For the 5.1 mixes, in the case I described, the left and center will contain an equal amount to present a queue of something in between those two channels. Eq-ing either one separately, left or center will mess up that balance.
If the content is not the same anymore in the left and center, that panned sound between left and right will exhibit the same hole as a mid/side processed stereo file gone wrong.
The biggest difference between my mid and side EQ is less than 3 dB, I'm not worried. I'm not widening the stage, and not narrowing it either. I just went for a bit of balance between bass and high frequencies, which makes wide spread backing vocals more full bodied while making the phantom center less dark.
Ah. Ok, I can see why that wouldn't work. Guess the discrete center channel is the only "clean" option in this case.
Could you post the EQ you are using on the mid/side? Are you boosting/cutting symmetrically ala Gerzon's article?
I didn't start with the Gerzon paper, I linked that one for the reference to Blumlein's work.
Earlier in this thread there was an interesting link to this:
http://www.sengpielaudio.com/FrequenzabhHoerereignisrichtung.pdf
I used that as a guide and went from there. I could post the curve but it works together with the basic EQ target I have. The 50% S curve was my first shot at this. From there it has been changed to taste (with a slight cut at 3700 Hz and ~7200 Hz in the center channel, see my thread for that)
Earlier in this thread there was an interesting link to this:
http://www.sengpielaudio.com/FrequenzabhHoerereignisrichtung.pdf
I used that as a guide and went from there. I could post the curve but it works together with the basic EQ target I have. The 50% S curve was my first shot at this. From there it has been changed to taste (with a slight cut at 3700 Hz and ~7200 Hz in the center channel, see my thread for that)
I think this entire subject (cross talk cancellation to achieve headphone style imaging) is off topic, maybe a new topic should be started? The ImmersAV technology is one related to producing content, not with playback of existing content.
I have a few problems with the ImmersAV technology, in particular:
1. reliance on generic (dummy head) HRTF, which can differ substantially from individual HRTF
2. headphones should produce flat response at the eardrum in this case (rarely the case, not optimal for normal stereo reproduction)
3. CTC loudspeaker files are encoded and distributed separately - not an efficient delivery system
4. not many binaural recordings exist; CTC is inappropriate for reproducing the majority of stereo recordings
Solutions:
1. record in ambisonic or other surround format and render to HRTF / surround speakers / CTC stereo as appropriate
2. headphone frequency and phase response should be compensated for the target (flat at eardrum)
3. see 1.. the new ITU and Dolby standards for immersive audio have a device-independent delivery format
4. enjoy stereo recordings as the producers of the music intended
or embrace "new" standards like surround sound or the upcoming immersive formatsI guess I don't like the general concepts in this thread - implement these non-standard ( though often valid) concepts in a playback system that is primarily tasked with reproducing existing stereo content. It's fun to experiment, but I think after 50 years of content being produced using a conventional (unprocessed) pair of speakers as a reference, using a non-standard listening setup can hardly be expected to improve the reproduction of the majority of recordings.
The intended audience of papers by Gerzon, Blumlein, etc is producers of stereo recordings..
When you are making these kind of changes in your playback system, I agree with Geddes that you are playing "artist" not "listener". That's cool, just realize that you are basically remastering the content and the quality of the outcome will vary substantially based on input.
I agree with some of your comments. I've been astounded at how well good headphones work with binaural recordings, so I think you're making too big a deal out of headphones not being good enough.I think this entire subject (cross talk cancellation to achieve headphone style imaging) is off topic, maybe a new topic should be started? The ImmersAV technology is one related to producing content, not with playback of existing content.
I have a few problems with the ImmersAV technology, in particular:
1. reliance on generic (dummy head) HRTF, which can differ substantially from individual HRTF
2. headphones should produce flat response at the eardrum in this case (rarely the case, not optimal for normal stereo reproduction)
3. CTC loudspeaker files are encoded and distributed separately - not an efficient delivery system
4. not many binaural recordings exist; CTC is inappropriate for reproducing the majority of stereo recordings
Solutions:
1. record in ambisonic or other surround format and render to HRTF / surround speakers / CTC stereo as appropriate
2. headphone frequency and phase response should be compensated for the target (flat at eardrum)
3. see 1.. the new ITU and Dolby standards for immersive audio have a device-independent delivery format
4. enjoy stereo recordings as the producers of the music intended
I guess I don't like the general concepts in this thread - implement these non-standard ( though often valid) concepts in a playback system that is primarily tasked with reproducing existing stereo content. It's fun to experiment, but I think after 50 years of content being produced using a conventional (unprocessed) pair of speakers as a reference, using a non-standard listening setup can hardly be expected to improve the reproduction of the majority of recordings.
The intended audience of papers by Gerzon, Blumlein, etc is producers of stereo recordings..
I once made a recording of me strumming on an acoustic guitar, with two very cheap omni condenser mics, placed on either side of a maybe 8 inch diameter clay flower pot, poorly mimicking a human head. Perhaps my first "binaural" recording. When I played the recording back through headphones, I was so startled by the rustling noise in the recording before I even started to strum the guitar, that I jumped up thinking someone was in my house. It sounded that real. I reacted before I even had time to think. It was the sound of me leaning back after pushing the record button on the little Sony stereo cassette recorder. Ever since I've been wanting to explore and learn from that. Some people seem to judge a stereo by how loud it goes, or how flat it's response is, or distortion... I got my biggest kick from good wideband imaging.
With any variations of matrixes, inter-aural cancellation, binaural recording, I do agree that at the end of the day, most of the program material we will be listening to may well sound just as good or better with a conventional stereo.
In your comment, " I think after 50 years of content being produced using a conventional (unprocessed) pair of speakers as a reference, using a non-standard listening setup can hardly be expected to improve the reproduction of the majority of recordings." I have to disagree. Some recordings are done very well, and many aren't. The recording end has all the same issues as at the playback end, dealing with acoustics, crosstalk, comb filter effects, etc. The recording end is always "painting a picture" so to speak, of an actual event, knowing the limitations of the reproduction process. They choose what they think is the best set of tradeoffs. They are often wrong, sloppy, or just optimizing for something different than what many of us have for playback. There are many variables (room acoustics, loudness compensation, speaker bass response and extension, etc.). As for the "playing artist not listener" statement, Fidelity is a good and necessary starting point, but then why not then make it sound it's best? Isn't the final goal enjoyment?
Last edited:
Some good points Bob. Sounds like we actually agree on more than I initially thought! Certainly, that enjoyment is the goal for most of us here
Unfortunately my experience with binaural reproduction has not been similar to yours. I simply cannot get a believable out-of-head front soundstage from any HRTF I've tried in any headphone even with measurement/EQ. Btw, the IRCAM set of real-head recordings is very cool for trying out different HRTFs. I did find one that kind of matched my own head and fooled around with some Ambisonic -> Binaural rendering. With head tracking, it becomes a lot more believable (no "head in vice" feeling, can move head to confirm localization), which is one of the reasons I don't think binaural *recording* is a viable technique compared to binaural *rendering*. There's a lot of development towards this kind of thing right now because of Virtual Reality and the upcoming Dolby/ITU standards.
These alternative / optimized stereo formats (CTC, binaural, M/S shuffling) seem to provide the most benefit to a subset of recordings - minimalist true stereo (microphone pair).
And yup, I agree that it's possible to improve upon the subjective quality of a given recording in one's own room, (and why not!) but these optimizations do not broadly apply and as such should be viewed as "personal remastering filters". (As opposed to an integrated part of the playback system.)
It seems all of us in this thread acknowledge that hi-fi could be so much better (more realistic, immersive..) than it typically is! This is encouraging. IMO the easiest way to sidestep these issues is adding channels, both playback and recording. Linkwitz would say that dipole stereo is all you need.
There's an obvious connection between the recording style and the optimal playback setup - but that's why I'm so passionate about standards! Reduce the movement of the target,and get a better shot at delivering a great experience for the most listeners.
Decware has a neat paper on setting up the speakers firing from a single corner.
DECWARE - Article about Setting up a Listening Room without Treatments
A buddy at work sits on floor, back to wall, facing speakers.
He and the speakers are across the narrowest of a shoebox shaped room.
That way the side reflections are very delayed in time and level.
Hardly the jukebox setup, but as a nearfield (or close field), that had it's own advantages.
DECWARE - Article about Setting up a Listening Room without Treatments
A buddy at work sits on floor, back to wall, facing speakers.
He and the speakers are across the narrowest of a shoebox shaped room.
That way the side reflections are very delayed in time and level.
Hardly the jukebox setup, but as a nearfield (or close field), that had it's own advantages.
The problem is real, and is unmasked more and more as the playback system becomes more phase coherent. Many systems don't have the problem, or very little (such as my current system) but others do.
And of course a lot of people simply don't notice it even when it is present.
The application of the effect, properly done, will not change to majority of recordings enough to notice. But for recordings with uneven tonal balance across L-C-R, it can really help.
If you are fixing problems that are ignored in the mix, then you are playing mastering engineer, not artist. Commercial recordings are the work of a committee - rarely some artist's untainted vision. I don't hold them sacred, I've seen and heard the sausage made.
The problem is real, and is unmasked more and more as the playback system becomes more phase coherent. Many systems don't have the problem, or very little (such as my current system) but others do.
And of course a lot of people simply don't notice it even when it is present.
The application of the effect, properly done, will not change to majority of recordings enough to notice. But for recordings with uneven tonal balance across L-C-R, it can really help.
If you are fixing problems that are ignored in the mix, then you are playing mastering engineer, not artist. Commercial recordings are the work of a committee - rarely some artist's untainted vision. I don't hold them sacred, I've seen and heard the sausage made.
Hey Pano,
Sorry I've been playing devils advocate here - I acknowledge the phantom center is a bit of a ****-show and a big step backwards from actual mono. The notches in frequency response as I move my head slightly around the center are irritating to be sure. Usually it sounds best a little bit off center.
"Mastering engineer" is more correct, yes. Still part of the creative process! I think you are right not to hold recordings sacred, but I was trying to point out that any shuffling probably can't offer a systematic improvement due to it not being auditioned upstream, in the production process.
If you're saying that the techniques outlined in FixingThePhantomCenter.pdf can help some recordings without harming the others, and/or that the pros outweigh the cons, then it has broader applicability than I thought. (There are a few techniques being discussed here - S curve EQ etc. - but I assume you are talking about FixingThePhantomCenter.pdf)
"Sausage" made me laugh.. a lot.
From what I read above you don't exactly seem to have a stable phantom center. It's quite possible to achieve that. No hearable notches, no drifting at certain frequencies etc. It does take a bit of work. I can wiggle sideways without any change and actually prefer that center sweet spot position. A mono recording comes from the center only, no bleeding etc... I didn't get there in one day though.
When you get to that point, that's when you start to notice the differences in tonality as mentioned in the beginning of this thread. From what you describe here, you have either too rough FR response or too many (early) reflections to notice or judge it now.
Personally, I've tried both options mentioned in this thread. Probably due to the nature of my speakers and/or processing, I preferred the S curve EQ. I tried it on many different songs to judge it and experience no obvious drawbacks. I wouldn't make a choice like this based on a couple of test tracks. Any solution has to work on the majority of music I listen to for me to adopt it.
As it is about a 2 to 3 dB difference (between mid and side) at max it's smaller than a lot of variations due to room stuff in the left/right channel in a room without any treatment/care.
A picture of my left/right balance:
It's an uneasy fix for problem most people don't even know about. For most people, most systems, it probably doesn't matter.
The main problem I've found is that the shuffler (and some other techniques) tend to brighten the center image to make it sound more like a single speaker. Recordings are not mastered with that in mind, so you have to re-EQ the system or it sounds too bright overall. Once the center matches the sides, the overall balance can be darkened a bit. That's important because there is so much important content right in the center.
So basically: Brighten the center to match the sides, then bring the whole system back to the original tonal balance. That done, it will sound very even, balanced and smooth.
The main problem I've found is that the shuffler (and some other techniques) tend to brighten the center image to make it sound more like a single speaker. Recordings are not mastered with that in mind, so you have to re-EQ the system or it sounds too bright overall. Once the center matches the sides, the overall balance can be darkened a bit. That's important because there is so much important content right in the center.
So basically: Brighten the center to match the sides, then bring the whole system back to the original tonal balance. That done, it will sound very even, balanced and smooth.
I'm a musician (guitarist) so fidelity is usually a 2nd priority after what song I'm listening to. I often listen to songs that were recorded horribly. Old classic jazz and blues, Motown; much of which is over-modulated, etc., so I want to be able to do somewhat of a repair job on much of what's out there, but also be ready for the occasional really good quality recording. Even binaural recordings (which I may get into creating). Trying to find a single best way to do reproduction seems impossible. Way too many variables. High-end preamps that have no tone controls seem obsurd to me.
As for the phantom center image (yes I remember), it may be pretty good with a floor to ceiling vertical line array since room acoustics become integral, but with point source speakers there will be comb filter effects at the highest frequencies. Each ear will get a different set of cancellations. Inter-aural cancellation techniques make this worse (or more obvious). With a phantom center, when you're off axis a bit, the center image tends to jump to either the right or left speaker. Actively steered multi-channel surround decoders occasionally do weird things when playing music (mine does anyway).
I've been reading about decorrelation techniques again. It sounds like they may stabilize a phantom center image better when listening off axis. But then how do you separate the center image from the rest without crosstalk causing phasing effects? If decorrelation is done with a 5 - 18mS delay applied to the L&R signals relative to L+R content, crosstalk may create a phaseshifter effect that I don't want, but one paper suggested a multi-delay or very short term "reverb" (like a vibrating vinyl disk might produce) limited to between 1 - 20mS, so cancellations would get filled in better thereby minimizing the phaseshifter sound effect..
Adding this very short term reverb decorrelation technique to a passive stereo L-XR matrix output, and perhaps also to an actual L+R center speaker seems like an interesting thing to try. There will still be some electrical crosstalk in the matrix (minimized by a continuously variable X factor in the L-XR), but it might work pretty well anyway.
I'm building such a beast right now (PCB's are due to arrive today). I'll report back with results. It will take a while to build since I'm an expert at making things more complicated than they probably needed to be (It also has a variation of the Carver Hologram circuit that can optionally be switched in for the L &R outputs).
As for the phantom center image (yes I remember), it may be pretty good with a floor to ceiling vertical line array since room acoustics become integral, but with point source speakers there will be comb filter effects at the highest frequencies. Each ear will get a different set of cancellations. Inter-aural cancellation techniques make this worse (or more obvious). With a phantom center, when you're off axis a bit, the center image tends to jump to either the right or left speaker. Actively steered multi-channel surround decoders occasionally do weird things when playing music (mine does anyway).
I've been reading about decorrelation techniques again. It sounds like they may stabilize a phantom center image better when listening off axis. But then how do you separate the center image from the rest without crosstalk causing phasing effects? If decorrelation is done with a 5 - 18mS delay applied to the L&R signals relative to L+R content, crosstalk may create a phaseshifter effect that I don't want, but one paper suggested a multi-delay or very short term "reverb" (like a vibrating vinyl disk might produce) limited to between 1 - 20mS, so cancellations would get filled in better thereby minimizing the phaseshifter sound effect..
Adding this very short term reverb decorrelation technique to a passive stereo L-XR matrix output, and perhaps also to an actual L+R center speaker seems like an interesting thing to try. There will still be some electrical crosstalk in the matrix (minimized by a continuously variable X factor in the L-XR), but it might work pretty well anyway.
I'm building such a beast right now (PCB's are due to arrive today). I'll report back with results. It will take a while to build since I'm an expert at making things more complicated than they probably needed to be (It also has a variation of the Carver Hologram circuit that can optionally be switched in for the L &R outputs).
Last edited:
Well that certainly looks awesome wesayso, how did you make that measurement? Is it similar a bit off-axis? I know I know.. its all in your thread.
So you think my early reflections and/or channel matching are to blame for my perception of the phantom center? I thought these were known problems - the phantom center comb filter is quite location dependent (strongest when the speakers are exactly equidistant). . The early reflections are responsible for the changes as I make small movements, although studies show that they should actually be helpful?
A problem with "brightening the center to match the sides" is the difference in HRTF of 30 degrees vs. 0 degrees. Having a darker center (due to comb filtering) could compensate for the difference between these response? This was brought up by DDF and rtf earlier.
I suppose if you are doing this EQ'ing by ear then we have not much to worry about.. perhaps more psychoacoustic studies are needed, rather than relying on measurements from the KEMAR head. So I do appreciate your subjective impressions - perception is complicated after all.
It would be incredibly helpful to have 3 matching speakers when doing these comparisons, not to compare the sound of L vs C, but L+R vs C if you intend on removing the center speaker while reproducing LCR content. 3 of basically *any* decent speaker should be adequate for this; I wouldn't think you need to build a 3rd line array for instance
Again this may all be a "fools errand" since center elements are often EQ'd brighter (or bright mics used), of course it depends on the recording.Bob, yes the phantom image is better with my corner line arrays than with conventional 2 or 3 ways placed away from the walls. In an untreated room mind you. I think the idea with the decorrelation techniques is that it isn't perceptible on a mono (L or R only) signal.
My experiences with a LR > LCR matrix haven't been positive but others especially Elias have reported good results, so maybe I am just doing it wrong. Not super interested in trying that again at the moment, but it could be a valid technique for dealing with these "phantom center issues", so interested to hear your impressions. Especially compared with the other techniques, a solo center channel, and a conventional L+R phantom center.
I'll have to revisit these techniques once I get my line arrays and room better set up, perhaps with some removable sidewall absorption. Recently I have been working/listening on a 2-way speaker which also may be causing some of these issues - RS180 + Neo8. The Neo8 does some wacky things in the direct sound around 10kHz as you move around (interference pattern related to the spacing of the holes I read somewhere, can't find the link), so I should confirm my impressions with my Infinity speakers (waveguide and phase shield dome tweeter, should not have these issues).
Very thought provoking thread.. my head hurts a bit.
Attachments
Last edited:
I really don't think that article is to be trusted for 2 reasons:
1. "the direct sound is ideally all you should hear" (paraphrasing). Uh, no.
2. those raytracing diagrams of "first reflections". What?? Is the speaker a laser?
I'm just afraid the short reverb, added to the center will cause our perception to notice that as an increase of distance to that voice. From what I've read a dry recorded voice appears to be closer to you while added reverb will create a sense of distance. Listening to my ambient channels confirms that, though I add (low level) late reverb to those.
The early phase shuffler as presented in this thread also had some shift in the center vocal, moving it slightly further away. It acted as (controlled) early reflections in a way. It caused my sides to stick to the speaker location whereas without it you wouldn't know or guess that exact position with your eyes closed. The later rendition with only a phase manipulation did much better in that regard.
I wish more people would play with the examples though. Sure it's fun to talk and theorize, but some experimentation might go a long way to see general trends. I realize my arrays are quite different from point sources and therefore I cannot claim what works for me would work for other types of speakers as well. But I guess to really want to play with this you'd need to hear the effect discussed here first to actually acknowledge it as a potential problem.
As Pano was playing with Altec Multicell horns during his first experiments it could in some way be more comparable with my multi source line arrays?
Or do we have any takers with point source speakers that notice the tonal differences?
Yeah, and the phantom center being perceived as further away from the listener than the L/R is already an issue (though some people like this as it is more "magical").
I would love to listen to some examples without having to implement the various techniques myself. Are they linked in this thread somewhere? (unprocessed wav and first technique/S curve pre-processed wav). IRs would also work.
I'd have to dig back thru the thread, but I did post some samples. As wesayso mentions, the later versions of the shuffler tend to work better for us. I made improvements on the method described in the paper.
Rather than look for samples, getting the convolver running is the best idea. That way you can easily listen to whatever music you have. I suggest JRiver Media Center as the best playback software, or you can use Foobar if you like.
I'm not sure that making a phantom fixer that does not affect mono recordings is possible. Since the dulling of the center image comes from comb filtering, mono recordings should suffer the worst center dulling as they will be the most correlated.
Rather than look for samples, getting the convolver running is the best idea. That way you can easily listen to whatever music you have. I suggest JRiver Media Center as the best playback software, or you can use Foobar if you like.
I'm not sure that making a phantom fixer that does not affect mono recordings is possible. Since the dulling of the center image comes from comb filtering, mono recordings should suffer the worst center dulling as they will be the most correlated.
Thanks Pano I'll find the samples. Do you have IR that I can use in the convolver? I'm guessing you would run a MS encoder, convolver, then MS decoder? (For S curve). And just LR for the decorrelation technique?
In the FixingThePhantomCenter.pdf, there is a paragraph that supports my assertion that the comb filtering of the phantom center changes as you move slightly off-center:
"When the listener is not equidistant from the speakers,
the comb filter peaks and nulls will be in different
frequency locations at the two ears, as shown in the
simple time-delay simulation of Figure 2. (In real life,
the responses would be significantly more complicated,
especially at high frequencies, due to head-shadowing,
pinna reflections and other effects.) "
I have switched to my Infinity R263 speakers to confirm some of what I was hearing with regard to the changing comb filtering. The effect is reduced - perhaps due to the wider dispersion, perhaps due to the tweeter having smoother response at small differences in angle. I'm talking about very small movements of the head, centimeters. And the effect is most noticeable at high frequencies. Note that I have untreated sidewalls; I have to install sidewall absorption to confirm that this is related to the two-speaker comb filtering, not first reflection comb filtering.
As for these techniques not affecting perception of mono recordings, that's not what I meant. Whats the point of fixing the phantom center... if it doesn't fix the phantom center? I meant that it wouldn't change perception of the hard L/R sounds. For the decorrelation technique.
Here's a test file which contains various treatments of pink noise. All are highpassed at 500Hz. Two of each clip are played in a row to give more auditioning time.
a) Right channel only, +3dB (to level match with stereo noise)
b) Both channels playing the same file (phantom center)
c) Both channels, one inverted polarity
d) Both channels, one delayed arbitrarily to decorrelate
I find clips (b) and (c) much more susceptible to the changes in frequency response induced by small head movements. In my untreated room, they are quite spectrally similar to (a) however.
There is always some change in frequency dips as I move around, even in the true mono (R channel only) file. This effect is not limited to speakers either; I can perceive it on other high frequency sources like a computer fan or high pitched whine from a power supply. I assume that is due to reflections in the room.
One more question for wesayso: is it actually a bad thing to precisely localize L and R speakers when they are playing hard panned elements? I would think that is completely natural and a good thing. If a sound should have a greater apparent source width than 0, that can be accomplished with various stereo techniques.
In the FixingThePhantomCenter.pdf, there is a paragraph that supports my assertion that the comb filtering of the phantom center changes as you move slightly off-center:
"When the listener is not equidistant from the speakers,
the comb filter peaks and nulls will be in different
frequency locations at the two ears, as shown in the
simple time-delay simulation of Figure 2. (In real life,
the responses would be significantly more complicated,
especially at high frequencies, due to head-shadowing,
pinna reflections and other effects.) "
I have switched to my Infinity R263 speakers to confirm some of what I was hearing with regard to the changing comb filtering. The effect is reduced - perhaps due to the wider dispersion, perhaps due to the tweeter having smoother response at small differences in angle. I'm talking about very small movements of the head, centimeters. And the effect is most noticeable at high frequencies. Note that I have untreated sidewalls; I have to install sidewall absorption to confirm that this is related to the two-speaker comb filtering, not first reflection comb filtering.
As for these techniques not affecting perception of mono recordings, that's not what I meant. Whats the point of fixing the phantom center... if it doesn't fix the phantom center?
I meant that it wouldn't change perception of the hard L/R sounds. For the decorrelation technique.Here's a test file which contains various treatments of pink noise. All are highpassed at 500Hz. Two of each clip are played in a row to give more auditioning time.
a) Right channel only, +3dB (to level match with stereo noise)
b) Both channels playing the same file (phantom center)
c) Both channels, one inverted polarity
d) Both channels, one delayed arbitrarily to decorrelate
I find clips (b) and (c) much more susceptible to the changes in frequency response induced by small head movements. In my untreated room, they are quite spectrally similar to (a) however.
There is always some change in frequency dips as I move around, even in the true mono (R channel only) file. This effect is not limited to speakers either; I can perceive it on other high frequency sources like a computer fan or high pitched whine from a power supply. I assume that is due to reflections in the room.
One more question for wesayso: is it actually a bad thing to precisely localize L and R speakers when they are playing hard panned elements? I would think that is completely natural and a good thing. If a sound should have a greater apparent source width than 0, that can be accomplished with various stereo techniques.
No
. But I tested the shuffler on a song where I'm used to the placement in the imaging. It changed my perception of that song. The later phase only version did not have that problem.The early version might have pushed back the singer slightly, but not by a lot from what I remember. I remember thinking this would be a pretty good solution for Home Theatre.
Here is one sample done with the impulse found in the paper.
http://www.diyaudio.com/forums/multi-way/277519-fixing-stereo-phantom-center-3.html#post4399979
Here is the original shuffler high passed as it should be
http://www.diyaudio.com/forums/multi-way/277519-fixing-stereo-phantom-center-4.html#post4400043
More samples:
http://www.diyaudio.com/forums/multi-way/277519-fixing-stereo-phantom-center-6.html#post4400796
And here is the phase only shuffle:
http://www.diyaudio.com/forums/multi-way/277519-fixing-stereo-phantom-center-10.html#post4425656
http://www.diyaudio.com/forums/multi-way/277519-fixing-stereo-phantom-center-3.html#post4399979
Here is the original shuffler high passed as it should be
http://www.diyaudio.com/forums/multi-way/277519-fixing-stereo-phantom-center-4.html#post4400043
More samples:
http://www.diyaudio.com/forums/multi-way/277519-fixing-stereo-phantom-center-6.html#post4400796
And here is the phase only shuffle:
http://www.diyaudio.com/forums/multi-way/277519-fixing-stereo-phantom-center-10.html#post4425656
OK.. I've been listening to the various IR filters. Thanks for making the easier to find Pano!
I loaded up all the IRs in jconvolver and togged between them with jack_mixer. Without a lag between switches, it made it a lot easier. I was able to match levels that way too. (It was only the high pass version that needed boosting.)
I spent quite a few hours comparing, and I must say I'm very impressed with the Rephase-2 impulse. The imaging is often more natural with it on, then off. Mono mixes ("River Deep Mountain High", Beach Boys Smile Sessions) have a heightened sense of realism. Able to hear more details within mono elements.
Speaker setup used is pictured below. The Neo8 2-ways spaced about 80cm (ctc) and 80cm from each ear.
Listening impressions from a bunch of lossless music and the pink noise samples:
1. The original shuffler and hipass both sounded really bad to me. I see now what wesayso meant about pinning L/R sounds to the speakers. It's a weird effect, not natural. Affects the frequency response of hard panned sounds and pulls the image apart (widening). Original pulls strongly to the left (screenshot below).
2. The Rephase-2 shuffler was leagues ahead of the rest. Usually brightened the center channel with not too much side effects (no pun intended). MUCH improved phantom imaging. Does move it back a bit, but not a lot. Moreso it moves it... in front of you. Phantom center has a stronger (vertical/central) position. I think when its sounds moved back, its because reflection details in the center have been unmasked.
3. Some music already has a brightened center mix and does not want further "opening up". Could always alter the voicing, keep the imaging effects but not the frequency response effect. (Perhaps with a MS eq)
3. For the rephase shufflers, they had a similar effects while listening off center, which surprised me.
4. The perceptually brighter center channel has the unfortunate side effect of unmasking compression and distortion artifacts along with reflection cues, harmonics etc...
5. It is difficult not to confuse the frequency and spatial effects, I will try some "voicing" on the Rephase-2 filter
I would love to hear wesayso's S-curve thing as well, since it is your chosen solution. Seems like it could be used *as well as* this rephase method even? They solve different problems, even related ones since the phase decorrelatioin perceptually brightens the center.
A MS IR would be fine, L channel for Mid and R channel for Side. To audition I can MS encode before convolution, MS decode after.
Thanks for the IRs, it was an easy way to try it, and I'm now a believer!
I loaded up all the IRs in jconvolver and togged between them with jack_mixer. Without a lag between switches, it made it a lot easier. I was able to match levels that way too. (It was only the high pass version that needed boosting.)
I spent quite a few hours comparing, and I must say I'm very impressed with the Rephase-2 impulse. The imaging is often more natural with it on, then off. Mono mixes ("River Deep Mountain High", Beach Boys Smile Sessions) have a heightened sense of realism. Able to hear more details within mono elements.
Speaker setup used is pictured below. The Neo8 2-ways spaced about 80cm (ctc) and 80cm from each ear.
Listening impressions from a bunch of lossless music and the pink noise samples:
1. The original shuffler and hipass both sounded really bad to me. I see now what wesayso meant about pinning L/R sounds to the speakers. It's a weird effect, not natural. Affects the frequency response of hard panned sounds and pulls the image apart (widening). Original pulls strongly to the left (screenshot below).
2. The Rephase-2 shuffler was leagues ahead of the rest. Usually brightened the center channel with not too much side effects (no pun intended). MUCH improved phantom imaging. Does move it back a bit, but not a lot. Moreso it moves it... in front of you. Phantom center has a stronger (vertical/central) position. I think when its sounds moved back, its because reflection details in the center have been unmasked.
3. Some music already has a brightened center mix and does not want further "opening up". Could always alter the voicing, keep the imaging effects but not the frequency response effect. (Perhaps with a MS eq)
3. For the rephase shufflers, they had a similar effects while listening off center, which surprised me.
4. The perceptually brighter center channel has the unfortunate side effect of unmasking compression and distortion artifacts along with reflection cues, harmonics etc...
5. It is difficult not to confuse the frequency and spatial effects, I will try some "voicing" on the Rephase-2 filter
I would love to hear wesayso's S-curve thing as well, since it is your chosen solution. Seems like it could be used *as well as* this rephase method even? They solve different problems, even related ones since the phase decorrelatioin perceptually brightens the center.
A MS IR would be fine, L channel for Mid and R channel for Side. To audition I can MS encode before convolution, MS decode after.
Thanks for the IRs, it was an easy way to try it, and I'm now a believer!
Attachments
Last edited: