Tides and Wireless Networking or Why WiFi and Tides do not mix! Version: 2004-12-25 Author: Rob Clark (clark@freenet-antennas.com) Summary I had been battling a short 2.4 GHz (WiFi) wireless link that was exhibiting a perplexing behavior for quite some time. This short link was my one and only connection the WA FreeNet (WAFN) a hobbyist WiFi Internet-like network. The link would go for a week with almost zero losses and then start failing every day for the next week. The cause was eventually tracked down to multi-path reflections off the river between the two end points of the link. This part of the river was close to the ocean and had a significant tidal variation. Outage History Figure 1 shows the outage history for a 1 month period. There is an obvious pattern, that repeats approximately twice per month. Tides are driven by the moon cycle, and repeat twice per lunar cycle which means approximately twice per month. It is also clear that when outages start occurring, they repeat approximately daily. In general, there is one high tide, and one low tide per day. Figure 1 - Outage history, showing a cycle repeating approximately twice per month.
Outages vs Tides The chart below shows the tidal variation at the port approximately 5km downstream. The red square wave shows the periods of signal outage. A value of zero means; the link was operational. A value of 1 means; 100% ping loss over the link. The correlation is obvious. There was one day where the predicted outage did not occur. That may (??) have been a day of high winds, causing the river to be a poor reflector. Tides and Connectivity 1.4 1.2 Changed antenna 1 Tide (m) 0.8 0.6 Tide Status 0.4 0.2 0 0.000 2.000 4.000 6.000 8.000 10.000 12.000 14.000 16.000 18.000 Day Figure 2 - Tide s vs Outages At the point shown, the relatively low gain (wide beam) 12 dbi panel antenna was changed to a higher gain (narrower beam) 19 dbi grid antenna. The outages completely stopped. More about the antenna change later. Trigonometry comes to the rescue The details of the link are: Distance: Antenna elevations: Polarization: Antennas APs 1km (1000m) approx 20m above mean river level Horizontal 1) 11 dbi, 360 degree, slotted waveguide 2) 12 dbi, panel antenna (initially), then changed to a 19 dbi grid antenna. 1) 60 mw Minitar MNWAPB 2) 60 mw Minitar MNWAPB The geometry is approximately as below.
1000 m Direct Path 20 m +/- Tides Reflected Path River Figure 3 - Geometry between the two antennas From basic physics, we know that if the difference in path lengths is an integral number of wavelengths, then the two waves combine constructively; that is the signal level is twice that via the direct path only. If the difference in path lengths is 0.5, 1.5, 2.5, wavelengths, then the waves arrive out of phase, and combine destructively resulting in a much lower signal level. Using a wavelength of 121mm (2.4 GHz) and some simple trigonometry, you can demonstrate that the received signal level at either antenna varies as in Figure 4. Relative Signal strength 1.2 1 Relative Signal Strength 0.8 0.6 0.4 0.2 0 19.2 19.4 19.6 19.8 20 20.2 20.4 20.6 20.8 21 Tide Level Figure 4 - Predicted relative signal strength vs. Tide Level
The Solution The fix used was to use an antenna which is much less sensitive to the reflected signal on one end of the link. Narrow beam antennas are more highly focused and hence less sensitive to off-axis signals. As a rule; the narrower the beam, the higher the antenna gain. Antenna gain is specified in units of dbi. The plot below shows the effect of changing the 12 dbi panel antenna (a relatively broad-beam antenna) to a highly focused 19 dbi grid antenna. This grid antenna also included a low-noise feature that is designed to reject off-beam signals in urban areas that potentially have quite a few unwanted signals from other signal sources. That feature ideally suited this problem. Figure 5 Daily plot showing the effect of changing the antenna. 0% Ping Loss is a perfect link. Figure 6 Final proof. You can see that Week 51 would have normally had regular outages. But after the antenna change (early week 50) - no outages.
The Antennas The antennas mentioned in this article have these properties. Antenna 12 dbi Panel Vertical Beam Width (when deployed with Horizontal Polarisation) 75 degrees 18 dbi Grid 12 degrees More Information WA FreeNet http://wafreenet.org/ (WAFN) Antenna http://store.freenet-antennas.com/index.php?cpath=22_24_26 Specifications RRDTool http://people.ee.ethz.ch/~oetiker/webtools/rrdtool/ (Plotting software)