Spectrogram related questions

[Kondarivan]

Hi!

I was looking over spectral analysis tools and I am wondering about this:
If a spectrogram only contains the data for frequency amplitudes and times, the phase data is not included, so if we reconstruct the waveform from this spectrogram, all frequencies are phased back to zero (right?). But the question is: is there an audible difference between two sounds with the same harmonic content, but with different harmonic phases? I tried playing with a triangle's phases in Oatmeal, but I couldn't notice any acoustic changes whatsoever.
I suppose that some configuration in aligning the frequencies differently might result in getting different peaks than in the initial wave (which possibly leads to clipping), but does it actually change amplitude acoustically?

...and since we're at it, here's a really silly question: What is the real spectral nature of sounds? Does a spectral analysis of any signal give a finite number of superimposed sine waves of different frequencies, or is it actually areas of the spectrum that have different amplitudes? If that sounds confusing, let me give an example:
Say I synthesize a middle C as a sine. At that point, the only thing that exists in the whole spectrum is one signal at 261hz. But if I synthesize a white noise and low-cut it at 260hz and hi-cut it at 262hz, I will have a range of frequencies between 260 and 262. But between 260 and 262hz the amount of possible frequencies is infinite, so it's really weird to figure out how exactly that sound can be defined, since there is no such thing as a "frequency quantum". So what exactly is going on here?

Thanks!

[BertKoor]

the question is: is there an audible difference between two sounds with the same harmonic content, but with different harmonic phases?

This has been researched. Afaik the conclusion was that you cannot hear these phase shifts. That is, some people can tell there is a difference when comparing sounds, but they can't really put the finger on it.

What is the real spectral nature of sounds? Does a spectral analysis of any signal give a finite number of superimposed sine waves of different frequencies, or is it actually areas of the spectrum that have different amplitudes?

Spectral analysis has it's limits. For instance if you want more precision for the frequency, you're losing precision in time. That's because you're looking at it from a whole other dimension. Time domain vs frequency domain.

But suppose you could get an analyser to have infinite precision and let it analyse white noise. I suppose you'll see amplitudes of all frequencies going randomly up & down. Noise is not deterministic in nature. It has a totally random moving amplitude with no real correlation to sine waves.

Another strange phenomena is the dirac spike: just one click of one sample long. It excites all frequencies of a spectrum analyser. But since there's no repetition, there really is no frequency in the signal. Oh well...

[tony tony chopper]

But between 260 and 262hz the amount of possible frequencies is infinite, so it's really weird to figure out how exactly that sound can be defined, since there is no such thing as a "frequency quantum".

that's why you need the phases plane, which is half of the FFT's result, so half of the importance

(as for audibility of phase shifts, they totally are, phase relationship between
neighboring partials matters a lot)