As you know, Fletcher at Bell Labs did a lot of the early stereo experiments in conjunction with an experiment in reproducing the Philly Orchestra at a remote concert hall, in real time.
Stokowski, Harvey Fletcher, and the Bell Labs Experimental Recordings
(Neat web site)
Their experiments showed that 3 channel stereo was better than two channel but also that 3 channel with a center derived, while not as good as discrete 3 channel, was better than 2 channel. This was with summed mono for the center.
One problem with a derived center is it tends to reduce seperation, by increasing the in-phase material, although you can spread the left and right speakers farther apart to compensate. You can also play matrix games such as feeding the left speaker L - .5 R and the right R - .5L to bring the separation up a bit. Finally, pro-logic steering can be used to give more isolation between L, center and R.
None of these solutions are ideal. As we were discussing comb filtereing a few pages back it reinforced that having a 3 channel system with wide speaker seperation would really be a big improvement.
I've just picked up a Denon AVR to get my home theater running again. I want to start experimenting with some of the matrixed and simulated sound fields, plus get my SACD RCA living Stereo 3 channel discs playing.
David S.
Which one did you get? I'd like to have someone knowledgeable to talk to about Audyssey.
The living stereo SACDs are fun, I have a few.
Somewhere here I have the schematic of the Dolby decoder as used in 1000s of cinemas before digital. IIRC, it was mostly just summed center, but I'd have to double check.
Bottom line: Does it matter? How was the track mixed? That might be the most important thing - unless you want to derive 3-4 channels out of a stereo mix. Many software players can derive multi-channels from 2. And as David says, there is always Pro-Logic, which was designed for small rooms and close speakers.
Somewhere here I have the schematic of the Dolby decoder as used in 1000s of cinemas before digital. IIRC, it was mostly just summed center, but I'd have to double check.
Bottom line: Does it matter? How was the track mixed? That might be the most important thing - unless you want to derive 3-4 channels out of a stereo mix. Many software players can derive multi-channels from 2. And as David says, there is always Pro-Logic, which was designed for small rooms and close speakers.
I've always liked the idea of a centre channel as well, but I see a major implementation problem with its application to 2 channel music (matrix decoded to 3 channels) as opposed to discrete 5.1 channel DVD's where the centre channel is generally dialogue which is uncorrelated with other channels.
And that problem is that centre channel speakers typically don't phase track with large floor standing Left and Right speakers.
For example if your main speakers were a 3 way design with 300Hz 2nd order L-R and 3Khz 4th order L-R crossovers, even if you were to use the same midbass driver and tweeter in your centre channel, the 2 way centre channel with only a 3Khz 4th order L-R is (in this example) half a phase rotation short of the 3 way, as the 3 way has additional phase rotation at the lower crossover point.
Therefore regardless of which way you phase the centre channel it is going to be out of phase with the main speakers at either low frequencies or high frequencies, with a transition in between.
As I said, not really a issue for 5.1 movies where the centre channel is discrete and uncorrelated, but if you are deriving 3 channels from stereo then Left, Centre, and Right need to be correlated at least at low and mid frequencies for continuous panning between centre and one side to be possible.
The problem could be solved with digital phase equalization, but that would require very careful measurements of the speakers to estimate the required phase correction - something very difficult (maybe impossible) to do accurately down to low frequencies in-situ in the living room. (Pre-defined presets for "known" models of speakers may be possible though, and close enough)
One other option might be to simply use identical small 2 way speakers for all three speakers and a separate sub, but that's not too appealing to those of us who like 3 way designs.
Yet another problem I anticipate with a centre channel derived from stereo is that for continuous panning between sides and centre there must be some phase correlation between centre and side speakers, yet for anything other than a directly centre or fully panned Left or Right sound source you now have comb filtering potential again - except this time the angular separation (between centre and one side) is only half the left to right separation, and probably not enough to minimize comb filtering sufficiently.
Has anyone solved these two issues ?
Last edited:
To pan part way between centre and one side with 3 channels, (derived from 2) there has to be phase correlation - at least at low and mid frequencies. If there is an extra 180 degree phase rotation in your 3 way Left and Right speakers at 300Hz, but not in your 2 way centre channel, surely this will screw up any sounds that are partially panned between centre and one side ?
Last edited:
Reason for the center speaker is to make 'imaging' more stable for off-center listeners.
But every different processor should be doing 're-panning' so, that 2 channel and n channel listening would give same localization ques. Search the AES for Avendano&Jot and Gerzon for different techniques.
Another issue is with envelopment, also called 'reverberance imagery'. With 2 channel listening, all this reverberance is coming from front - it is a serious limitation and without getting this correct, you cannot have believable 'you are there' sensation - on other hand, if reverberance imagery is supplied by your (live)listening room, you can have believable 'they are here' sensation.
Usher, Johnston, Griesinger and others have researched this issue, and found that to (re)create a different reverberance image than your living room supplies, you have to have correct signals reproduced from at least 4 loudspeakers - a good potential for multichannel music, but we all know stories how research and 'sound engineers' go hand-in-hand.
On another note, Wave Field Synthesis isn't very practical this day, but has anyone heard some recent development of Perceptual Soundfield Reconstruction?
AES E-Library Perceptual Soundfield Reconstruction
But every different processor should be doing 're-panning' so, that 2 channel and n channel listening would give same localization ques. Search the AES for Avendano&Jot and Gerzon for different techniques.
Another issue is with envelopment, also called 'reverberance imagery'. With 2 channel listening, all this reverberance is coming from front - it is a serious limitation and without getting this correct, you cannot have believable 'you are there' sensation - on other hand, if reverberance imagery is supplied by your (live)listening room, you can have believable 'they are here' sensation.
Usher, Johnston, Griesinger and others have researched this issue, and found that to (re)create a different reverberance image than your living room supplies, you have to have correct signals reproduced from at least 4 loudspeakers - a good potential for multichannel music, but we all know stories how research and 'sound engineers' go hand-in-hand.
On another note, Wave Field Synthesis isn't very practical this day, but has anyone heard some recent development of Perceptual Soundfield Reconstruction?
AES E-Library Perceptual Soundfield Reconstruction
Last edited:
His video (slides) is interesting too:
The Loudness War: Background, Speculation and Recommendations
The Loudness War: Background, Speculation and Recommendations
No, it way predates that. Its an AVR 3801. Mint and less than a quarter of its new price. I just wanted something that would decode DD and DTS and allow me to plug in an SACD player.
I was digging through a Sony manual of a recent AVR with network connectivity, Audyssey, internet radio, Digital radio, Satellite radio. Does anybody use all that? (Or figure out how it works?)
David
Consider this situation - we have a signal mixed equally in the LR channels to yield a phantom image in the center. Now if we mix the LR into a center I now have three widely spaced sources of identical signals across the front and this will make a mess of any attempts at directivity control. Only if the parts of the L and R signal that are sent to the center are removed from the L and R will the directivity of the result be what it is supposed to be. With discrete 3 channels this is not an issue, but when C is derived from LR then it can be a big issue.
I have a strong preference for three channels when they are recorded that way. I worry about what happens in the derived case, and you can't get any information about how the AVRs do the derivations.
I have a strong preference for three channels when they are recorded that way. I worry about what happens in the derived case, and you can't get any information about how the AVRs do the derivations.
My plan for a center channel signal is to just add L&R, knowing that whatever program material is recorded L-R (which would include any prologic material that is intended by the encoder to be sent to the side or rear speakers), will be mostly cancelled out. So it's got something going for it since I will be doing a five speaker half circle arrangement.
As far as the phase problems when mixing with the acoustic signals from the left and right speakers, I'll see how big a problem that actually is psycho-acoustically in my room, and then I plan to experiment with putting between 1-20mS delay on the center speaker, relative to the L&R speakers, and see what effect that has. The math is over my head, but I have a suspicion that it might help the center speaker get out of the way of timing related imaging cues in the lower midrange better. The delay will shift in frequency where the cancellations will land, and change the psycho-acoustic effects. It may give me a subjective improvement anyway.
If I want steering, I've already got that in a Yamaha AV rcvr, in several forms (DTS, Dolby and Yamaha's own brew). Actually, the electronics I'm building will have the option of up to 3dB of steering on the center speaker signal only, based on the rectified L+R to L_R difference, or just the L+R signal level. It's all experimental.
As far as the phase problems when mixing with the acoustic signals from the left and right speakers, I'll see how big a problem that actually is psycho-acoustically in my room, and then I plan to experiment with putting between 1-20mS delay on the center speaker, relative to the L&R speakers, and see what effect that has. The math is over my head, but I have a suspicion that it might help the center speaker get out of the way of timing related imaging cues in the lower midrange better. The delay will shift in frequency where the cancellations will land, and change the psycho-acoustic effects. It may give me a subjective improvement anyway.
If I want steering, I've already got that in a Yamaha AV rcvr, in several forms (DTS, Dolby and Yamaha's own brew). Actually, the electronics I'm building will have the option of up to 3dB of steering on the center speaker signal only, based on the rectified L+R to L_R difference, or just the L+R signal level. It's all experimental.
I think this article is linked before in this thread:
www.sfxmachine.com/docs/FrequencyDomainUpmix.pdf
It uses some vertor-based signal decomposition algorithm. It should be easy to write some code in matlab/octave to test it...
Griesinger (exHarman, Logic7) and Johnston (exTHX, Neural) are both AES fellows, upmixing methods are well documented and tested. This doesn't happen with correlated signals, ask them.
Your excample could more likely happen with an incompetent mixing engineer with multichannel material than with a competent upmixing algorithm.
Worse yet, correlated signals are mixed to all channels! Thats why most 'pop' multichannel recordings sound better, when you upmix the 2 channel version instead of the discrete version.
Last edited:
Everything but satellite and digital radio. Used to use satellite back in its infancy but it's lost ground. Slacker radio has changed my mind about what else might be necessary. I highly recommend it. Nothing too wrong with Pandora either. I know some great people work there at Slacker. Yes, they are into psychoacoustics for sure. Chose their broadcast by the evidence even. Had hrs worth of conversation about it(and some other things) with some of them in person.
Digital radio is not impressive yet.
Dan
The essence of steering (Dolby Pro Logic, DTS Neo) is to constantly look at the 4 channel (derived) signal vectors, and if any channel dominates, let it through while feeding the vectorally opposite signal to the other 3 channels. This gives full separation between the dominate channel and everything else. Otherwise you are stuck with the usual 3dB max separation between adjacent channels.
You may not like the effect, and it won't be as good as discrete, but they claim to have well optimized algorithms for music with minimal leakage or steering artifacts.
http://www.dolby.com/uploadedFiles/..._Logic_II_Decoder_Principles_of_Operation.pdf
Dolby white paper
Dolby Pro Logic II vs. SRS Circle Surround
Comparison of DPLII with Circle Surround
David S.
Aargh, those evil speaker managers always asking for more!
Not evil, just misunderstood.
I played with that wave tank simulator for over an hour yesterday. Very cool for visualisation!
David S.
Dolby PL: http://www.dolby.com/uploadedFiles/..._Logic_II_Decoder_Principles_of_Operation.pdf
You can catch Roger Dressler over at AVS Forum
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.