pitch dependent on volume?

Topic: pitch dependent on volume? (Read 5722 times) previous topic - next topic

0 Members and 1 Guest are viewing this topic.

pitch dependent on volume?

2003-02-21 18:43:24

I listened to a 300 test tone today. Somebody on rec.audio.opinion had recorded it from vinyl and compressed it to mp3 to demonstrate that his turntable had no audible speed stability problems.

The interesting thing about it is that when I listen to it at different volumes, the pitch (subjectively) changes. JohnV verified this effect. The pitch increases as the volume decreases.

I'm wondering why this is. Anybody know?

You can download the sample from the link below. It's a 10 second sample compressed using --alt-preset standard.

http://ff123.net/export/freq1.mp3

pitch dependent on volume?

Reply #1 – 2003-02-21 19:04:56

I'm supposing that this effect happens because masking of the hearing masks the tone the more the volume is increased, thus the pitch seems to be changing.

pitch dependent on volume?

Reply #2 – 2003-02-21 19:11:09

The tone appears to me to have the same pitch, regardless of the volume.

But what I really find interesting is that Winamp 2.81 shows a bit rate for the sample varying between about 128 and 192, with an average of 173; what possible criteria could a codec use to vary the bitrate across a constant-frequency sample? If the sample is the same from beginning to end, shouldn't any codec (even a VBR) assign a fixed bitrate, except perhaps for a few frames at the beginning and end?

pitch dependent on volume?

Reply #3 – 2003-02-21 19:13:50

I'm not sure if I understand the question (have to download the file), but maybe you describe something related to the Doppler-effect: As the volume decreases you subjectively expect the pitch to go down.

Just an idea...

pitch dependent on volume?

Reply #4 – 2003-02-21 19:21:57

Wow, wouldn't of believed it if i hadn't heard it. I hear the pitch change effect too.

pitch dependent on volume?

Reply #5 – 2003-02-21 19:49:24

I'd say it's because the basilar membrane of the inner ear, which does the "frequency analysis" by resonating to specific frequencies at specific points, vibrates in a linear way if the volume is low. On high volumes the vibration becomes non-linear for mechanical reasons, so the point of greatest resonance is shifted which means that the frequency "reported" to the brain now will be the frequency corresponding to the new "greatest-resonance-point".

pitch dependent on volume?

Reply #6 – 2003-02-21 20:54:22

I think I got it, possibly... a natural oscillation will make a string move more at louder amplitudes. Therefore it stretches it more, thus raising its pitch. Same goes for bells and anything else physical. And we compensate for this, right?.. So with a sine wave, which does not do this because the speaker is not actually a resonant system and because the electrical oscillator is not using massive particles to make its sound, we don't hear that rise with loudness. But we do stil hear our internal compensation. Result: net lowering of percieved pitch with raised amplitude.

Edit: This sort of thing is very common in synthesizer programming, especially the pseudo-doppler effects in filters and in modulation indexes in FM synthesis, which, oddly, make very similar phase shifts with sweeps across the frequency band, but with opposite effects to each other on the percieved pitch. Not quite what this thread is about, but still interesting I hope.

pitch dependent on volume?

Reply #7 – 2003-02-21 21:24:47

Quote

I think I got it, possibly... a natural oscillation will make a string move more at louder amplitudes. Therefore it stretches it more, thus raising its pitch. Same goes for bells and anything else physical.

So far this is what I was talking about in my post above. But it's not the brain "knowing" that loud sounds of the same source have a higher frequency and trying to compensate.

It's the inner ear that behaves as you describe. Let me give an example: Let's say we have P400 = the point of the basilar membrane that has a resonance frequency of 400Hz on low volume (=low amplitude what means that the vibration is linear (or harmonic (?) I don't know in English): The distance you move it from its "rest position" is propotional to the force in the opposite direction.) and P380 = ... the same ...

On high volume there's non-linear vibration: Starting from a certain distance from "rest position" the force grows faster then the distance. The result is a higher frequency (raising pitch, as said by The_Doctor). So at high volume P400 has a resonance frequency let's say of 420Hz while P380's r.f. is 400Hz. This means that the inner hair cell (= ear's neuronal sensor) that is located at P380 and therefore responsible to report a 380kHz tone to the brain, reports a 400Hz loud tone. This is what leads to decreasing pitch on increasing volume.

pitch dependent on volume?

Reply #8 – 2003-02-21 21:31:51

Quote

The tone appears to me to have the same pitch, regardless of the volume.

But what I really find interesting is that Winamp 2.81 shows a bit rate for the sample varying between about 128 and 192, with an average of 173; what possible criteria could a codec use to vary the bitrate across a constant-frequency sample? If the sample is the same from beginning to end, shouldn't any codec (even a VBR) assign a fixed bitrate, except perhaps for a few frames at the beginning and end?

That would be because it's a sine reproduced from vinyl, and the noise is varied enough to make a different workload on the VBR encoder used to compress it at different moments in the sound.

pitch dependent on volume?

Reply #9 – 2003-02-21 21:42:12

Tigre, I like that idea a lot, but I don't agree with it. I admit it will make a lot of difference at truly large sound levels, but it would be very nonlinear, as you say.
And one thing you'd not hear if it was purely physical is the same shift beteen low and full level on a fader whose final output is mediated by a main volume control. But you do hear it even then, so it's relative to the range on the fader, not the absolute volume heard at the ear. What is more, that apparent change is very linear. The non-linearity you mention would make an equally non-linear shift in which hair-sensors are going to respond, and this would acount for the roughening perceived in a sine at very loud levels, but that, also is not heard at low levels, even though this shift effect is.

This would also be why mrosscook could hear it as if there was no pitch change. I can't be sure, but my guess is he (?) listened to it at various momentarily fixed levels rather that sweeping the level steadily, but I'm not sure. It's the change in level rather than any momentary value that cues us into compensating for the pitch change in a natural oscillator that is losing or gaining energy.

pitch dependent on volume?

Reply #10 – 2003-02-21 22:19:57

Quote

And one thing you'd not hear if it was purely physical is the same shift beteen low and full level on a fader whose final output is mediated by a main volume control.

I'm not sure if I understand correctly: Do you mean the perceived pitch shift of a 20-60dB and a 50-90dB difference is the same?

Quote

But you do hear it even then, so it's relative to the range on the fader, not the absolute volume heard at the ear.

Well, for me and my hearing on low volumes the pitch doesn't change, but the more the volume raises to a level where it becomes anoying, the pitch decreases noticeably (for me I'd say it's about a half tone (or is it called semitone) from lowest audible to highest standable volume). I think it would be hard to find a way of abx testing this - a true challenge.

Quote

What is more, that apparent change is very linear. The non-linearity you mention would make an equally non-linear shift in which hair-sensors are going to respond, and this would acount for the roughening perceived in a sine at very loud levels, but that, also is not heard at low levels, even though this shift effect is.

I think you confuse linear: amplitude/force=constant; precondition for harmonic oscillation with linear: volume difference/pitch difference=constant. I was talking about the first "linear". Sorry If I got you wrong.

Quote

This would also be why mrosscook could hear it as if there was no pitch change. I can't be sure, but my guess is he (?) listened to it at various momentarily fixed levels rather that sweeping the level steadily, but I'm not sure. It's the change in level rather than any momentary value that cues us into compensating for the pitch change in a natural oscillator that is losing or gaining energy.

If your assumptions made here were correct, this would be a good argument, but another conlusion would be that it's just harder to notice small differences on fixed levels.

pitch dependent on volume?

Reply #11 – 2003-02-21 23:28:29

I was wrong about the change being constant for a given change on the fader. That sample wasn't long enough for me to try much, but I just tried it on a synth, snd you're right about that deeper change with louder real level. It still has a sort of linearity about it though, just like multiplying rather than straight addition. About that non-linearity, I meant that as the ear gets stressed a more chaotic response happens, like the roughening heard in the sweep of a siren at close range. I was just thinking that this other thing is far too smooth and neatly arithmetical to be explained by that mechanism.

pitch dependent on volume?

Reply #12 – 2003-02-22 00:24:34

The Doctor is correct that I was listening to the tone at a succession of fixed levels rather than doing a volume sweep. When I do sweep the volume, I think that I can hear an effect similar to what you are describing, but I don't know that I would call it a change in pitch; the sound has a deeper "color" to it at higher volumes, if I can use such a term. This is probably caused by nonlinearities somewhere, but whether they are in my ears or my speakers, who knows?

If the effect is not related to the noise on the vinyl recording, it should be seen for any constant-frequency sample, and there are lots of those around -- does anybody hear an apparent change in pitch on a different sample with a change in volume?

I'm still troubled by the variable bitrate -- the mp3x frame analyzer shows variations in bitrate between 160, 192, and 224 kbps; this seems to me like a very large variation in bitrate for very small amounts of noise. Maybe there is an mp3 guru lurking here who could give an opinion whether this is reasonable behavior or not?

pitch dependent on volume?

Reply #13 – 2003-02-22 01:32:19

I can hear this effect pretty well when I put on/off my headphones. It has nothing to do with Doppler effect anyway, since Doppler effect affects only moving sources.

Edit : I've noticed another interesting effect, it is much more easy for me to distinguish wow and flutter at very low listening levels.
In a given night club, the sound was 120 db, and I even lost the track of the melody, so confused was the pitch of each note. It sounded like a garbled MP3.

pitch dependent on volume?

Reply #14 – 2003-02-22 01:37:06

In order to ABX it, one should generate close tones, closer than the difference heard because of the volume.
In the test, the subject would listen to the two tones in a random order and say if the second one is higher or lower in pitch than the first one.
The tones played at the same level would validate the ability of the subject to distinguish between them at both low and high volumes.

Now the lower pitch tone is played at the lower volume, and the high pitched one at the higher volume.
If the effect is real, the subject should mistake all the time the low pitched tone for the high pitched one.

pitch dependent on volume?

Reply #15 – 2003-02-22 09:24:18

I've just read that above 1000Hz the effect is the other way round: Increasing the volume increases the pitch. I tested on a 2000Hz sine tone, but can't say if it's like this for me. I just can say my ears don't like this kind of test.

pitch dependent on volume?

Reply #16 – 2003-02-22 16:26:59

I agree it is different above a certain pitch. I just tried it with the sine wave made by an FM synthesizer, heard through loudspeakers, and I found that at high frequencies the pitch appeared to rise with increasing volume, and that my 'break even point' was at 3.6 KHz.

pitch dependent on volume?

Reply #17 – 2003-02-22 17:06:44

The following PDF paper has a lot of nice graphs, including one in the second part which shows how pitch varies with amplitude.

Part 1
http://www.acoustics.hut.fi/~mak/KA2001/KA6a.pdf

Part 2
http://www.acoustics.hut.fi/~mak/KA2001/KA6b.pdf

So apparently, this is a well-known an studied phenomenon. But it sure took me by surprise.

ff123

pitch dependent on volume?

Reply #18 – 2003-02-22 22:14:03

This is something I've noticed on my own many times just listening to music while out and about. It has been painfully clear that for some reason as I turned up the volume the overal pitch of a song would seem to change! And I never knew if I was just imagining this or if this was really happening. So thanks! Now I know I'm not alone

/me bookmarks this thread...

pitch dependent on volume?

Reply #19 – 2003-02-22 22:54:24

This is really interesting stuff! Never noticed it before! The question I have about this is...

Can this "effecT" be used to further enhance lossy audio compression? AFAIK, MP3 encoders separate the spectrum into different frequency bands, and then it just stores the data in the loudest band. I mean, just in a very simplified sort of way. Well, if the perceived pitch changes with amplitude, would it change what frequency band it should be encoded in depending on it's amplitude?

Say MP3 defines one section from 300-400Hz and another from 400-500Hz. A 390Hz signal is present, so normaly it would be encoded in it's range. But it is a loud signal... so should it really belong on the range above it? Hmm... kinda hard for me to explain since I don't know the proper terms etc. I guess we would need some sort of equation for calculating the "percieved" pitch of a signal... maybe it's in those PDFs... havent had a chance to read them yet.

Again... real insteresting...

pitch dependent on volume?

Reply #20 – 2003-02-23 00:36:35

@LoKi128:
Some points why your idea probably doesn't work:

1. You don't know at what level the encoded file is played back later, so you don't know the pitch shift

2. The encoded file should sound as close to the original as possible. If the original contains a frequency of e.g. 400Hz, it's not masked and it's loud, so it will be perceived, when played back loud too, in a way that its pitch sounds the same as a low volume 380Hz tone. In order to be close to the original, the lossy compressed copy must contain (almost) the same frequency (400Hz) at (almost) the same level, so that it's perceived as 380Hz too when played back loud. Otherwise there'd be the danger that the lossy compressed copy sounds better than the original©.

pitch dependent on volume?

Reply #21 – 2003-02-23 16:29:00

Quote

Otherwise there'd be the danger that the lossy compressed copy sounds better than the original©.

I love that.

Actually, (slightly offtopic) I've considered this. I've used the Hyperprism harmonic enhancer to recreat harmonics after FFT NR on radio recordings. Purists will scoff, but that won't alter the fact that apart from the loss of dynamic range in FM transmissions I have got results that are very close to the original CD's (Notably in a copy of 'Evening Land' from the Visible World CD by Mari Boine and Jan Garbarek. I tested this after recording the song from BBC Radio 3's Late Junction, and having bought the CD some months later..)

My point is that if the HF is reduced before compression, and a plugin in a player uses this harmonic regeneration, all in a controlled manner like that used for EQ and level by Dolby B encoding, there might by a way to get ferocious reductions in file size while still having beautifully sharp HF in playback. And if the original was a tad muted, the player could have a user-tweak so that the result could indeed sound 'better' than the original.

pitch dependent on volume?

Reply #22 – 2003-02-23 16:47:14

Quote

The following PDF paper has a lot of nice graphs, including one in the second part which shows how pitch varies with amplitude.

Part 1
http://www.acoustics.hut.fi/~mak/KA2001/KA6a.pdf

Part 2
http://www.acoustics.hut.fi/~mak/KA2001/KA6b.pdf

So apparently, this is a well-known an studied phenomenon. But it sure took me by surprise.

ff123

To me, it's a well known phenomenon since many years, i knew this pitch change phenomenon at high volumes and accepted it, so this didn't surprise me at all.
But, nice to see there are studies about it, thanks! This will be interesting reading.

Hey now i see that my post is kinda simillar to SNYder's .

pitch dependent on volume?

Reply #23 – 2003-02-23 17:08:54

tigre: heh true true... I guess I got overly excited about the whole pitch thing! first thing I did was start playing all my music at low volumes then at high volumes...

anyway, I read those PDFs... and there are a lot of masking phenomenae aparent in audio. I guess my new question is, do the current encoders use all of these? COULD they use all of these? Just wanna gauge how much potential for improvement do the current compressors have, or are they as good as they are going to get.

Notice