an FFT can be extremely helpful.
Of course, nobody said the opposite. But the information it provides is absolutely useless in music production. It is a great tool to measure atomic elements, but music is much more than a fundamental with harmonics. The transients, the dynamics envelope and the context is totally ignored.
it can be used to reverse synthesize sounds that you may be lost as to how to remake.
Yes you can use FFT to do all kinds of crazy things, but we're talking about FFT visualizing here.
I seriously doubt the OP will start developing his own resynthesis algorithms in the next days..
it can provide the fundamental of a timbre or melody so that you can know exactly what notes are played, as notes are measured in hertz...
No it can not. The FFT measures the amplitude of a previously fixed amount of "bins", usually 64, 128 or 256. These are frequency regions (like a crossover), no absolute frequencies. This practically means you won't be able to distinguish between 200Hz and 220Hz. Sadly, the ear is highly sensitive to these small variations. This is even worse at low frequencies where it is simply impossible to distinguish between 20Hz and 150Hz. Again, totally useless for music - the ear provides much more information.
Open your FFT and feed it with a 50Hz sine and a 100Hz Sine, and post a screen-shot
it can pinpoint if an instrument has outright phase or intentional detune (since it is a subtle difference at some levels) and how much, it can show you if a sound has unison.
No, no and no. Classic FFT visualizers don't show any phase information. You cannot pick up the detune because of the inaccuracy (fixed frequency bins/grid) described above. Finally, you won't be able to recognize an unison in an fft when analysing a piece of music. Most of all because an FFT is too inaccurate.
it can show harmonics and how few and far between they are, and to an educated eye, that means showing visually what waveform it is.
It can show the harmonics of a single tone - which is quite useful for distortion measurement, but absolutely meaningless for a dynamic and usually very complex audio signal ("music").
Also, the actual waveform is independent of the fft you see in your spectrum analyzer:
A. Because the FFT is a time average (defined by the "window size" parameter you'll find on all FFT tools).
B. Because classic FFT visualisations don't show the phase, and thus, an unlimited amount of waveforms can have the exact same graphical representation in an FFT.
So, no it doesn't tell you how the waveform looks like.
it can provide stellar feedback as to what you are actually doing to the sound when you are putting on that filter, when you are turning up that detune, when you are accidentally phasing your waveforms, when you are doubling up that A3 saw.
Nothing the ear couldn't do with MUCH more accuracy. A FFT doesn't show any kind of musical information the ear couldn't perceive.
it is my belief that it is important to understand what you are doing when you synthesize or modulate in every form possible and an fft provides a great visual reinforcement.
Yes, in a static, pre-defined environment (like educational and technical testing purpose). But not in a music production context.
it can also be used, for instance of drum programming, to make sure that you are making a well-balanced kick with ample space and no phase.
So, how does the FFT of a good kick look like? how does a bad one look like?! That's a point where you'll have difficulties with your argumentation.
it can even locate muddiness if put on a master bus.
How? How does muddiness on a master bus look like? how does it look like for a bass solo, how does it look like for a drum solo? Pls explain us.
For example, you can create a track with perfect pink-noise frequency distribution and still make it sound muddy. The best example IS pink noise: It sounds muddy, but looks perfect on the FFT.
it can explain to you sound!! nothing is more important than being in tune with sound and how you are interacting with it if you are making music.
I'm sorry, but that's exactly the kind of stuff an FFT cannot do. Lol, feel free to post an fft of a production which is out of tune. How do you recognize such
a detune in the FFT?!
by the way, bass always interferes with the kick drum =) which is why you duck or compress the bass...among many other things
No it doesn't. It happens when people don't know how to make music, but this is not mandatory. Ever heard a life orchestra or big band clashing? No. Your claim is wrong.
Going further, a realtime FFT is too inaccurate in the lows, which makes it exceptionally useless for visual bass and bass-kick inspection.
frankly i don't see any sort of backing or even validity to the above statements stating the uselessness of a fourier.
Please don't mix up different things. We were clearly talking about the usefulness of the
FFT visualization in music production.
All this gets pretty clear as soon you understand what the FFT really does, especially the restriction due to the frequency bins and the time precision vs spectral accuracy trade-off.Read this:
The Fourier transform is independent of the signal under examination in that it requires the same number of operations no matter if the signal we are analyzing is one single sinusoid or something else more complicated. This is the reason why the Discrete Fourier transform is called a nonparametric transform, meaning that it is not directly helpful when an ‘intelligent’ analysis of a signal is needed (in the case where we are examining a signal that we know is a sinusoid, we would prefer just getting information about its phase, frequency and magnitude instead of a bunch of sine and cosine waves at some predefined frequencies).
We now also know that we are evaluating our input signal at a fixed frequency grid (our bins) which may have nothing to do with the actual frequencies present in our input signal. Since we choose our reference sine and cosine waves (almost) according to taste with regard to their frequency, the grid we impose on our analysis is artificial. Having said this, it is immediately clear that one will easily encounter a scenario where the measured signal’s frequencies may come to lie between the frequencies of our transform bins. Consequently, a sinusoid that has a frequency that happens to lie between two frequency ‘bins’ will not be well represented in our transform. Adjacent bins that surround the bin closest in frequency to our input wave will try to ‘correct’ the deviation in frequency and thus the energy of the input wave will be smeared over several neighbouring bins. This is also the main reason why the Fourier transform will not readily analyze a sound to return with its fundamental and harmonics (and this is also why we call the sine and cosine waves partials, and not harmonics, or overtones).
Simply speaking, without further post-processing, the DFT is little more than a bank of narrow, slightly overlapping band pass filters (‘channels’) with additional phase information for each channel. It is useful for analyzing signals, doing filtering and applying some other neat tricks (changing the pitch of a signal without changing its speed is one of them explained in a different article on DSPdimension.com), but it requires additional post processing for less generic tasks.
Stephan Bernsee (Prosonic/Dirac inventor)
The DFT “à Pied”: Mastering The Fourier Transform in One Day : The DSP Dimension
