I posted earlier about the wireless LAN bridge we installed from our campus to our new leased office a couple of miles away.
These radios operate at such a high frequency (80 GHz) that nearly everything — a tree, a dense flock of birds, an errant golf ball (okay, I just made that one up) — is opaque to the signal. Consequently, they suffer from signal attenuation due to rain. This is a well-known phenomenon called “rain fade” that the RF engineers have to take into account in their designs.
We’re having some trouble getting sufficient path clearance (getting the base signal high enough) to avoid a link drop during periods of intense rain. Right now the base level is approx. -44 dBm. We were expecting rain fade of approx. 16 dBm, which should make it bottom out around -63 dBm. According to Bridgewave, these radios are supposed to be able to maintain the 100 Mb/s link down to -72 dBm. Therefore, we should have approximately 9 dBm of safety margin.
But we have two problems. First, we’re seeing rain fade of 22 dBm. Second, we’re seeing the link drop when the signal goes below -65 dBm. With 6 dBm more rain fade than expected and the link dropping 7 dBm before it should, our 9 dBm of safety margin is shattered.
Here is a graph showing two times when we had brief periods of intense rain. Each time the link dropped for a few minutes and then came back up after the intense rain passed:
I gave a requirement of four 9s of availability (99.99% up time). Clearly, with the current base signal level, the amount of fade we’re experiencing, and the link drop level, we’re not going to be close to four 9s. Thankfully, we’re installing this well ahead of staff moving into the office so we have time to improve the situation. Also, I’m grateful that we’re installing in the Spring when we have rain every couple of days. Each time we make a change, we don’t have to wait too long before a real-world test presents itself.
We’re out there on the bleeding edge with this one! Please pray with us that this wireless LAN bridge story has a happy ending.