Yes, sort of. Generally, you get relative anplitudes, often
on a logarithmic scale. Because the data is sampled and many
things like windowing can affect the results, it's not the
kind of thing one uses for talking absolutes. The dataset to
dataset variation is enough to discourage that. Still very
usefull, but more indicator than meter accuracy. The heavy
duty math types can feel free to dispute my opinion.
This is one of those areas where not only are there no simple answers, but there are really no simple questions.
What an FFT does is it splits a signal up into different frequency "buckets". The time element is gone, and replaced by a frequency element.
*However*, when you look at that FFT data, it's not always what you would expect to see based on simple (or perhaps simplistic) theory. You also find that there are restrictions on sampling times, sample sizes, windowing, and other factors which limit what you can do in practice.
FFTs are a nice tool in the right circumstances, but you really need to know what you are doing from a theoretical point of view before just cramming some sample data in one end and cranking numbers out the other. Then again, a lot of signal processing is like that.
