My rule of thumb for FSK is: "Bandwidth is greater than Shift." On the surface, that may seem quite simplistic :) But, the deeper one looks, the more one finds that this is just about as far as one can go in making a universal statement of truth in the relationship between bandwidth and shift in real world transmitted FSK.
Mathematical computation of "How much greater?" requires knowledge of many variables, as well as what type of bandwidth and how that bandwidth is to be described. There are many types of bandwidth definitions, for different applications. It is possible to make a very simplistic definition of bandwidth, without variables, but it probably won't be an accurate definition of real world signals emitted by real world transmitters. In actual practice, making good spectrum analysis measurements of some types of FSK signals can also be somewhat difficult. The instrumentation and adjustment of the test gear greatly affects the measurement. Here is an example of a spectrum analyzer measurement on an 8FSK waveform that I made, using 30Hz resolution bandwidth, peak hold, for 240 seconds: http://hflink.com/technical/ As most digital mode operators know quite well, the adjustment of audio and transmitter is a huge variable in the bandwidth of a digital signal. How many hams have a spectrum analyzer with accurate 30Hz resolution bandwidth? As DSP receivers have become widely available, it is possible to use them for spectrum analysis, and this can control some of the instrumentation variables more closely than with traditional spectrum analyzer methods. Still, operator knowledge of how to get good calibration and use reference standards is quite important to derive useful measurements. Bonnie KQ6XA
