You're welcome!
I find the opposite to be true. The better and clearer the speakers and room, the more I notice the dip. Some of that is explained in the paper.
Better is better, for sure. The phantom image is usually more stable and distinct with better systems and rooms, but the tonal dip still remains.
I am hoping that someone, anyone, here will try this phase shuffler. I may have to spoon feed the files to get anyone to try it. 🙂
I am hoping that someone, anyone, here will try this phase shuffler. I may have to spoon feed the files to get anyone to try it. 🙂
I tried to try it, but I haven't used convolution in MC for quite a while. Lots of clicks and pops with the file loaded and convolution on........I'll need to sort that out as I'd like to look into lin phase crossovers soon.
Thanks Nate. I've had the clips and pop with convolution on small processors, like an Atom. Usually not bad on medium fast processes. But it could be something else causing it. Wifi makes my laptop pop and click. Seems to be a resource hog.
Tomorrow I will try to post a file that is normal, and one that has been convolved. That way anyone should be able to play it.
Tomorrow I will try to post a file that is normal, and one that has been convolved. That way anyone should be able to play it.
So, er, according the paper, room reflection is good to stereo. Now we're sort of adding some artificial reflections back to the signals? Did I read that right?
Yes, you did. Seems strange, I know. But the added "reflections" are pure and very brief. See the impulse file. They really only serve to decorrelate the phase, thus eliminating the comb filter.
If the source were actually right in front you, there would not be any comb filtering. We just get it when creating a central image with two speakers spread apart.
If the source were actually right in front you, there would not be any comb filtering. We just get it when creating a central image with two speakers spread apart.
OK.
What's the difference between this (digital add-on) and the natural reflections (in a properly reflective room)?
Is there any advantage to avoid reflective boundaries and add these (controlled) impulses into the signal?
BTW, these also come to mind:
http://www.diyaudio.com/forums/multi-way/121385-advantages-floor-coupled-up-firing-speakers.html
Virtual Acoustics and Audio Engineering: Optimal Source Distribution
(Ambiophonics was already brought up.)
What's the difference between this (digital add-on) and the natural reflections (in a properly reflective room)?
Is there any advantage to avoid reflective boundaries and add these (controlled) impulses into the signal?
BTW, these also come to mind:
http://www.diyaudio.com/forums/multi-way/121385-advantages-floor-coupled-up-firing-speakers.html
Virtual Acoustics and Audio Engineering: Optimal Source Distribution
(Ambiophonics was already brought up.)
Here are 2 files. Very simple. A voice counting to 12. Panned hard left, center, right. One file is dry, the other has been convolved.
Do you hear any differences between the two? If I did it right, you should.
Thanks for listening!
Do you hear any differences between the two? If I did it right, you should.
Thanks for listening!
I see that you put up a couple of songs. Did not try those yet.
I tried your new V2 version on 4 different songs, much cleaner than the first version I'd say. I love what it does for central voices... It seems to work very well.
But... it might be my particular setup but the sides move back to my speaker ground plane where as before they were way past the side speakers.
My speakers are the kings of comb filtering of coarse, but with the DRC room correction the sides were more clear than the mid voices. Generally following the description of your case here. If I could find a way to only process the mid part of the audio with this trick I'd be very happy. My center image seems more clear to me, less dark. But I kind of like the engulfment the expanding sides give me (enhanced with a bit of mid-side processing in my case, using Voxengo's MSED).
I'm probably not a standard case though.
I tried your new V2 version on 4 different songs, much cleaner than the first version I'd say. I love what it does for central voices... It seems to work very well.
But... it might be my particular setup but the sides move back to my speaker ground plane where as before they were way past the side speakers.
My speakers are the kings of comb filtering of coarse, but with the DRC room correction the sides were more clear than the mid voices. Generally following the description of your case here. If I could find a way to only process the mid part of the audio with this trick I'd be very happy. My center image seems more clear to me, less dark. But I kind of like the engulfment the expanding sides give me (enhanced with a bit of mid-side processing in my case, using Voxengo's MSED).
I'm probably not a standard case though.
Last edited:
Thanks for the report! Interesting that you sense the sides moving back. I wonder if I can fix that? Below is the impulse but with the echo part high passed at 650Hz. It sounds different to me, let us know if it sounds different to you. It "should" apply the effect only above 650Hz
Can acoustic recording itself have the same problem. Maybe panned bursts of pink/white noise could be useful?
//
//
Bizarre ... I posted about the voice counting files, it went up, I noted a wrong word and did an edit, and that killed the post! Did I somehow invoke Delete? Anyway, I got almost zero perceptible difference, perhaps because of the laptop, with upward firing speaker; if anything the Normal sounded "better", the Convolved seemed to have lost a touch of resonance, had a very slightly deader sound ...
I think in a way the early reflections (that's how they look to me) work to enhance the central intelligibility for me. But it makes the sides collapse to speaker locations. Makes sense though. I did what I could to get early reflections down, by damping panels, reduce diffraction as much as I could and Room processing. And now adding these synthesized early reflections takes some of that work away again. And as it is the same on every part of the song it get's noticed as a fixed boundary close to the speakers?
I'll try the high passed one later on, but I'm still trying to come up with some way to only affect the mid shared part of the left and right signal. I do like what it does there.
The mid side processing I do is loosely based on the article: BSG qøl Signal Completion Stage | Stereophile.com
By boosting the sides just a little, you create an anti phase copy at lower volume in the opposite channel. Sort of acting like cross talk cancelation, but without the delay. Also known as a Blumlein Shuffle. On my speakers (line array) it does work very well to make the sides more pronounced. Without the mid side processing the positions of the side sounds is the same but with it, the sides come alive, sound more real.
I started with JRiver's Surround Field Effect. But that is a fixed algorithm and it's only loosely based on the Blumlein Shuffle principle according to it's author. The author's of JRiver didn't want to tell me what it does exactly, and I can't blame them. A mid side plugin does something similar, yet a bit different. Too much and none of those enhancements sound good.
But what I'm missing is to be able to "cure" the phantom center. The trick in this thread seems to be able to do that, if it was/is possible to leave the sides alone.
I also tried some cross talk cancelation routines but they suffer the same draw back if you keep the speakers in a traditional position (which I want). The sides work better but the center is falling behind in tonality. I notice in my Car I have a very good center voice. Probably because of early reflections (there's ton's of them in a car) filling in the bits and pieces?
I'll try the high passed one later on, but I'm still trying to come up with some way to only affect the mid shared part of the left and right signal. I do like what it does there.
The mid side processing I do is loosely based on the article: BSG qøl Signal Completion Stage | Stereophile.com
By boosting the sides just a little, you create an anti phase copy at lower volume in the opposite channel. Sort of acting like cross talk cancelation, but without the delay. Also known as a Blumlein Shuffle. On my speakers (line array) it does work very well to make the sides more pronounced. Without the mid side processing the positions of the side sounds is the same but with it, the sides come alive, sound more real.
I started with JRiver's Surround Field Effect. But that is a fixed algorithm and it's only loosely based on the Blumlein Shuffle principle according to it's author. The author's of JRiver didn't want to tell me what it does exactly, and I can't blame them. A mid side plugin does something similar, yet a bit different. Too much and none of those enhancements sound good.
But what I'm missing is to be able to "cure" the phantom center. The trick in this thread seems to be able to do that, if it was/is possible to leave the sides alone.
I also tried some cross talk cancelation routines but they suffer the same draw back if you keep the speakers in a traditional position (which I want). The sides work better but the center is falling behind in tonality. I notice in my Car I have a very good center voice. Probably because of early reflections (there's ton's of them in a car) filling in the bits and pieces?
Which is odd, because it actually has more resonance. It sounds odd on headphones, that's for sure.
The high passed impulse file should only affect frequencies above 650Hz. That's an octave lower than the filter used in the paper, but it seemed like a good start.
I'll have to do some analyzing to find if it is really only active above 650.
Thanks for trying it out!
Going away for a while and will not be able to audition before that. I'll make sure to revisit this thread when I get back.
Hope some others bite too and give their opinion on it.
Thanks for a fun thread! Always refreshing to look at things in another way.
I did manage to look at both impulses in REW to compare:
Judging from what we see here I expect the new one to have a better balance overall. 1=left, 2=right, 3=left, 4=right.
Attachments
Last edited:
I played the two full bandwidth counting tracks and have to say any differences were bordering on being imperceptible.
I have Linkwitz Lx521 dipole speakers with diffusers behind them, so there is quite a lot of decorrelated energy in the room to fill in the interaural suckout.
Thanks for being interested in a subject that I have been trying to come up with an original solution for. Maybe a stereo to MS side channel where we have a 2kHz odd bandpass filtered version of the M signal feeding a speaker/speakers pointing up at the ceiling.
Keith
I have Linkwitz Lx521 dipole speakers with diffusers behind them, so there is quite a lot of decorrelated energy in the room to fill in the interaural suckout.
Thanks for being interested in a subject that I have been trying to come up with an original solution for. Maybe a stereo to MS side channel where we have a 2kHz odd bandpass filtered version of the M signal feeding a speaker/speakers pointing up at the ceiling.
Keith