VideoHelp Forum

Our website is made possible by displaying online advertisements to our visitors. Consider supporting us by disable your adblocker or Try ConvertXtoDVD and convert all your movies to DVD. Free trial ! :)
+ Reply to Thread
Page 1 of 2
1 2 LastLast
Results 1 to 30 of 52
Thread
  1. Im wanting to create a stereo wave with matrixed surround information in it, for encoding to Pro Logic II. Someone on this board once said I have to do this to the surround channels:

    LS -->(50%L + PhaseInverted 50%R) w/ +90 PhaseShift (if you know how to do that, otherwise ignore the phaseshift)
    RS -->(50%R + PhaseInverted 50%L) w/ -90 PhaseShift (same as above)

    LS -->100%L, but down by -3dB
    LS --> 100%R, Down by -3dB, AND Phase Invert
    Same for RS

    What is "phase" in the world of audio, and why does Phase Invertion and Phase Shifting have to be used? What does this process do?

    Thank you as always.
    Quote Quote  
  2. Member Cornucopia's Avatar
    Join Date
    Oct 2001
    Location
    Deep in the Heart of Texas
    Search PM
    Well, I'm the one that gave you that info, so I might as well be the one to further explain...

    I won't go into the whole physics of compression of air, etc. I'll just try to clarify what you do with phase on a computer.

    Let's say you've got your sine wave file. Looks something like this:

    This would be like how the waveform displays in an audio editor. It puts each sample on the graph, depending on sample time (x axis) and sample volume/intensity (y axis). Notably, the y axis can be positive or negative (meaning "compressing" or "decompressing" the air relative to normal atmostpheric pressure--but I said we weren't going to get into that).
    Positive y = Positive analog voltage = Speaker Moving Outward = Compression,
    Negative y = Negative analog voltage = Speaker Moving Inward/Backward = Decompression
    (Sorry for that little bit more. )

    Now let's look at a different sine wave file.


    It looks just about the same. For most purposes, it is the same. One minor exception seems to be the initial delay. It's not really delay, it's "Phase Angle". Since the Sine function relates to the Phase Angle of a circle, where you start on the circle determines what the given intensity is at that point in time. This one is out-of-phase with the first one by a small amount.

    Look what happens when you superimpose to two:

    This more clearly shows the relationship between the 2.

    However, when mixed (digitally, electronically, or aurally), they combine to show a new wave that is the vector addition of the 2 waves:


    Because of the vector addition (which incorporates the phase angle) and because of the fact of having the possibility of Positive and Negative intensities, weird things can start to happen (called interference)...

    If they are identical in phase (basically just 2 copies of the same file), they have constructive interference and add together to be twice as loud.:


    If they are completely opposite in phase (i.e. "Reverse Polarity" or "Inverted"),

    then together they have destructive interference and subtract, creating a total wave of ZERO or complete silence.:


    Of course, realworld signals have multiple frequencies and harmonics and vary their levels, so nothing is so cut & dried, it's much more complex than that.

    Phase inversion/shifting allows you to do the complex mixing required so that you can add (matrix) files together at one stage, and (nearly completely) recover the original files at a later stage (decoding). Dolby Surround/PL/PLII is based on this priciple of 4:2:4 matrixing.
    There are other psychoacoustic uses for phase adjustment as well.

    Hope that clarified things...
    Scott
    Quote Quote  
  3. OK thats all very heavy, lol! So why are surround channels in a PL mix phase inverted and then phase shifted? Does this putting of the two surround channels out of phase create a more "open" sound field?
    Quote Quote  
  4. Member Cornucopia's Avatar
    Join Date
    Oct 2001
    Location
    Deep in the Heart of Texas
    Search PM
    2 reasons:

    1. the 90 degree phase shift is used so that simple circuits won't completely lose the surround signal when decoding (un-matrixing), there is still recovery!

    2. Yes, in most settings it does EXACTLY make it sound more spacious!

    Scott
    Quote Quote  
  5. Ah so I got one thing right lol. Can phase shifting be used on discrete surround channels when doing 5.1 encoding as well?
    Quote Quote  
  6. Member Cornucopia's Avatar
    Join Date
    Oct 2001
    Location
    Deep in the Heart of Texas
    Search PM
    Yes, 5.1 can make use of phase shifting, but with more channels comes more complexity. Believe it or not, Dolby EX uses matrixing on the back channels (using a form of phase shift) to get extra channel(s).
    With 5.1, it's at the point where it's quite unpredictable how phase shift will affect the final mix. Much more predictable is 5.1 surround panners, 5.1-randomized reverbs & delays, etc. Even HRTF effects are more predictable.

    Scott
    Quote Quote  
  7. Originally Posted by Cornucopia
    Yes, 5.1 can make use of phase shifting, but with more channels comes more complexity. Believe it or not, Dolby EX uses matrixing on the back channels (using a form of phase shift) to get extra channel(s).
    So in theory, I can make matrixed 6.1 tracks using phase shift? Is it possible?

    ...5.1 surround panners, 5.1-randomized reverbs & delays, etc. Even HRTF effects are more predictable.
    What are surround panners, and all those others you list?

    See, at the mo I do my 5.1 mixes without any phase shift on the rear channels. I do, however, apply this Audition setting to my music on the rears:



    The only thing IS, I dont know what thats really doing. Is that keeping the stereo elements of the music on the same channels, but reversing the (natural) surround encoded information contained within the audio?

    I also dont know what those numbers represent, I tried the Help file but it doesnt explain. I mean what does -50 mean on that scale, for instance?

    What would apply phaseshifting do if I put it on my rear channels in a 5.1 mix? Would it make the sound field more open and therefore more immersive?

    A few questions there for you, hope you dont mind!
    Quote Quote  
  8. Member 3dsnar's Avatar
    Join Date
    Jan 2006
    Location
    Proxima Centauri
    Search Comp PM
    Originally Posted by GavSalkeld
    Im wanting to create a stereo wave with matrixed surround information in it, for encoding to Pro Logic II. Someone on this board once said I have to do this to the surround channels:

    LS -->(50%L + PhaseInverted 50%R) w/ +90 PhaseShift (if you know how to do that, otherwise ignore the phaseshift)
    RS -->(50%R + PhaseInverted 50%L) w/ -90 PhaseShift (same as above)

    LS -->100%L, but down by -3dB
    LS --> 100%R, Down by -3dB, AND Phase Invert
    Same for RS

    What is "phase" in the world of audio, and why does Phase Invertion and Phase Shifting have to be used? What does this process do?

    Thank you as always.
    This is not true that you MUST do the 90 deg phase shifthing.
    It is important to mix the surround channels with certain procedure,
    and they must differ by 180 degrees. So you may shift the phase +90 and -90,
    or simply do nothing with the right channel and change sign of the signal in the left channel.
    The 90 deg shifts are computationally much more complex (you need to perform bunch of FFTs in order to calculate Hilbert transform, do overlapp-add procedure, etc.),
    and they do not enhance the channel separation efficinency whatsoever


    This is the normalized DPLII matrix downmixing procedure.
    Lt = 0.2645 L + 0.1871 C + 0.1871 LFE - 0.2291 BL - 0.1323 BR
    Rt = 0.2645 R + 0.1871 C + 0.1871 LFE + 0.1323 BL + 0.2291 BR

    The coeffs can be understood as gain changes (by certain amount of percent).
    E.g. LT
    take left channel and change its gain by 26.45%
    add to it center channel with gain changed by 18.71%
    add to it LFE with gain changed by 18.71%
    revert phase (change sign) of the surround left, change its gain by 22.91% and add to the rest,
    revert phase (change sign) of the surround right, change its gain by 13.23% and add to the rest.

    Do similar thing to create the right DPLII channel (with reference to the Rt equation),
    and that is all.

    More about optimum weights used for DPLII downmixing can be found here:
    http://forum.doom9.org/showthread.php?s=&threadid=57988

    BTW. Here is an example of 90 deg. phase shift of more complex waveforem (square).


    I hope this is helpful
    Cheers, 3d




    Quote Quote  
  9. Member Cornucopia's Avatar
    Join Date
    Oct 2001
    Location
    Deep in the Heart of Texas
    Search PM
    Originally Posted by 3dsnar
    Originally Posted by GavSalkeld
    Im wanting to create a stereo wave with matrixed surround information in it, for encoding to Pro Logic II. Someone on this board once said I have to do this to the surround channels:

    LS -->(50%L + PhaseInverted 50%R) w/ +90 PhaseShift (if you know how to do that, otherwise ignore the phaseshift)
    RS -->(50%R + PhaseInverted 50%L) w/ -90 PhaseShift (same as above)

    LS -->100%L, but down by -3dB
    LS --> 100%R, Down by -3dB, AND Phase Invert
    Same for RS

    What is "phase" in the world of audio, and why does Phase Invertion and Phase Shifting have to be used? What does this process do?

    Thank you as always.
    This is not true that you MUST do the 90 deg phase shifthing.
    It is important to mix the surround channels with certain procedure,
    and they must differ by 180 degrees. So you may shift the phase +90 and -90,
    or simply do nothing with the right channel and change sign of the signal in the left channel.
    The 90 deg shifts are computationally much more complex (you need to perform bunch of FFTs in order to calculate Hilbert transform, do overlapp-add procedure, etc.),
    and they do not enhance the channel separation efficinency whatsoever

    I hope this is helpful
    Cheers, 3d
    I respect your work a great deal (thanks, BTW), but I am old enough to remember Quad, and one of the problems with all of the different matrixing schemes at the time was the separation (or lack) upon decoding, and the sometimes weird panorama/soundstage dimension changes that might occur with different sources upon decode. Dolby (who bought out the Scheiber QS? license)incorporated the 90degree phase shift specifically to alleviate some of those problems.
    So, leaving it out should still be quite workable, but it does leave one open to the same type of anomalies as some of the Quad productions had.

    Scott
    Quote Quote  
  10. Thanks, guys, although all those 0.285867 things mean nothing to me lol. Can anyone answer the other questions I posted?
    Quote Quote  
  11. Member 3dsnar's Avatar
    Join Date
    Jan 2006
    Location
    Proxima Centauri
    Search Comp PM
    Originally Posted by Cornucopia
    I respect your work a great deal (thanks, BTW), but I am old enough to remember Quad, and one of the problems with all of the different matrixing schemes at the time was the separation (or lack) upon decoding, and the sometimes weird panorama/soundstage dimension changes that might occur with different sources upon decode. Dolby (who bought out the Scheiber QS? license)incorporated the 90degree phase shift specifically to alleviate some of those problems.
    So, leaving it out should still be quite workable, but it does leave one open to the same type of anomalies as some of the Quad productions had.

    Scott
    Hi Scott, thanx!
    Hmm, you may be right. I.e. probably some things has changed with the prologic technology
    develpment. I have tested with my HT amplifier the -90/90 phase shifting and the downmixing
    procedure suggested by Doom9 devs (the one that I've described), and both seem to be similarly effective on my equipment.
    But I think the 90 deg phase shift could have been of importance (could have been used as a source of information for channel separation), if the surround panorama, by its origin, would have had identical phase in each channel, and only different panning. I think this is the oldstyle way of creating 5.1 mixes, and many years ago Dolby developed their hardware under such assumption?
    Nowadays, reach 5.1 mixes (as you already explained to GavSalked) have much more complex phase relations between channels to enhance the 'spatiality', so the modern DPLII decoeders cannot assume any fixed phase shifts between the downmixed sources, and therefore I think (based on my experiments) that the downmixing procedure which uses the 90 deg. phase shifting could be obsolete.
    What do ya think?
    Cheers, 3d.

    180vs90.7z
    Quote Quote  
  12. Member 3dsnar's Avatar
    Join Date
    Jan 2006
    Location
    Proxima Centauri
    Search Comp PM
    Originally Posted by GavSalkeld
    Thanks, guys, although all those 0.285867 things mean nothing to me lol. Can anyone answer the other questions I posted?
    What is the other question?
    Please read my post carefully. I have explained what (for example) 0.2645 means.
    It means changing gain by 26.45%, and I think you can do it with any wave editor,
    including the one that u use.

    Actually, the VST channel mixer seems to be very much suitable for that.
    Can you mix 6 channel sources with it?
    If yes, please do the following
    ----------------------------------------------------------------------
    CREATING DPLII LEFT CHANNEL (sliders settings)
    26.45% for input LEFT
    18.71% for input CENTER
    18.71% for LFE
    -22.91 for SURROUND LEFT
    -13.23 for SURROUND RIGHT
    =======================================
    CREATING DPLII RIGHT CHANNEL (sliders settings)
    26.45% for input RIGHT
    18.71% for input CENTER
    18.71% for LFE
    13.23% for SURROUND LEFT
    22.91% for SURROUND RIGHT
    ---------------------------------------------------------------------

    So you can see that all 6 channels are utilized in order to create the new stereo source.
    If you cannot set gain in % so precisely, the rounded values are still fine (e.g. 26% instead of 26.45%, or -23% instead of -22.91%, etc, etc.)
    I hope this will work.
    Cheers, 3d

    fifthelem_6chnl.7z

    Quote Quote  
  13. Sound level are normally expressed in dB, because it is in log scale.

    These % numbers have an absolute sum of 100%. So are these numbers basically, a normalised dB scale ?

    Is the mixing ratio came from experimental result, or based on calculation with phase and magnitude then normalised ?
    Quote Quote  
  14. Member 3dsnar's Avatar
    Join Date
    Jan 2006
    Location
    Proxima Centauri
    Search Comp PM
    Usually gain changes are represented in two ways,
    in dBs or in % (see the picture provided by GavSalkeld).
    % refers only to a multiplication factor. I.e. changing gain by 50%
    means multiplying signal by 0.5 (changing gain by 10*log10(0.5^2) = -6 dB).

    I guess the mixing ratios come from information provided by dolby and other sources.
    But there have been some confusion, so
    they were also verified empirically by some poeple:
    http://forum.doom9.org/showthread.php?s=&threadid=57988

    I used the '4.8 dB' downmix, as suggested, but with some normalization
    (to avoid clipping)

    squares.7z
    Quote Quote  
  15. For modulated signal, phase is part of a strict signaling sheme. I always wonder how phase is considered in audio, which is a base band signal with no rules. A phasor diagram will be impossible to describe audio signal.

    The "pair of ears" are modeled as receiver, that are often placed in an pre-determined location in a room, which are often not replicated in real-life. It does seem to include quite a bit of hand waving. I am pleasant surprised that with all these conditions that we managed to get surround sound that feel so real.
    Quote Quote  
  16. Originally Posted by 3dsnar
    Usually gain changes are represented in two ways,
    in dBs or in % (see the picture provided by GavSalkeld).
    % refers only to a multiplication factor. I.e. changing gain by 50%
    means multiplying signal by 0.5 (changing gain by 10*log10(0.5^2) = -6 dB).
    10*log(Out/In) is used to compare ratio of two voltages. Why not use 20*log(Out/In) for Sound level, that I though sound (aka moving air) is power scaled ?

    Originally Posted by 3dsnar
    I used the '4.8 dB' downmix, as suggested, but with some normalization
    (to avoid clipping)
    0.707, 0.86, are all square law of 0.5, 0.75..., so it is a take of how sound attenuate due to distant. aka power scaled.
    Quote Quote  
  17. Member 3dsnar's Avatar
    Join Date
    Jan 2006
    Location
    Proxima Centauri
    Search Comp PM
    I wrote 10*log10(0.5^2) = 20*log10(0.5)
    The downmixing weights refer to amplitude (not energy), thus they can be represented by %
    I.e. as u you noticed, multiplying signal by 0.5 changes its energy by
    0.5^2
    ---
    But the weight values are not that important for the prologic concept.
    But the DPLII decoder assumes that they are given, therefore you have to use those that I provided (so the decoder will work properly).

    Ofcourse to enjoy fully the 5.1 panorama, you must stay in the sweet spot,
    so there is an assumption of your position against speakers


    The prologic does not have anything to do wigh source direction perception.
    This is just a procedure which allows to roughly regenerate 6 (5 actually) signals from the stereo source and feed them to 6 (5) speakers.

    Simulating 5.1 on headpohnes (stereo source) is a HRTF bussiness, but that is a different story.
    Quote Quote  
  18. [quote="3dsnar"What is the other question?[/quote]

    Well if you scroll up, you cans see it was a response to an earlier post, whereby I asked the following (please see the original post for the context these were posted in):

    So in theory, I can make matrixed 6.1 tracks using phase shift? Is it possible?
    ...
    What are surround panners, and all those others you list?
    ...
    The only thing IS, I dont know what thats really doing. Is that keeping the stereo elements of the music on the same channels, but reversing the (natural) surround encoded information contained within the audio? I also dont know what those numbers represent, I tried the Help file but it doesnt explain. I mean what does -50 mean on that scale, for instance?


    Thanks fellas.
    Quote Quote  
  19. Member 3dsnar's Avatar
    Join Date
    Jan 2006
    Location
    Proxima Centauri
    Search Comp PM
    OK.
    Yes, it is possible to generate 6.1 sound (with surround center),
    or 7.1 (with surround side and surround back)
    or 8.1 (3 fronts, 2 side, and three back) speakers.
    Actually, the DPLII downmix is sufficient for this.
    In fact, I could extend our (i.e. aud-x direct show filter)
    DPLII decoding algorithm to 6.1, 7.1 or 8.1,
    but I do not think that there is demand for this,
    (nearly nobody has so many speakers connected to a PC)
    so I decided not to waste time for such experiments...maybe in the future
    ---
    -50% means multiplying the signal by -0.5. This means reverting
    the signal (changing its sign), and changing amplitude by half (50%).
    So it is not a rocket science
    ---
    Panning (simple stereo example):
    If you have a monophonic sound. and you copy it to both channels, while listening from stereo speaker set you will perceive it as coming from a center (the same level from both speakers, with identical phase). Now, if you would start to smoothly increase the amplitude in the left channel and decrease the amplitude in the right channel, you would perceive it as the source moving towards the left speaker... etc.
    ---
    Why is the downmixing procedure so important in order to recreate the 5.1 panorama?
    In general, all these operations, that are performed during DPLII downmixing allow to analyze the signal complex spectra (for left and right channel) and with some DSP operations 'reverse' the downmixing procedure. The decoder assumes that certain weights were used during downmixing, and therefore it is so important to follow the procedure that I've described (i.e. the exact gain manipulations are so important)
    The 'reversing' does not allow to retrieve the original perfectly,
    due to many ambiguities, but still allows for some basic 5.1 panorama reconstruction.
    Therefore DPLII should be viewed rather as some cheap alternative to a proper digital multichannel format.
    The decoding is pretty much complicated, and if you are not DSP expert, it will be difficult to understand it.
    Please do some tests with downmixed 5.1 signals on your HT amplifier, or with the use of our DirectShow filter to find out how it sounds, and how it differs from original 5.1 source.
    Quote Quote  
  20. Originally Posted by 3dsnar
    -50% means multiplying the signal by -0.5. This means reverting
    the signal (changing its sign), and changing amplitude by half (50%)
    Well setup picture you saw there that I posted was applied to a stereo form and it didnt sound like a simple channel mix. The audio sounded all... weird (for want of a better word!) whereby the audio sounded like it was "in my head". Why did it do that?

    Also, does anyone know how I could matrix a rear surround channel into my 5.1 discrete mixes that I do? If it is possible, Id sure like to know how to do it.
    Quote Quote  
  21. Anyone got any ideas at all?

    Surely a rear surround channel would be possible to matrix into the surround channels easily?
    Quote Quote  
  22. Member 3dsnar's Avatar
    Join Date
    Jan 2006
    Location
    Proxima Centauri
    Search Comp PM
    Hmm, it would be the best to have a surround mixer, where you can adjust level of each channel, according to the procedure that I've described.

    If you do not have a such a mixer, I suggest to split the 5.1 signal into 6 mono waves,
    and apply the amplitude modification to each wave. Than downmix them to create the DPLII left channel and DPLII right channel.

    Cheers, 3d
    Quote Quote  
  23. Im talking about 6.1 sound, mate. Wondering if a matrixed centre surround channel can be feed into the LS and RS.
    Quote Quote  
  24. Member 3dsnar's Avatar
    Join Date
    Jan 2006
    Location
    Proxima Centauri
    Search Comp PM
    Sorry for late reply.
    Yes. Exactly. to create the surround channels (stereo) from 2+1
    mix the center to the left and right identically as it is done for the front 2+1.
    Cheers, 3d
    Quote Quote  
  25. Originally Posted by 3dsnar
    Yes. Exactly. to create the surround channels (stereo) from 2+1
    mix the center to the left and right identically as it is done for the front 2+1.
    Cheers, 3d
    OK, just to be clear:

    If I do that in a 5.1 mix, that will make a 6.1 matrixed track?
    Quote Quote  
  26. The amount of phase and the type of mixed can be your own ratio if you are simulating matrix sound from stereo source.

    Sony and other 6.1 surround amplifiers can interperate stereo, 5.1 and 6.1 system.
    Quote Quote  
  27. Just to be absolutely clear: if I mix a mono signal with 50% going into both the LS and RS of a 5.1 mix, that will create a matrixed 6.1 mix, yes? Do I need to apply ANY processing to this matrixed centre-surround channel for it to work on a 6.1 processor?
    Quote Quote  
  28. Member 3dsnar's Avatar
    Join Date
    Jan 2006
    Location
    Proxima Centauri
    Search Comp PM
    I have no idea if there is a 6.1 DPLII decoder.
    Theoretically it should be possible though.

    Moreover, the DPLII downmix (in terms of information that it carries)
    allows for decoding also to 7.1
    (FrontL, FrontC, FrontR, SideL, SideR, BackL, BackR, LFE)
    and also to 8.1
    (FrontL, FrontC, FrontR, SideL, SideR, BackL, BackC, BackR, LFE)

    But I am not aware if there are algorithms to resolve this.
    Actually I could write something like that, as an extention to a regular DPLII decoder
    but I think that there is no big demand for such solution

    Cheers, 3d.
    Quote Quote  
  29. Originally Posted by 3dsnar
    I have no idea if there is a 6.1 DPLII decoder.
    Nooooo, Im not talking about PLII anymore now lol. I asked about a 6.1 discrete mix.

    Again, just to be clear:

    Speaking strictly about a discrete 5.1 mix, is it possible to matrix a centre surround channel into the surround channels of a 5.1 mix, therefore making a 6.1 matrixed mix (which I can then encode to DTS 5.1 ES or Dolby Digital 5.1 EX)?

    If so, what do I need to do?

    Do I need to apply ANY processing to this matrixed centre-surround channel for it to work on a DTS 6.1 home cinema system?
    Quote Quote  
  30. Member 3dsnar's Avatar
    Join Date
    Jan 2006
    Location
    Proxima Centauri
    Search Comp PM
    Hmm, I completely misunderstood you...
    DPLII is a so cald matrix based downmixing...
    So discrete is something different than matrix surround.

    So what do you mean by:
    "is it possible to matrix a centre surround channel into the surround channels of a 5.1 mix, therefore making a 6.1 matrixed mix (which I can then encode to DTS 5.1 ES or Dolby Digital 5.1 EX)"

    Do you want to add a new channel, creating 6.1 discrete multichannel signal,
    or do you want to downmix the center surround channel to the surrounds L and R (of a 5.1 mix), resulting in (still) 5.1 signal, but containing audio from the center?

    Anyway, I am not a DTS expert.
    Maybe lack of responce on this forum is caused by the fact
    that nobody understands what you really mean...
    You can also try on Doom9 forum.
    Cheers, 3d.
    Quote Quote  



Similar Threads