Fixing the Stereo Phantom Center

Pano

Administrator
Paid Member
2004-10-07 6:05 am
Panama
Stereo is a marvelous illusion. Done right, it works remarkably well. The illusion of sounds coming from the center is an important part of stereo, but one that causes many arguments. With good speakers and a good room, it can be very convincing. But it isn't perfect.

One thing that has bothered me about the phantom center over the years is its darker tonal balance, compared to the sounds coming strictly from left or right. Or to put it another way, a sound coming from between two speakers often sounds darker than a sound coming from a single speaker.

You might not notice this on a lot of music, mostly because sounds don't move around in the mix and so you don't get to hear and compare them coming from two speakers or one. But with dialog it can be very noticeable. If a voice is panned from left to center to right, you'll hear a dip in tonality as the voice crosses the center. A voice played off to the side, in just one speaker, will sound brighter and clearer than a voice in the center. Of course it's not just voice, most sounds have this effect when panned. With voice it is the most noticeable.

Why is that, and can anything be done to fix it?
 

Pano

Administrator
Paid Member
2004-10-07 6:05 am
Panama
This middle dip was something I wanted to fix. Perhaps there was some clever way to EQ just the phantom center? Using a matrix and EQ it might be possible to EQ only what is common to both channels.

Figuring that someone had already worked on the problem, I went searching. After some digging Google found not only the fix, but the explanation of the cause! This paper explains it well.
http://www.sfxmachine.com/docs/FixingThePhantomCenter.pdf
 

Pano

Administrator
Paid Member
2004-10-07 6:05 am
Panama
Look at section 3.4 where an impulse response is used to fix the problem.
The dip in center tonality seems to come from comb filtering between the two speakers and our ears. If the room is very reverberant, the large number of reflections will fill in and kill the comb to a large extent. But in rooms and speakers with lower reflections, the comb filtering can be strong enough to cause the audible dip.
If we add some very short echo to both channels, but decorrilate those echoes, the comb filtering is lessened and the center tonal dip goes away. Figure 3.4 shows the impulse response that can do this.
The file attached below is simply an impulse file I made in imitation of what is illustrated. If you look at the file in a wav editor, you'll see how simple it is.
 

Attachments

  • phase shift stereo -2dB.zip
    667 bytes · Views: 231

Pano

Administrator
Paid Member
2004-10-07 6:05 am
Panama
How do yo use this file? It's meant to be used in a convolution engine of a music player.
JRiver Media Center has a convolver built in to its DSP. All you have to do is unzip the file from above, then point the convolver to it.
Foobar also now has convolution built in, so you can use that. If you don't know how to find this feature or how to use it, just ask.

The file is meant for 44.1K sample rate, so try it with normal ripped CD files. You should be able to hear the difference. Let me know if you do or don't.

After comments we can explore some audio files that should make the effect quite obvious, as well as any flaws and improvements. The file is very basic, it can be tweaked to be made better, I think.
 
I did a brief test with one song that I applied the convolution to. As I already use the convolver in JRiver for my regular EQ. I've never been successful to run 2 convolvers from within JRiver so this was the easy way for me. It did change things... I can't say anything useful yet though. I did notice off axis listening changed too.
The central voice didn't follow me like before but sounded more like out of phase way off axis. Pano, could you show a picture of the left/right impulse? Importing it as raw in Audacity gave me a view of it, but I have to gamble the settings.
Is it a PCM signed 24-bit little endian, 44100 Hz stereo file?

Ideally I'd like to combine this with my DRC correction in one file to use in my convolver. How did you create the impulse?
 
ambiophonics, try a barrier.





since playing with a barrier, I've just used 1 speaker.
Going from barrier to separated speakers, the voice was a hazy blob in the middle.
Curse you acoustics !!!!!
Even 1 speaker can sound entirely different depending on what side of the tv based on setup.

Or you can try a mono center speaker between your speakers.

Norman
 
I did a brief test with one song that I applied the convolution to. As I already use the convolver in JRiver for my regular EQ. I've never been successful to run 2 convolvers from within JRiver so this was the easy way for me. It did change things... I can't say anything useful yet though. I did notice off axis listening changed too.
The central voice didn't follow me like before but sounded more like out of phase way off axis. Pano, could you show a picture of the left/right impulse? Importing it as raw in Audacity gave me a view of it, but I have to gamble the settings.
Is it a PCM signed 24-bit little endian, 44100 Hz stereo file?

Ideally I'd like to combine this with my DRC correction in one file to use in my convolver. How did you create the impulse?

Offline convolution of left channel filter with your left channel correction filter, and right channel impulse with your right channel correction filter will work. Normalization/gain of resultant filters may be necessary.

For import into REW first convert to integer format in Audacity.
 

boldname

Member
2011-12-03 11:15 am
Stereo is a marvelous illusion. Done right, it works remarkably well. The illusion of sounds coming from the center is an important part of stereo, but one that causes many arguments. With good speakers and a good room, it can be very convincing. But it isn't perfect.

One thing that has bothered me about the phantom center over the years is its darker tonal balance, compared to the sounds coming strictly from left or right. Or to put it another way, a sound coming from between two speakers often sounds darker than a sound coming from a single speaker.

You might not notice this on a lot of music, mostly because sounds don't move around in the mix and so you don't get to hear and compare them coming from two speakers or one. But with dialog it can be very noticeable. If a voice is panned from left to center to right, you'll hear a dip in tonality was the voices crosses the center. A voice played off to the side, in just one speaker, will sound brighter and clearer than a voice in the center. Of course it's not just voice, most sounds have this effect when panned. With voice it is the most noticeable.

Why is that, and can anything be done to fix it?

The single speaker comparison is very good. It is all coherent. Espeically with coaxial single speasker as we all know. No cancellation within the direct radiated sound. So in theory with two, the immmediate resultant banding is dulling down the sound, and this can be artificially reduced with the centre speaker or even a 7:1. But then this has other penalties. This is also going to degrade the wave of sound with distortion from the banding. The single speaker is sometimes the best way to listen to something. Sakuma found this and it also makes technical sense
For many it may not matter.
 

Pano

Administrator
Paid Member
2004-10-07 6:05 am
Panama
This is what the impulse looks like. It's not clipped, I've just expanded the scale a bit. Still hard to see the very last impulse.

It's a standard wave file. Should not need any special treatment.
Also attached is the file (zipped) with an older style wave header. That might work better for you?
 

Attachments

  • phase_shuffle.png
    phase_shuffle.png
    11.7 KB · Views: 1,337
  • phase shift stereo V2.zip
    622 bytes · Views: 60

Pano

Administrator
Paid Member
2004-10-07 6:05 am
Panama
Good question. I've run across it a few times on the web, but never seen or heard one.
The manual does not seem to be online now, making it hard to know what he is doing.
But from the various descriptions, it sounds like he is doing some phase tricks, the main difference being that the Trinaural needs a center speaker, the phase shuffler does not.

I'd say the Trinaural is very different approach, tho it may use some of the same techniques. The phase shuffler eliminates the comb filtering of 2 channel mono or stereo, without the need for a center speaker. It alters the center phantom image, it does not replace it.
 

Pano

Administrator
Paid Member
2004-10-07 6:05 am
Panama
I keep wondering how you can change the center phantom image without changing the R&L channels themselves?
It's in the paper I linked. You can also look at the impulse file.
Briefly, it's a very short echo applied to left and right. The left echo is in phase, the right echo alternates phase.
This makes no change in frequency response, it sums to zero. But for 2 ears separated by a head, it does make an audible difference because it eliminates some comb filtering in the midrange.
There is a change to left and right (short echo) but it should be inaudible in either of those two channels. It's when they acoustically combine into the center that there is a difference.