frequency splitting

[dobro]

Okay, chrisharris, bless him, taught me how to use the frequency splitter in Cool Edit. I'm amazed at what I found, but I'm still not sure how to interpret it. I'd really appreciate some informed opinion on this one - there's a lot to learn here, I think.

I split an unaffected, un-EQed vocal track into eight bands:

0-50 Hz
50-100 Hz
100-250 Hz (Approximately the range of the sung notes)
250-500 Hz
500-1000 Hz
1K-3K Hz
3K-5K Hz
5K - 22K Hz

Here are the results (I'm gonna ask questions as I go along).

1 There's stuff going on in the 0-50 Hz range. Huh? I don't get it. The lowest note on the song was at 110 Hz. What's stuff doing down below 50 Hz? Does the human voice have harmonics that are lower than the sung note? Or is something in my room resonating with my voice, but two octaves lower?

2 There's stuff going on in the 50-100Hz range too. This stuff, and the stuff in the 0-50 range matches the sung material exactly in pitch, but it's unintelligible. You can't make out the words. Why's that, I wonder?

3 In the 100-250 Hz range, you can make out the words, but they're really muddy, really dark. Again, I don't get it. This is the sung range of the song, so the sound should be really clear it seems to me.

4 The 250-500 Hz range is much clearer than the previous range, but still muddy.

5 500-1K: clearer again, kinda like telephone sound.

6 1K-3K: okay, now we're talking! Of course, it's high and thin, but really CLEAR. This is where the clarity of this track resides. WTF? What's it doing way up here? Sorry to labor the same point, but I don't get it. LOL Also, this range and the 100-250 Hz range, were the biggest waveforms. More energy in these two ranges than the others. Okay, I can understand why there'd be lots of energy in the 100-250Hz range, cuz that's the range of the notes I sang, but 1-3K?

7 3K-5K: okay, very small and tinny, but perfectly intelligible.

8 5K-the top: the intelligibility is starting to disappear, but you can make it out.

Okay, some of this must be happening because of my gear, my signal chain. This track was sung through a Rode NTK and a Mindprint Envoice with nothing on it but a 50K cut and a hard limiter that probably never kicked in during the tracking. Can you explain to me what all that stuff is doing down below and up above the range of the sung notes? I understand overtones and harmonics, but I don't understand why there's *so much* energy above 250 Hz, and also why all the clarity of the song resides way up past 1K.

Now, if I haven't bored the trousers off you by this point, I'd like to talk about a comparison test I ran. I also did a frequency split on the same track after I'd EQed some bottom off it (I set a highpass filter at 95 Hz) and added some air at 8K. I also put some mild reverb on it. The results were almost identical. I can *hear* that there's less bottom on the EQed track, but the frequency splitter shows there's still stuff going on down below 50 Hz. I would have thought a filter would have just about zeroed sonic activity down there. Damn. What's going on?

 

[jrosenstein]

in any finite signal, there will be some frequency content at ALL frequencies, not just the fundamental that you were aiming for. but i'd be curious .... are you using a shockmount?

and yes, the human voice has a ton of harmonics and overtones. what you are probably hearing is the difference between the fundamental frequencies and the formant frequencies. when you sing (or speak), your voice has some base pitch associated with it. but there are also many "formant" frequencies present, which are what allow you to identify different vowel sounds at the same fundamental frequency. thus, singing the word "me" in different registers is still clearly identifyable. most of these formants are significantly higher than the fundamental.

so when you say that your higher frequency bands are "clearer", what you are really seeing is that the formant frequencies (and sibilants) that define the WORDS you are singing are actually much higher than the pitch you sing.

 

[jrosenstein]

and filters are not exact (yes, even on precious digital editors). a high-shelf is more like a high-plateau. frequencies near the cutoff are not attenuated as much as frequencies far below it.

 

[dobro]

No, I don't use a shockmount, but I track on carpet in a rather quiet room. The two lower ranges in the frequency split I did don't seem to have any rumble - everything that's audible in those two ranges is an aspect of the vocal track, but lower than the sung notes. I always thought that overtones and harmonics were *higher* than the fundamental.

And yes, I now understand that filters are nowhere near as precise and comprehensive as I'd imagined.

 

[jrosenstein]

yes, harmonics only exist above the fundamental frequency. but that does not mean that when you sing at 220 Hz, you won't see any frequency content below it. we only speak of harmonics in reference to some fundamental. with a flute or recorder, this may actually be a good approximation of the sound wave it produces. but not with a human voice.

your voice is creating a series of finite signals. a finite signal is unlike an idealized infinite signal, in that it has continuous frequency content. if you were to plot the frequency content (fourier transform) of a 50Hz sine wave, you would get one spike at 50Hz, with no energy anywhere else. plotting the frequencies of a square wave yields spikes at 50Hz, as well as a number of harmonics. in between these spikes, again, there is no energy.

if you plot the frequency content of just ONE of those square pulses, what does it look like? it may surprise you that the frequencies are not all at 50Hz. in fact, ALL frequencies are represented (well, except for a few), though the frequencies close to 50Hz are the strongest.

on a somewhat related subject, when you think about frequencies, remember that they very rarely behave linearly. most of the time they are logarithmic related. if you start at A-440hz, then an octave up would be 880, but an octave above that is 1760, not 1320. what this means is, 150hz is equally as "close" to 75hz as it is to 300hz.

 

[Raydio]

Dobro, I want to know how to use the Frequeny Splitter also in Cool Edit Pro; who ever can help me will be appreciated. Just reading this post makes it sound like an interesting plugin, LOL. You can reach me on AIM also at "lilraydio@aol.com"

Peace!

 

[dobro]

It isn't a plugin. It's a feature built into 2.0. You can learn how to use it in the Help. The trick is to left click to highlight the track, and then left click and drag across the track to double highlight the portion you want the splitter to analyze.