQ is unstable (between 50 and 100%, rarely below 50%), reflecting to constant renegotiation of Tx speed. It's effect is not so disastrous like previous case, but surfing experience for end users of my hot-spot is far from perfect.

For me it seems impossible to fix the latest problem making changes only from my side of the link, without changes or additional installations at my provider's premise. Am I right? Unfortunately, any changes there are not an option.

That's why I found a new place at the other side of the bay, where I can put my own equipment, thus having full flexibility over band - 2.4, 5.x, polarization, antennas, etc. The downside of this new spot is that it's just 15 meters above the sea level (compare to 80 meters with the current spot). Distance is little shorter - about 4.5km, 100% over the sea.

I'm ready to spent money on this new installation and changing (if needed) wireless card and/or antenna on my side of the over-sea link, if it can help me fix my problems. Plus I would like to increase the connection speed from 36 to 54 Mbps.

I need your suggestions how to stabilize (7x24) my uplink at full 54 Mbps despite usual over-the-sea problems (fresnel effect, heat-haze, tidal influence, thermal ducting to name a few), keeping in mind not-fully-solved problem with omni to grid interference.

I'll be more than happy to hear your well-grounded recommendations about the 54 Mbps, 4.5km over-the-sea link, 3.5 meters above see level from one side, 15 meters on the other:

1. 2.4 or 5.x GHz band? 5Ghz seems better in order to avoid interference with my own 2.4 APs plus 12 more WiFi's I can "hear" when scanning with my 2.4GHz grid antenna (some of them with SNR 20+ dB). But the local Mikrotik distributor's technician swears that 2.4 is better for over-the-sea links. 5.x GHz is quiet here, but 2.4 GHz is cheaper to build.
2. Vertical or Horizontal polarization? Please keep in mind that my neighboring omni antennas are horizontal.

Any recommendations about specific gear, especially cards and antennas are greatly appreciated.

Thank you in advance!

Regards,
Nikolai

P.S. Please apologize me for this very long posting full of my "English"
P.S.2. Transmission power is not a problem in my country...

What kind of tide is there in your sea? Does high tide match your low CCQ time?

Tom

What kind of tide is there in your sea? Does high tide match your low CCQ time?

Tom, AFAIK, taide of the Black Sea (which is the sea under question) is very, very low - somewhere between 3 and 8 cm. Waves go higher, but even they rarely go beyond 1 m.

IMHO the problem is somehow connected to the heat-haze, due to it's start time (3pm) - just after the hottest part of the day (2pm). The highest temperature exceeds 35 degrees Celsius just few days per summer. Usually it's around 30.

Thank you.

Is your antenna direction pointed generally in the direction of the sun during the bad signal times? Solar reflections off the water causes some headaches for me occasionally. You might want to try changing the polarization on one of the antennas also.

Is your antenna direction pointed generally in the direction of the sun during the bad signal times? Solar reflections off the water causes some headaches for me occasionally.

Hm... depends. It's on the back of my antenna, thus it should be in front of the remote antenna. If this is the case, what should I do?

You might want to try changing the polarization on one of the antennas also.

Changing the polarization is not an option. My uplink is connected to my upstream provider, whose AP is used by other customers.
The polarization of the hot-spot omni antenna obviously cannot be changed either, due to the nature of omni antennas.

After receiving here all the recommendations about my possible new remote point, where I'll control all the equipment, I can change the polarization of the uplink whenever I like.

Thank you.
Nikolai

I've had a customer with a 4.5km link over sea water running for a couple of years now.

1> Go Horizontal.. not that it makes much of a difference, but H-pol tends to work better over water
2> Go 5Ghz.. smaller fresnel, less reflections
3> Don't use MT for backhaul... MT doesn't support antenna diversity, which improves signal stability when links are over varying terrain (eg. sea). I can't say which product to use, but until the great guys at MT convince themselves that diversity DOES in fact make a huge difference, you should consider opting for a product which has antenna diversity and wiring the MT hotspot via Ethernet.

3.5> Not sure if 11n uses diversity as we know it, but it does use 2 antenna, so you could try changing your wireless card to 11n.

Hi ...

Over-sea propagation have some tricks. If the day is calm, no winds, you can expect a kind of reflecting layer around 40m ASL and lots of multipath. As the wind speed increase or sea became rough this layer spread a bit but now between 5 ... 15m ASL instead. Less reflections then. This is called oceanic evaporation duct.

So you can use antenna heights to use this duct (e.g from 1 to 5 meter ASL both sides) or much higher than the 40m ASL line.

On 2.4GHz you can expect fades from 5dB @ 2.5m sea waves to 7dB @ 1.2m sea waves. On 5.8GHz the 7dB value occur with 0.8m sea wave height. Calm than that <> deeper the notches.

When inside the duct, you can expect some unusual interferences sometimes from stations that are tens of miles away (even NLOS). Never one antena at - let say - 55m and another at 5m. Both inside the duct or both outside.

Circular polarization - if avialable - is the better, followed by horizontal polarization. Do not use vertical 'cause errors due to Brewster angle effects may arise.

I had a paper from JPL, 900 pages, about such subject somewere. This ducts creates lots of trouble to radars, generating false echoes. Another paper from an australian company told abt a 10,5GHz link between telco backbone and a cell-phone tower on an island few miles away that interfere at another instalation 45 miles ahead 'cause this duct act as a natural waveguide.

May be 2 parallel links, with antennas spaced (spatial diversity), bridged both sides handle that. Or a 2.4 + 5.8 link, bridged interfaces.

Anyway, on a terrestrial 2.4GHz PMP scenario when I locked the AP and clients speed to a single value (12Mbps - OFDM), G mode, long preamble, some signals that came through dense vegetation (tropical weather) CCQ jump from 20 ... 40 to 95 ... 100 even under rain showers or the dense overnight fog (leaves are vy wet, sometimes we think its raining but is simply the fog => leaves => drops). RSL varies from -60 to -85dBm with CCQ above 95. I will experiment on 5.8 with turbo-channels too. Wider the channel less the chance of narrow band fades affect the OFDM signal.

Regards

When Marcus talks... the world (of MikroTik) listens

Interesting as always... thanks for the explanation. Excuse my ignorance, but what is ASL? I remember using that term way back on IRC, but I don't think radio waves have any 'age, sex, location'...

On the topic, so you're saying that when a link is going over water it's better to place the antennas at 1-5m height above sea level than at 15m for example?

asl = above sea level (afaik)

Despite the naysayers, this thread http://forum.mikrotik.com/viewtopic.php?f=9&t=31644, explains how I created a fast and reliable 5 GHz link across 20 miles of pacific ocean.

On the topic, so you're saying that when a link is going over water it's better to place the antennas at 1-5m height above sea level than at 15m for example?

Sort of. If - for any reason - you can't mount antennas above 45m both sides better have both from 5 to 15m.

Amateur radio uses this ducts to establish PTP voice comms at vy long distances above 1.2GHz for instance. 100 ... 400 miles. But this setup are usefull for amateur purposes: a short time lasting two way voice contact just to 'break distance records'.

There is lots of concern on some professional markets about this ducts: the sea radar market, low orbiting satellites when sat-to-ship comms have problems (inmarsat) due to sat low elevation angle, microwave PTP links between oil rigs, etc.

And all of them reach more or less the same conclusions: both antenna inside the duct or both above the spray layer; circular pol, then horizontal, never vertical; wider channels and lots of possible throughput to have room for a stable but lower throughput; narrow & clean beamwidht antennas (e.g. high gain and few sidelobes) to avoid reflected signals (multipath) even if you need to reduce power due to high gain of antennas, etc.

The topic original setup is: one side at 5m and the other at 75m. So one antenna is inside the duct, another outside. You can expect this effect:



And - finally - there is the 'marriage' between 2.4GHz and water. So other frequencies (900M / 5.8G) will use this duct anyway but at least will not suffer extra attenuation as 2.4G due to evaporated water from sea surface.

Despite the naysayers, this thread viewtopic.php?f=9&t=31644, explains how I created a fast and reliable 5 GHz link across 20 miles of pacific ocean.

Interesting solution: frequency diversity. Both are ok: frequency or spatial. Those multipath nulls 'walks' around the spectrum at some slow speed. This drive us to wider channels and OFDM with as much carrires as possible if someone want to try to use single freq + single antenna.

I deploy a 140Mbps (ETH + 16E1) 76Km 8GHz PTP link across a big city some time ago ... not ocean ... but this huge city develops lots of thermal inversions etc. One side ant is 15m @ 1.2m diameter. Another is 125m or so, 3.6m diameter. There is a lot of narrow band nulls along the day but we bet on modem resources, able to handle two 16dB nulls along the 30MHz channel spectrum (64QAM). And it works, SLA > 99,999 (3 minutes possible outages a year).

Regards;