Adding concrete walls to simulate the shelves/aisle is a good idea. That was helpful. VisualRF still does not reliably predict correct AP placements, even with the walls drawn in. I got an eval copy of another modeling platform and it does a slightly different job, and does account for the walls, but one has to tweak it a bit to arrive at even roughly similar AP counts as VisualRF resolves (still predicts over twice as many APs as VisualRF for the same environment with the same variables.)
The design guides do a good job of offering hints for avoiding some common pitfalls, choosing the right APs/Antennas, etc., but no real insight into AP placement.
What I got from the guide is:
Assuming a 20 db loss, coverage radius is typically 60-80 meters best case.
Coverage at the top of a 10 meter shelf will be lost likely prior to the 4th row over.
Coverage at the bottom of a 10 meter high shelf will be lost likely prior to the 2nd row over.
Use a lower gain downtilt Omni to limit AP-AP Arm conflicts causing tx power reduction resulting in reduced RF coverage at floor level.
Use a lower gain downtilt Omni can limit ground coverage range to a predicable area, but without a site survey there is no way to predict what that predictable area will be.
Basically go WAG it, and maybe over-saturate APs, and just let Arm work out the problems the lack of design fundamentals caused.
Or:
|-------------------------------------------------------------------------------|
| AP AP AP |
| || || || |
|AP || || AP || AP |
| || || || |
| || || || |
| || AP|| || AP AP |
| || || || |
| || || || |
|AP || || AP || AP |
| || || || |
| AP AP AP |
|-------------------------------------------------------------------------------|