Skip to main content
Topic: Downmixing 5.1 Music to Stereo (Read 16607 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Downmixing 5.1 Music to Stereo

I have some SACDs with both 5.1 and stereo tracks. Sometimes the stereo tracks are highly compressed (DR5-7) while the 5.1 tracks are not (DR10-12).

I want to make good stereo versions from the 5.1 tracks, but it is not that simple:

After ripping the SACD with a Sony PS3 and converting to 6 channel wavs with foo_input_sacd I have tried the following:

A: Extracting left and right channel with eac3to and merging them with Wavewizard.
This sounds fairly good on some tracks, but not when there is a vocal in the center track.

B: Convert 5.1 to Stereo DSP in foobar2000
This introduce thousands of clipped samples.

C. -downStereo option in eac3to
Lose quite a lot of volume and doesn't sound right.
I think the surround channels are too loud.

Should I try other software for downmixing?

Should I reduce the volume of the surround channels first? If yes by how much?

Or should I remove the surround channels and just downmix 3.0 or 3.1. Should I include LFE?

Downmixing 5.1 Music to Stereo

Reply #1
If already using foobar2000, you could try the Channel Mixer component.

Should I include LFE?

I left it out of my downmixes as they come out too bassy otherwise, the front stereo channels already carry sufficient low frequency information.

For the surround channels, experiment with and without them and see which version you prefer. If it's a live album they generally consist of audience cheer and clapping, so you can leave them out. For studio albums however, you should include them. But there's no definitive answer...

Downmixing 5.1 Music to Stereo

Reply #2
With any matrix system (foobar Matrix Mixer or Channel Mixer, ffdshow), the regular formula (not normalized) is:
FL' = FL + 0.7071 x FC + BL
FR' = FR + 0.7071 x FC + BR

Or with SoX (normalized):
1v0.3694,3v0.2612,4v0.3694 2v0.3694,3v0.2612,5v0.3694

AFAIK eac3to is doing it right but it also detects clipping and so may automatically adjust the gain accordingly.
Dolby recommends not to add LFE to the downmix.
For the record, not SACD related, DTS may need phase shifting for rear channels in some cases.

Downmixing 5.1 Music to Stereo

Reply #3
I have done David Bowie's Heathen with plain "eac3to -downStereo". Great result.

But the back channels on Epica's 2005 album are way too loud. I have tried dropping them and the result is acceptable. It is not a live album, so maybe they should just be mixed with lower volume.

Downmixing 5.1 Music to Stereo

Reply #4
Quote
B: Convert 5.1 to Stereo DSP in foobar2000
This introduce thousands of clipped samples.
Mixing is done by summation.  That increases the levels and you can get clipping.  If your mixdown software is not automatically compensating for that, you'll have to normalize manually.  Almost all audio editors use floating point so they can temporarily/internally go over 0dB without clipping.  So the trick is to normalize before saving in an integer format.

On DVDs, there is metadata instructing the decoder how to mix-down....  It's up to the DVD producers.  The LFE* is always left out.  I don't know about SACD.    You may have to use your ear and your judgment.

If you think about how the sounds mix acoustically, the rear channels would be mixed equally with the front, but the center channel needs to be reduce by 3dB because copying it to both front-left & front-right without reducing it would double the center-channel sound.


For DVDs, I believe the normal default is -3dB for the center & rear channels, and 0dB for the front-left & front-right channels.


*  The LFE is for "booms & explosions".  The other 5 channels contain full-frequency audio including the "regular bass".    On most home systems (with small surround speakers), the receiver uses bass management to re-direct all of the bass to the subwoofer.

Downmixing 5.1 Music to Stereo

Reply #5
Hi, guys!
Need your help here.
I have a MKV movie file with 5.1 AC3 audio stream and I'd like to downmix it properly to stereo (to have good stereo sound on my stereo TV set and not low voices, loud music... you already know this).
Reading your advice, I'm trying to use FB2000 to do this... but What I get sounds very awful.
Now, I tell you what I do.

Using MKVToolNix, I can extract the audio part in MKA format BUT FB2000 seems unable to recognize this format.
Using VideoSplitter, I can extract the audio part in AC3 format BUT FB2000 seems unable to recognize this format.
I try to convert those files (both AC3 and MKA) with FormatFactory into AAC 5.1 format... it seems it works but I can't check the waveform into Audition to know if the general waveform of 6 channels is maintained.
Then I use those AAC files within FB2000 to have them stereo downmixed.
My FB2000 has those DSPs:
- channel mixer
- downmix AC3/DTS
- downmix channels to stereo

Using the last one, I obtain something weird, left very different from right... if I can make me clear. I don't have picture  of the waveform because those files weight about 1.4Gb... must delete if unneeded.
Then, using the Channel Mixer (output channels = 2,  downmix F1, C0.7, R1, Sw0.8... a little improvised), it looks quite the same. Here is a picture of what it looks like.


I don't know if the downmix is not done well by FB2000 and makes this not-equilibrated thing  OR  if it's just FormatFactory that already modifies the AC3 5.1  to AAC 5.1 into this not-equilibrated thing then downmixed by FB2000.

Sorry, I know I can't make myself very clear but I'm sure this forum is the place where I could receive some great help... am I wrong?

Hope to read you soon!


Re: Downmixing 5.1 Music to Stereo

Reply #6
I am replying to this because this is on page 1 of google for "downmixing 5.1 music."

I recently downmixed the 5.1 release of "(What's the Story) Morning Glory?" by Oasis for myself, since the stereo release is rather horribly brickwalled and compressed sounding. I did a lot of research into how to do this properly and I determined the following method to be "doing it properly":

Extract each channel to its own mono file, then mix them into a stereo file as follows:

Left channel to left channel
Right channel to right channel
Center channel down the middle evenly, -3 dB
LFE down the middle evenly, +10 dB
Left rear to left channel
Right rear to right channel

-3 dB seems to be an industry standard used by Dolby and the like for center channel downmixing. This is apparently the optimal value for compensating, volume-wise, for the fact that you are now reproducing something, designed to be played through one speaker, through two speakers at once. The same standards I saw using this -3 dB value had the left, right, left rear and right rear channels all equally loud and all 3 dB louder than the center channel.

Surround equipment automatically plays the LFE channel 10 dB louder than every other channel, so the surround channel is encoded into the 5.1 audio 10 dB lower than it should be reproduced. Hence, you boost it by 10 dB when downmixing.

Adobe Audition's multitrack mode makes this process quite easily and you can prevent the mix from clipping easily with its master volume control. I'm quite sure this method is possible with Audacity as well. I used foobar2000 to convert the 5.1 audio to a format that Audition could read.

Anyways, the result of this method was a truly excellent sounding downmix.

With regards to Dolby recommending not including LFE in the downmix, from what I gather the reasoning behind this is it can increase dynamic range too much to be practical for movies. Not including the LFE channel in my Oasis downmix resulted in a tonal balance almost utterly devoid of bass, whereas it sounds absolutely perfect (and I am by no means a basshead; I prefer a close to neutral frequency response for my headphones and speakers) with the LFE downmixed and boosted 10 dB, as this is obviously how it was intended to be heard through a surround system which automatically gives the encoded LFE data a 10 dB boost. I'm going to go ahead and recommend including that LFE channel at +10 dB for 5.1 music, but recommend keeping in mind that Dolby recommendation for TV/movies.

With regards to phase shifting mentioned by Brazil2, after doing some reading it seems this only applies to DTS ES.

Re: Downmixing 5.1 Music to Stereo

Reply #7
With any matrix system (foobar Matrix Mixer or Channel Mixer, ffdshow), the regular formula (not normalized) is:
FL' = FL + 0.7071 x FC + BL
FR' = FR + 0.7071 x FC + BR
FC = Center? BL and BR are surround-left and surround-right?
And can you provide a source or explanation for the sqrt(2)/2 factor for Center?


Re: Downmixing 5.1 Music to Stereo

Reply #9
Actually, correct formula is FL=FL+0.707*FC+0.707*SL
You can find it in ATSC (AC3) standart - https://www.atsc.org/wp-content/uploads/2015/03/A52-201212-17.pdf (read from page 95) and in ITU-R BS.775-3 - https://www.itu.int/dms_pubrec/itu-r/rec/bs/R-REC-BS.775-3-201208-I!!PDF-E.pdf
Sill makes me wonder how they came up with \frac{\sqrt{2}}{2}, though. I'm trying to think of some geometric reason, but that only makes sense, when each front speaker is half the distance away between the center and the surround speaker, on an arc, that is. But I doubt that's the case.

Re: Downmixing 5.1 Music to Stereo

Reply #10
Actually, correct formula is FL=FL+0.707*FC+0.707*SL
You can find it in ATSC (AC3) standart - https://www.atsc.org/wp-content/uploads/2015/03/A52-201212-17.pdf (read from page 95) and in ITU-R BS.775-3 - https://www.itu.int/dms_pubrec/itu-r/rec/bs/R-REC-BS.775-3-201208-I!!PDF-E.pdf

Huh. I guess I should be mixing in the left and right rear/surround channels at -3 dB then. After doing a bit more reading it appears as if the "standard" 5.1 setup has the front left and right channels at 30 degree angles and the left and right surround channels at 110 degree angles. I guess this 3 dB is to compensate for a signal coming from 110 degrees to the right sounding louder to the right ear than a signal coming from 30 degrees to the right.

Re: Downmixing 5.1 Music to Stereo

Reply #11
Aha, you are finding out that this is not as simple as it may seem, and you really have to do this by listening and mixing artfully for good sound, not simply applying a mathematical formula.  For instance, the amount of center channel "buildup" you gets depends on the coherence of the material in the left and right channels.  For an experiment, put a tone in both channels and look at the mono summed "center channel" signal.  It is 6 dB louder, not 3, as coherent signals sum 6 dB hotter.  If you have totally incoherent sound like different sources of white noise, or hard panned audio from different instruments, they will sum 3 dB louder, not 6.  So in the reverse, downmixing depends on how coherent the original audio that was mixed to make the center channel was.  If the center channel is the vocal, as is often true, you may find it too loud if mixed 3 dB down to left and right, an that 4, 5 or 6 dB sounds better to you. 

You might look at some editors that have selectable pan laws, because of the differences in how coherent and non-coherent signals sum.

Regarding the rear channels, that's even more of a crap shot.  Depends on what the original producer/mixer used them for, where the listener was assumed to be in the mix (in the middle of the band or back at the rear somewhere), and how much "special effect" or whatever the mixer/producer chose to put in the rear. 

LFE is not SUPPOSED to be used for bass management, the main channels in true 5.1 are supposed to be full bandwidth, but as you discovered, some producers either don't understand, or purposely mis-use it.

So the real answer is set up a mixer with stereo out, left and left rear panned full left, right and right rear panned hard right, center panned center, and listen to the result as you adjust the faders.  And yes, the summing will add level so you'll have to re-normalize or loudness compensate it before saving it in a fixed point format to avoid clipping.  You will probably actually have to pull the master down while you are mixing to avoid clipping in the D to A converter in your monitor system while mixing.

And, oh yes, you may want to put the LFE on a fader and see if it sounds better with that added to the mix, but don't get fooled by the psychoacoustic 'more bass always sounds better' effect.

Re: Downmixing 5.1 Music to Stereo

Reply #12
LFE is not SUPPOSED to be used for bass management, the main channels in true 5.1 are supposed to be full bandwidth, but as you discovered, some producers either don't understand, or purposely mis-use it.
I think you probably didn't mean to include LFE as one of the main, full-bandwidth channels, but it might have sounded that way.  So just to be clear, LFE is <120Hz, and is not available as a full-bandwidth channel.
https://www.dolby.com/uploadedFiles/Assets/US/Doc/Professional/38_LFE.pdf

 
SimplePortal 1.0.0 RC1 © 2008-2018