I have seen people doing site surveys with the ACU in different ways especially when doing the site survey at more then one data rate. Some just look at the signal to noise ratio in dB levels, while others look at the packet success rate (percent successful).
What is your experience when doing a site survey with different data rates? What leads to the best results?
To answer your specific question, the users don't care about dB levels, they care about dropped packets - so that's the final metric.
Whether you use ACU, Mini Stumbler, or a Spectrum Analyzer there are several things you need to know to get the best results.
- AP manufacturer is a big one. A Cisco AP may have 5dB more Transmit power, but it also has 3dB greater receive sensitivity at 11Mbps. That difference of 8dB means more than double the distance at the same data rate, even with a non-Cisco client. That could be good or bad - either way you need to know. For client device manufacturer, it is safest to assume a 30mW card, not 100mW to ensure that built-in modules in handhelds or laptops get good service.
- Desired overlap or non-signal areas. If you need double or triple coverage for reliability's sake that will have a big impact on your survey and design. Since the "non-overlapping" channels actually have 100% overlap in their side bands, you would ideally survey each cell with the adjacent APs in place. Similarly, if you need to avoid reaching an area, that will affect your AP placement and antenna choice, with a cascading affect on adjacent cells.
We start with both a blueprint and a client meeting to make sure we are surveying with the right goal in mind. Armed with that information and an understanding of RF propagation, I find that Mini Stumbler on an iPaq makes for a very convenient Survey tool, though you can spend more for similar results with a dedicated device from BV Systems. Confirm your results by spot-checking with ACU and watching packet success rate in ACTIVE mode (to test for problems in either direction).