I recently had to design a number of space diversity hops for a CLEC. Not a huge deal, I've done it before, however this was in a different part of the globe then I've designed for before, which raised some interesting questions regarding the "rules of thumb" most designs rely on.
Pathloss has a number of multipath propagation reliability models that it can use and it's implementation of them are explained in the manual. However it's manual is quite barren on real information to the application of them in real work design work. This prompted me to research the Vigants-Barnett model as it tends to be the most widely used in North America and the Caribbean.
Most of what we know about the model of space diversity comes from article in the Bell System Technical Journal during the 1970's. In trying to understand this completely I was perplexed at the difficulty finding these reference article. I have located some of them and made them available as OCR'd PDF scans below.
Space diversity is costly to implement and I have worked with a number of engineers who have made statements to the effect of "in Florida any path over 10 miles needs to be full diversity", but have nothing backing these "rules of thumb" up. An excellent treatment of 6 GHz prediction calculations is in the Space-Diversity Engineering BSTJ text below. It is important to calculate the improvement factor and see if diversity will make a cost justified improvement.
Pathloss can be used to do these calculations, but I found my understanding of the results to be lacking without an through understanding of the models. Also complicating the fact was the ITU-R P.530-9 and Vigants-Barnett models predicted much different results over different paths. I suspect the Vigants-Barnett model to be more accurate as it takes a real terrain roughness calculation into account. Vigants-Barnett seams to be more conservative as well.
As a general rule to spacing of the top TX antenna to the bottom diversity antenna, it's advantageous to get as much spacing as possible. The recommendation of Trevor Manning's book "Microwave Radio Transmission Design Guide" is at least 200 wavelengths or ~30ft (10M) at 6ghz. This agrees with the recommendation of Vigants in his BSTJ article. Over water paths tend to be of different design with much more predictable reflection points, and I would suggest reading the appendix of Vigants model for his recommendations on this (and to make these paths as a high-low shot clearance permitting).
Here are the relivent links to the technical articles.Recently I attended a seminar that Harris conducts on a semiannual basis. The seminar goes in depth on microwave theory and systems design in different regions. I'd suggest that anyone seeking a better understanding of these concepts attend.