posted 07 June 2001 01:14            

phys_panning.kym

The attached soundfile contains two Sounds.

In the first, I produce a mono blend of inputs and then enable you to pan the sound anywhere you like +/- 30 deg from dead ahead. This sound uses both a simple minded HRTF (head related transfer function) and delays to plant the sound realistically in the stereo field. The HRTF and delays are not frequency selective. The 30 degree variation was chosen to match the placement of stereo speakers arranged in an equilateral triangle with the head.

Using this kind of panning creates more realistic soundfields than simple level controls as found on most mixers. It also spares the ears from the pressure differential when listening through headphones.

In the second Sound, I was attempting to apply the same simple physics to the apparent placement of a stereo blend for headphone listening. Instead of panning, this Sound allows you to control the apparent separation of left and right channels. You can select all the way from a pure mono mix, to the original stereo signal, by varying both the HRTF and the !Spread controls.

The HRTF is based on a pair of ellipses centered in the head with each ear having a preference for its side. The HRTF is controlled by the !Eps parameter which is nothing more than the eccentricity of those ellipses.

When !Eps is zero, there is no apparent level shift as the sound pans around. But there is still a delay to take advantage of the precedence effect.

When it is 1, the level shift is extreme -- just like that provided by the pan control on most mixers. But again, there is a delay that helps the ears.

I usually run the !Eps around 0.3 but I have no idea if this is anywhere close to reality. That produces a level shift of about 5 dB between when the sound is on the same side as the ear, to the opposite side.

The stereo spreader at a !Spread of 0.33 selects a simulation of speakers placed in an equilateral triangle with the head.

When !Spread = 1 and !Eps = 1 you get the original stereo signal out (Left and right are 180 degrees apart and no feed from one side to the other). When !Spread is 0 and !Eps is 0 you get a pure monaural sound.

I have no idea how close this is to reality but it sounds awesome with the !Spread = 0.33 and !Eps = 0.3. I have to say I don't think it sounds like speakers in an equilateral relationship to the head. It sounds even better than that to me through headphones.

Enjoy!

- DM

[After listening carefully to a recording of a harp concerto, I swear it sounds more 3-dimensional to me with the stereo spreader. I originally developed these two sounds for application in the music studio where one instrument at a time was playing, possibly synthesized in stereo.

But it works just fine on recorded material too. And I swear it feels like I could just reach out and touch each instrument. Instead of sounds layering on top of each other and interfering or obliterating each other, they seem to physically spread out and give each other room to be heard.

I think, however, that this approach is "better than reality" because we are performing a coherent mixing of sound, whereas the ear normally performs an incoherent sum in the brain. The ears are not phase sensitive detectors, while mixing waveforms on a computer certainly is phase sensitive. We get contructive and destructive interference on the computer that simply doesn't happen with our ears.

Oh well... (but it sure sounds awesome!)]

[...er, yes I know... that last statement about phase coherence is pure rubbish! Else we couldn't experience noise cancellation, and other neat effects... sorry! ]

[This message has been edited by David McClain (edited 07 June 2001).]

posted 07 June 2001 11:49            

phys_panning.kym

With this new scaling of the amplitudes, an HRTF epsilon (eccentricity) of 0.66 gives roughly 3 dB variation in amplitude as a sound slides from one extreme to the other, with separation angles +/- 30 deg from straight ahead.

- DM

posted 14 August 2001 14:09              

I quite fancy the hrtf files on this thread but the server has swallowed them ...any chance of uploading them again,

best wishes,

jachin

quote:

---

Originally posted by David McClain:
Hi,

The attached soundfile contains two Sounds.

In the first, I produce a mono blend of inputs and then enable you to pan the sound anywhere you like +/- 30 deg from dead ahead. This sound uses both a simple minded HRTF (head related transfer function) and delays to plant the sound realistically in the stereo field. The HRTF and delays are not frequency selective. The 30 degree variation was chosen to match the placement of stereo speakers arranged in an equilateral triangle with the head.

Using this kind of panning creates more realistic soundfields than simple level controls as found on most mixers. It also spares the ears from the pressure differential when listening through headphones.

In the second Sound, I was attempting to apply the same simple physics to the apparent placement of a stereo blend for headphone listening. Instead of panning, this Sound allows you to control the apparent separation of left and right channels. You can select all the way from a pure mono mix, to the original stereo signal, by varying both the HRTF and the !Spread controls.

The HRTF is based on a pair of ellipses centered in the head with each ear having a preference for its side. The HRTF is controlled by the !Eps parameter which is nothing more than the eccentricity of those ellipses.

When !Eps is zero, there is no apparent level shift as the sound pans around. But there is still a delay to take advantage of the precedence effect.

When it is 1, the level shift is extreme -- just like that provided by the pan control on most mixers. But again, there is a delay that helps the ears.

I usually run the !Eps around 0.3 but I have no idea if this is anywhere close to reality. That produces a level shift of about 5 dB between when the sound is on the same side as the ear, to the opposite side.

The stereo spreader at a !Spread of 0.33 selects a simulation of speakers placed in an equilateral triangle with the head.

When !Spread = 1 and !Eps = 1 you get the original stereo signal out (Left and right are 180 degrees apart and no feed from one side to the other). When !Spread is 0 and !Eps is 0 you get a pure monaural sound.

I have no idea how close this is to reality but it sounds awesome with the !Spread = 0.33 and !Eps = 0.3. I have to say I don't think it sounds like speakers in an equilateral relationship to the head. It sounds even better than that to me through headphones.

Enjoy!

- DM

[After listening carefully to a recording of a harp concerto, I swear it sounds more 3-dimensional to me with the stereo spreader. I originally developed these two sounds for application in the music studio where one instrument at a time was playing, possibly synthesized in stereo.

But it works just fine on recorded material too. And I swear it feels like I could just reach out and touch each instrument. Instead of sounds layering on top of each other and interfering or obliterating each other, they seem to physically spread out and give each other room to be heard.

I think, however, that this approach is "better than reality" because we are performing a coherent mixing of sound, whereas the ear normally performs an incoherent sum in the brain. The ears are not phase sensitive detectors, while mixing waveforms on a computer certainly is phase sensitive. We get contructive and destructive interference on the computer that simply doesn't happen with our ears.

Oh well... (but it sure sounds awesome!)]

[...er, yes I know... that last statement about phase coherence is pure rubbish! Else we couldn't experience noise cancellation, and other neat effects... sorry! ]

[This message has been edited by David McClain (edited 07 June 2001).]

[This message has been edited by David McClain (edited 07 June 2001).]

---