Hafler and matrix surround is bringing back old old memories. I worked near Atlanta at a theater equipment manufacturer and we bought tons of Hafler amp modules. Probably kept them in business for a long while.
We also developed our own Dolby Surround decoder. It was kind of amazing how Dolby's original surround and our updated version worked. None of it had the benefit of DSP technology. I'm not sure what Dolby used but we built our matrix detector using Voltage Controlled Amplifiers (VCA's). Essentially op amps with gain adjusted by a DC voltage. The sums and differences of various inputs would produce varying DC levels to the L, C , R and surround VCAs.
Dolby's matrix steered very hard and as a result lost a lot of ambience. The focus was on effects so not very surprising. It was also noisy as hell, but I digress.
In the lab we had a version which allowed us to turn a pot to adjust how hard the steering would be and I can't tell you how much better softer steering sounded. It still steered but wouldn't kill the side channels as hard.
Anyway, my point is that if the OP wants to experiment with modest steering he'd probably be pleased. Otherwise in the modern world I've found Neo6: Music and movie modes to be outstanding for this type of work.
We also developed our own Dolby Surround decoder. It was kind of amazing how Dolby's original surround and our updated version worked. None of it had the benefit of DSP technology. I'm not sure what Dolby used but we built our matrix detector using Voltage Controlled Amplifiers (VCA's). Essentially op amps with gain adjusted by a DC voltage. The sums and differences of various inputs would produce varying DC levels to the L, C , R and surround VCAs.
Dolby's matrix steered very hard and as a result lost a lot of ambience. The focus was on effects so not very surprising. It was also noisy as hell, but I digress.
In the lab we had a version which allowed us to turn a pot to adjust how hard the steering would be and I can't tell you how much better softer steering sounded. It still steered but wouldn't kill the side channels as hard.
Anyway, my point is that if the OP wants to experiment with modest steering he'd probably be pleased. Otherwise in the modern world I've found Neo6: Music and movie modes to be outstanding for this type of work.
I'm thinking about trying to cook up my own upmix method using DSP. It's fun to think about ways it could work. Exactly what should it do with signals that are completely out of phase, or partially out of phase? I've got some ideas but have to learn some coding to execute it. At least this way I'll know what's going on and perhaps come to a better understanding of what sounds good to me and what doesn't.
So was the voltage control coming from another channel to create steering? And this was a matrix detector for material already encoded with surround?
So by kill the side channels, you mean pull all of the center channel info completely out of them?
Hi OP:
Glad you found any of my ramblings interesting. Before going on too much longer I want to give credit to the engineer who actually designed the circuit I'm discussing, Oscar Neundorfer. I only did the PCB design and reverse engineering of the Dolby boards of the time.
Please keep in mind the goal of the matrix decoder was to give everyone in a theater an excellent experience, no matter how far left or right or back they sat.
To answer your questions:
No. Dolby Surround used exclusively 2 channels from which a center and surround signal were derived and from which all 4 were steered. The only case I'm aware of where a separate control channel was used was for Sensurround.
Yes, exactly. A good decoder would perform reasonably well with mono or stereo signals, but also work for those with full 4 channels encoded.
The control signals were used to turn the 4 channels up or down as needed. Lets consider the simple case of the center. It's very simple, it's L + R.
The problem is what do you do when you have a L only signal? That's where the analog logic comes in. It should notice that there's no R at all, or very low level given noise, and essentially shut the C off. It should also be aware when L and R are on but the signals are not highly correlated.
A similar thing happens for Surround, which was L - R (or R-L, can't remember, but not relevant). Any time that L and R are not identical this would produce a signal in the surrounds so the matrix' job was to keep the Surround off until there was an equal and opposite signal in the L and R channels.
Both the Dolby and the other circuit I'm thinking of used a number of derived pseudo channels, fed by VCA from a variety of combinations, and each output op amp would eventually have multiple inputs to sum. If you haven't learned about op amp theory and how to mix multiple signals now is a good time.
There were additional conditions possible than discussed above. For instance, if a strong center was detected, L and R should be suppressed unless L and R had significantly uncorrelated signals. This is where hard steering comes in. Dolby would be ruthless in turning L and R down, while Oscar was showing that being more gentle could really let the sound track breathe and create a more immersive experience.
The hard steering effects of Dolby Surround were the bane of many a movie sound engineer. It took a great deal of control away from them, something that's been completely restored with the advent of Dolby Digital and DTS. 70mm film with discrete magnetic tracks and the very rare 35mm discrete magnetic track films (The Natural) really showcased how limiting Dolby Surround was.
Hope this helps!
Glad you found any of my ramblings interesting. Before going on too much longer I want to give credit to the engineer who actually designed the circuit I'm discussing, Oscar Neundorfer. I only did the PCB design and reverse engineering of the Dolby boards of the time.
Please keep in mind the goal of the matrix decoder was to give everyone in a theater an excellent experience, no matter how far left or right or back they sat.
To answer your questions:
No. Dolby Surround used exclusively 2 channels from which a center and surround signal were derived and from which all 4 were steered. The only case I'm aware of where a separate control channel was used was for Sensurround.
Yes, exactly. A good decoder would perform reasonably well with mono or stereo signals, but also work for those with full 4 channels encoded.
The control signals were used to turn the 4 channels up or down as needed. Lets consider the simple case of the center. It's very simple, it's L + R.
The problem is what do you do when you have a L only signal? That's where the analog logic comes in. It should notice that there's no R at all, or very low level given noise, and essentially shut the C off. It should also be aware when L and R are on but the signals are not highly correlated.
A similar thing happens for Surround, which was L - R (or R-L, can't remember, but not relevant). Any time that L and R are not identical this would produce a signal in the surrounds so the matrix' job was to keep the Surround off until there was an equal and opposite signal in the L and R channels.
Both the Dolby and the other circuit I'm thinking of used a number of derived pseudo channels, fed by VCA from a variety of combinations, and each output op amp would eventually have multiple inputs to sum. If you haven't learned about op amp theory and how to mix multiple signals now is a good time.
There were additional conditions possible than discussed above. For instance, if a strong center was detected, L and R should be suppressed unless L and R had significantly uncorrelated signals. This is where hard steering comes in. Dolby would be ruthless in turning L and R down, while Oscar was showing that being more gentle could really let the sound track breathe and create a more immersive experience.
The hard steering effects of Dolby Surround were the bane of many a movie sound engineer. It took a great deal of control away from them, something that's been completely restored with the advent of Dolby Digital and DTS. 70mm film with discrete magnetic tracks and the very rare 35mm discrete magnetic track films (The Natural) really showcased how limiting Dolby Surround was.
Hope this helps!
Last edited:
It may not be possible to apply DPL on the receiver unless it "knows" exactly which channel is which, and is capable of re-mapping appropriately. However, you could apply DPL from the music player itself, before the 3-way crossing split occurs.
Just a quick question >
With consumer surround amps & receivers, have they yet gone beyond 7.2 ?
With consumer surround amps & receivers, have they yet gone beyond 7.2 ?
DPL had two modes and I'm not sure which mode you're referring to, ... When I tried DPL cinema mode on music, the steering did work well but I did not use any surround channels (3/0 LCR only). However, tastes change from person to person.. ,
Yes. DTS and Dolby Digital (Atmos) have gone to object based sounds. In original DTS/DD with 5.1 for instance, each channel was discretely encoded, and you had perfect separation. However with the advent of cheap DSP and compute power overall movie sound tracks have gone to encoding an object, for instance a helicopter or flying bug as an object in space, and leave it up to the processor to determine which speakers that goes to.
In theory, this would allow you to have 100 surround speakers, and the processor would determine how that bug or helicopter pans around the speakers.
This is part of the reason why you now see crazy number of output channels on high end processors.
Sorry, I should clarify. In my posts above I was referring to "Dolby Surround" as used in motion picture auditoriums. For reference look up the Dolby CP 50/CP 55 processors.
For home use, the original Dolby Surround label was used for a system which did not have the full capabilities of the motion picture version. Consumer level "Dolby Surround" had a gated surround channel but no other steering if I remember correctly. When Dolby finally brought all of the motion picture goodness to home processors the consumer products were re-badged as "Dolby Pro Logic."
Consumer Dolby Surround = Motion picture Dolby Surround without the full steering capabilities
Consumer Dolby Pro-Logic = Motion picture Dolby Surround
Consumer Dolby Pro-Logic = As above, but I think it introduced better steering and a music mode. Not 100% sure.
As far as I know, "Dolby Pro Logic" was never a label used in motion picture auditoriums, it was always called "Dolby Surround"
Going back to my discussion on Dolby Surround and op amps being used for the final mix downs.... when a strong surround signal was detected, that needed to be removed from the front channels. Your plane or bullets can't fly over if they are also playing through the fronts.
I've found used Q-Sys processors on ebay to be a great way to experiment with such.
For instance, it's been easy to implement any of the LCR matrices in this link..http://elias.altervista.org/html/3_speaker_matrix.html
The drag and drop method of placing components into a processing schematic allow for endless, easy to try, experimentations.
I've been thinking about this issue. What we really want is to tease out in phase from out of phase sounds between the channels. If sound is in phase and equal level in both channels at a particular frequency, it should be hard steered to the middle. But opposite phase content should be left entirely in the side channels. That's my thinking.
But what do you do when a frequency is playing in both channels that's partially out of phase? It seems to me that all three channels should play in that case. So L+R gives you the in phase portion of the signal. Half of that should be extracted from each side channel. This seems strange to me. For the left channel this results in L - .5(L+R) = .5(L-R). and right channel is R - .5(L+R) = .5(R-L). So all three channels should be playing, and it's back to what I started this thread with, a matrix of L-R, L+R, R-L, except the side channels have to be turned down by .5.
This seems to work perfectly for things that are equal volume in left and right channel regardless of phase differences. But it doesn't work when things are panned left and right by volume. If left is playing alone, then the derived mix should also only play the left channel. If left is louder than right and in phase, then only left and center should play. So it seems what I might want to do is add steering logic to the simple summing matrix.
If only everything in a 2 channel recording were either pefectly in phase, or perfectly opposite phase, this would be easy. Or, if everything were equal volume but only varied by being in phase or out of phase by various degrees, this would also be easy. When there's volume panning and sounds that are various degrees out of phase it gets complicated. Almost all recordings have various degrees of in and out of phase content between channels as well as volume panning. Darn it!
Oops! It seems there's a problem here with math ideas not being properly applied. If L is equal in volume but opposite in phase to R, they will cancel out so .5(L+R) will equal zero, so L - .5(L+R) = L - 0, which equals L! So you can't treat the L in the (L+R) as equivalent to the stand alone L and jump to .5(L-R). But .5(L+R) is correct. Hmm.... You have to do the stuff in the parenthesis first and then complete the equation.
Well, it turns out they are equivalent. If the signals are out of phase, .5(L-R) = .5(L-(-L)) = L
Algebra really works, and my sanity is restored.
So this is different than my original matrix, but this one also assumes that the side speakers are placed apart in a normal listening triangle. And, it doesn't work correctly for any signals that have a difference in level from left to right.
I've owned both of those systems. They are what got me thinking about issues of crosstalk. What really got me going was when I tried a physical barrier in front of my face and heard what that could do when reasonably well dialed in. Now I'm thinking about Blumlein's point that coincident pairs of microphones where volume panning rules will produce the correct phase response across the head for all panned positions. So we really don't want to reduce crosstalk at the ears. The problem is comb filtering for any position that isn't hard panned to one speaker or the other. So more speakers means reduced comb filtering, and a more stable off axis image.
I've played with Dolby up-mixing and was extremely impressed with some settings, yet also not completely satisfied that it wasn't somehow degrading the sound. I didn't perceive that effect from native multi-channel recordings, so it seemed logical to me that it was an up-mixing issue. I kept returning to unprocessed 2 speaker playback for 2 channel recordings in the end, which makes me want to get more control over exactly what is happening with the up-mixing and see if I can come up with a result I can unequivocally say I prefer to just two speakers. Plus it's a fun project. If I can figure what I think ought to happen, then I get to learn how to write software that will do it for me.
I think I know what to do for certain situations, so I have a set of specific algorithms. I just can't figure out a universal algorithm that will take all phase and level scenarios into account in a way that seems coherent to me. There's one idea I have that's super simple and will work as I think it should on all in-phase or perfectly out of phase material. On stuff that's partially out of phase it will produce distortion that will have to be filtered out. That might be good enough! I'd have to try it and listen. I fear though that it's going to destroy some subtle phasing effects in a way that's going to be less than satisfying.
Last edited: