Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[dvogel1]

I have seen several references to this output option but I don't fully understand what it does.

I am using DSP Studio (Parametric Equalizer) filtering for a multi-way stereo speaker system (4 drivers per side for a total of 8 channels). The filtering works perfectly for stereo sources but, of course, does not work correctly for 5.1 sources and I can't find a combination of settings to convert 5.1 to stereo.

I have made several attempts to utilize the "2 channels (inside a 7.1 channel container)" output format option. I have this option selected together with "JRSS mixing".

When I play a 5.1 DVD video Audio-Path indicates: "Convert from 6 channels to 8 channels". The output is no different from the "7.1 channels" option and the center channel (voices) are very low. Should the conversion be from 6 channels to 2 channels?

Wish list items:
In the DSP Studio/Output Format section it would be beneficial to have an option to mix any multi-channel source (2.1, 5.1, etc.) to stereo. Also, please add the ability to rename the channels.

Thanks,
Don

[6233638]

I think JRSS is what mixes everything to use all your channels. Try setting 7.1 and disabling it.
If you want to downmix to stereo, try a 2 channel output.
It does seem like this option is not working as intended though.

[dvogel1]

My standard setup is 7.1 and no mixing. That works as expected for stereo sources.

If 2 channel output is selected it is not possible to send any data to the other 6 channels.

Agreed, "2 channels (inside 7.1 channel container)" does not work as explained in other posts. Maybe there is a basic misunderstanding about it's purpose.

[6233638]

Oh I'm sorry, I read over your post quickly and did not properly understand what you were trying to do.
It's a 2 channel (stereo) system which is split over 8 channels.
 
The "2 channels inside a 7.1 container" should only send audio to the front left/right channels, which doesn't help with your setup.
 
 
I think you will have to use an 8 channel output and use the parametric EQ to downmix everything, with JRSS disabled.
I'll look things up for a proper downmix and see if I can come up with a solution.
 
It's a bit difficult for me to test though, as I only have a 2 channel setup.

[dvogel1]

It would be helpful to know exactly what "2 channels (inside 7.1 channel container)" does when playing a multi-channel (5.1, etc.) recording. I have wasted a lot of time trying to figure it out.

[mwillems]

It would be helpful to know exactly what "2 channels (inside 7.1 channel container)" does when playing a multi-channel (5.1, etc.) recording. I have wasted a lot of time trying to figure it out.

"2 channels inside a 7.1 container" downmixes to stereo, and then adds six blank channels.  It shows "converting 6 channels to 8 channels" with a 5.1 source because the output of the downmix is eight channels, it's just that the last six are blank.  

I have a bi-amped setup with a stereo sub, but I like to control my sub routing so I use "2 channel in a 5.1 container" with JRSS mixing.  It downmixes all audio to stereo, and gives me four blank channels. Then I do my channel routing/sub-mixing in parametric equalizer.  "2 channel in 7.1 container" with JRSS mixing should do exactly what you want: everything downmixed to stereo with six blank channels so you can do your channel routing and crossovers in PEQ.  The center channel does occasionally sound a little soft in my setup, but is normally just fine.

If the center channel sounds very soft that probably means that the center channel isn't getting downmixed correctly (i.e. if JRSS mixing is off that will happen, or if something is glitched with the setting).  I can confirm that "2 channel in a 5.1 container" works exactly as expected, and all downmixing is handled correctly.  I just did a quick test with 2 channel in a 7.1 container with JRSS mixing on, and it seemed to work correctly in my setup, but I haven't "lived with it" like you have.

[6233638]

OK, I think this is what you need for your 7.1 downmix.
Use a 7.1 output and disable JRSS
 
 
The "Copy Left to ___" and "Copy Right to ___" assumes that the correct channels are being sent to your left speaker and right speaker, change those as necessary.
 
 
EDIT: Or if I am correctly understanding what mwillems has written, you would simply be able to use "2 channels inside a 7.1 container" with JRSS enabled and copy the channels, without everything else in the Parametric EQ I listed. (the downmix)

[mwillems]

EDIT: Or if I am understanding what mwillems has written, you would simply be able to use "2 channels inside a 7.1 container" with JRSS enabled and copy the channels, without everything else in the Parametric EQ I listed. (the downmix)

That's exactly what I do.  I use 2 channel in a 5.1 container, and copy my left and right onto the blank channels (S, C, SL, SR) and then do my crossovers, etc.

[6233638]

That's exactly what I do.  I use 2 channel in a 5.1 container, and copy my left and right onto the blank channels (S, C, SL, SR) and then do my crossovers, etc.

Well that certainly simplifies things, without having to calculate what the correct downmix from 7.1 to 2.0 is...

[mwillems]

Well that certainly simplifies things, without having to calculate what the correct downmix from 7.1 to 2.0 is...

It's the point of the x channels in the x container options.  It allows you to specify the up or downmixing target while still having blank channels for custom mixing.  I kind of wish that the mixing target and output container were able to be specified separately, as there are currently some unachievable combinations that could be useful (i.e. 2.1 in a 4-channel container would be useful for some setups, or 4 channel in a 7.1 container, etc.).

[dvogel1]

When I use this output format the center channel (voice data) is not present in the down-mixed 2 channels, with or without JRSS mixing enabled. The remaining stereo content is present but voices are very low (10% of the other content).

I have experimented with "adding" the center channel to the L and R channels before the other filtering occurs. In this configuration the voices are present at normal levels, however, the voices float around. By float I mean that the voices move from left to right to center and back in a random way.

[dvogel1]

Here is a partial listing of the filtering that I have configured. Left and Right initially contain the full stereo signal, that is, with a true stereo source.

L/R = left and right woofers
Center/Sub = Left and right lower midranges
RL/RR = Left and right upper midranges
SL/SR = Left and right tweeter

[mwillems]

When I use this output format the center channel (voice data) is not present in the down-mixed 2 channels, with or without JRSS mixing enabled. The remaining stereo content is present but voices are very low (10% of the other content).

I have experimented with "adding" the center channel to the L and R channels before the other filtering occurs. In this configuration the voices are present at normal levels, however, the voices float around. By float I mean that the voices move from left to right to center and back in a random way.

That isn't normal (at least not in my setup); I've only had the "missing" center you described when I had JRSS turned off, so something is going wrong in the downmixing.  It's worth figuring out if it's not downmixing at all, or if there's just some issue with the way the mix is working.  

I'll suggest a few troubleshooting steps:

1) A quick experiemnt (which you may have already tried). Try setting the output channels to 2 channels in a 7.1 container, enable JRSS, turn off parametric equalizer and all other DSP, and then mute your amplifier(s).  Then play a 5.1 file, and have a look at analyzer: do you see seven lines, or just three?  If you see three, that's the two stereo channels and the "top" line, which means the downmix is working at least that far.  If you see seven, that means no downmixing is occurring and something is glitched.  

2) I've found that output format changes don't necessarily "take" unless media is stopped and restarted.  I assume you've tried making the changes when nothing is playing or paused, but I thought it was worth a check.  

3) Could you post a screencap of your output format settings?

[dvogel1]

1) With the PEQ off there are just 3 output lines (RMS, left and right). That confirms that the remaining 6 channels have no data (good).

Update from previous post: when I added the center channel to the L and R channels I must have had only the 7.1 channel option selected (no down-mixing). Sorry to mislead the investigation.

Conclusion: the 2 channels (inside 7.1 channel container) option is blanking channels 3 through 8 but not mixing properly (see A below).

2) Yes, I do toggle playback when making setting changes (other than PEQ filter changes). For overkill I frequently shutdown MC and restart.

3) Please review the attached output format information.

New data:

A) I set the output format to "2 channels (inside 5.1 channel container)". In my configuration the filtered tweeter output is missing (channels seven and eight) but there is enough information in the remaining output to figure out what is happening. In this configuration the center channel is present, however, it floats around like I described earlier. Based on this test I think that the mixing for the 7.1 option is incomplete.

This raises a new question. Why does the mixing produce sound that floats? If this is normal, expected behavior then I may be trying to implement a configuration that won't be acceptable. I never experienced anything like that when playing 5.1 material through my CD/DVD player using only L/R stereo output.

[6233638]

It looks like there is a bug with the "2 channels in a 7.1 container" option - I just tested it using the converter.
 
7.1 files are downmixed correctly* but 5.1 tracks only output the left and right channels.
Using the custom downmix I specified in my previous post with the output set to 7.1 and JRSS disabled works correctly.
 
 
 
*The mix does not seem to be normalized and could potentially clip. Same goes for 5.1.
An option to normalize the mix would be nice.

JRSS stereo downmix on top, LAV stereo downmix on the bottom.

[dvogel1]

I moved my trouble shooting to another computer to help eliminate variables. This system has a conventional 2 speaker audio system, that is, no PEQ filtering is used. The PC motherboard outputs to a stereo receiver. I have the receiver connected to only the L/R (green) output.

With 2/7.1 + JRSS mixing selected the center channel (voice) data is very low.

With 2/5.1 + JRSS mixing selected the center channel data is present at normal volume.

Clearly 2/7.1 is not mixing correctly.

More importantly, with 2/5.1 + JRSS mixing selected the center channel (voice) data does not float. That implies that the mixing is not working within the PEQ system properly.

[6233638]

With 2/7.1 + JRSS mixing selected the center channel (voice) data is very low.

The center volume isn't low, it isn't present. The option is bugged when you play back 5.1 tracks, and only the left and right channels are used.
 
The 2 channel in 5.1 option works correctly.
 
 
Edit: attached images of the bug. (every 3 seconds the channel changes; L,R,C,S,Ls,Rs,Lr,Rr)

[Hendrik]

I'll check in the coming week whats up with the 2 in 7.1 option, its probably an easy fix somewhere.

[dvogel1]

The 2/5.1 option does not work correctly with the PEQ. With the PEQ enabled with multiple filters the center channel data floats.

[mwillems]

Glad we've started to get to the bottom of it.  It looks like the 2 in 7.1 wasn't working correctly just with 5.1 sources, which would explain why it seemed to work for me when I tested it yesterday (I was using a 7.1 source).  

The 2/5.1 option does not work correctly with the PEQ. With the PEQ enabled with multiple filters the center channel data floats.

That hasn't been my experience with that setting.  I use quite a few PEQ filters (obviously I have a bi-amp filter bank not that different from yours), and haven't experienced the center channel "floating" like you're describing.  The voice data does move around a little, but it's typically things that are in the original surround mix (i.e. it's always correlated to what's happening on the screen, voices will sound like they're coming from the left when the person speaking is at the left edge of the screen, etc.). That's achieved in the original mix by moving some of the dialog out of the center channel and into one of the mains; that kind of thing is why you heard some vocal content when the center was entirely missing (not all voice is in the center channel).  But I'm guessing that's not what you mean when you say the center info is "floating around."

I can think of two avenues for inquiry:

1) It may be that the center channel mixing in 2/5.1 container is glitched, and I haven't noticed it in my own setup over the last year.  That's entirely possible: my speakers are only about 6 feet apart and my listening position about 12 feet away, so I tend to get a slightly "too stable" phantom center.  

You mentioned above that you got the "float" even when you didn't use JRSS, but just mixed the center into the two sides manually using PEQ, right?  An easy way to test would be to change output format to 7.1 with no up or downmixing.  Then, copy the center channel to two stereo channels from your setup (overwriting what was already there).  Then mute all other channels.  You could then just listen to the center channel info by itself in stereo on one "stage" of your setup and see if it still floats.  If I were you, I'd try it with different stages of my setup (i.e. try it with the two vocal range drivers first, but then with some of the other driver stereo pairs as well).  Whether it floats or not under those circumstances, it will help isolate what is happening when; if it floats with nothing but the center channel playing you may want to investigate what I mention in the next paragraph.

2) When you run your setup with conventional stereo music does the "stereo illusion" typically work from your listening position (i.e. does most of the sound seem to be coming from directly in between the speakers, rather than seeming to be localized at each speaker)?  If not that may be the issue; the Haas precedence effect makes having a stable phantom center challenging if you're different distances from your speakers, or the timing between stereo channels is off to some extent.  http://en.wikipedia.org/wiki/Precedence_effect.  In that situation even slight variations in the "position" of sound in the mix will cause the center to "collapse" into one side or the other, which causes stereo content that should sound "centered" to "jump" around from left to right to center, etc.  In your case, it might be more noticeable when the mid-high frequency drivers are engaged because HF is much more positional and is a big part of how we "spatialize" sound.  That would also possibly account for why the issue "went away" when you tried different speakers (presumably located in a different room with a different listening position, etc.)

I know I'm probably not telling you anything you don't already know, but I wanted to mention this issue because it could cause the kind of issue you're describing (in case you hadn't already thought about it and ruled it out).

[Hendrik]

The mixing will be fixed in the upcoming build.

*The mix does not seem to be normalized and could potentially clip. Same goes for 5.1.
An option to normalize the mix would be nice.

The mixing matrix is actually normalized, however its actually smart and uses the per-channel True Peak information if analyzed, so that the normalization is only applied as strong as it needs to be to not clip, to preserve as much volume as possible.
In short, it should never clip due to mixing.

You can probably see the difference if you clear the peak level field.

Of course mixing also works smart with volume leveling, and any volume it removes in mixing is accounted for in volume leveling and not removed again.

Edit:
Actually, seems the smart normalization is broken (until the next build), it should apply full normalization all the time right now.

[csimon]

I kind of wish that the mixing target and output container were able to be specified separately, as there are currently some unachievable combinations that could be useful (i.e. 2.1 in a 4-channel container would be useful for some setups, or 4 channel in a 7.1 container, etc.).

+1

[dvogel1]

2) When you run your setup with conventional stereo music does the "stereo illusion" typically work from your listening position...

Yes, the stereo illusion is rock solid.

You mentioned above that you got the "float" even when you didn't use JRSS, but just mixed the center into the two sides manually using PEQ, right?

Yes, I used Add.

An easy way to test would be to change output format to 7.1 with no up or downmixing.  Then, copy the center channel to two stereo channels from your setup (overwriting what was already there).

With this configuration the sound is perfectly stable and centered. Note that I did not make any changes to the PEQ filtering.

[mwillems]

With this configuration the sound is perfectly stable and centered. Note that I did not make any changes to the PEQ filtering.

So you mean that you conducted the "copy and mute" experiment as described, but left the rest of your filters in place (just verifying the methodology)?  If that's the case, then to summarize (please correct anything that's incorrect in my restatement):

1) Using JRSS downmixing along with subsequent PEQ filtering gives you an unstable center channel, and
2) Turning off JRSS downmixing, and manually adding center channel information to the mains in PEQ gives you an unstable center channel, but
3) Leaving JRSS downmixing on, but disabling PEQ, gives you a stable center channel (with a different set of speakers), and
4) Turning off JRSS and copying the center channel to a stereo pair in PEQ (overwriting what was already there) and listening to just the center channel in stereo gives you a stable center image.

That seems to suggest that there's something specifically goofy in the channel summing (whether JRSS or PEQ derived), but not all the time (see number 3), and not necessarily in other channel mixing options, like copy (see number 4).  If it weren't for number 2), I'd conclude that the issue was somewhere in the interaction between JRSS and PEQ, but number 2) rules that out.  That is indeed odd, and hopefully will give the devs something to go on if the channel summing is bugged.  

I tried last night to replicate the issue, but my center remains stubbornly centered while using both JRSS and substantial PEQ filtering. and I can't think of any other things to rule out.  

Any other folks who use stereo downmixing and/or bi-ampers (or tri-ampers, or quad-ampers) out there having similar issues with the center?  

Hopefully the devs will have some ideas.

[Hendrik]

Mixing is completely independent of the PEQ. The output format conversions including resampling and mixing are always performed long before the audio reaches the PEQ. Mixing does not change when you add a PEQ, so something in your PEQ must be causing this.
Since you have a rather long list of filters as it seems, I would suggest to try to isolate it step by step, instead of just enabling/disabling PEQ as a whole.

JRSS downmixing just merges the center into the L/R channels with a -3dB attenuation, there is no other special processing here.
If even manually adding the center to your mains causes the same symptoms, I'm somewhat inclined to blame either your source audio or something in your audio setup. The mixing code is really quite straight forward.

[mwillems]

Mixing is completely independent of the PEQ. The output format conversions including resampling and mixing are always performed long before the audio reaches the PEQ. Mixing does not change when you add a PEQ, so something in your PEQ must be causing this.
Since you have a rather long list of filters as it seems, I would suggest to try to isolate it step by step, instead of just enabling/disabling PEQ as a whole.

JRSS downmixing just merges the center into the L/R channels with a -3dB attenuation, there is no other special processing here.
If even manually adding the center to your mains causes the same symptoms, I'm somewhat inclined to blame either your source audio or something in your audio setup. The mixing code is really quite straight forward.

Hmm, if that's the case then I'm stumped. Hopefully dvogel can step through the filters and figure out what might be causing it.

If it's not the channel mixing, and it's not the speaker placement, that only really leaves asymmetrical volume or timing fluctuations (whether due to an unfortunate combination of PEQ filters or something else in the signal path) as a potential cause of an unstable center image.  I'm not sure what could be causing variable asymmetrical volume or timing fluctuations that wouldn't also lead to much more noticeable issues with stereo playback almost immediately.  Maybe a limiter somewhere in the PEQ chain, combined with some boost that isn't offset (or isn't offset until after the limiter)?  Just brainstorming.

[dvogel1]

Fixed parameters: 5.1 source, 7.1 channels, JRSS Mixing

1) PEQ disabled, simple L/R speaker setup

     Move center to front L/R enabled: Result has stable center channel
     No center move: Result has no center channel

2) PEQ enabled, simple L/R speaker setup with minimal filtering

     No center channel mixing: Result has no center channel
     Center channel added to L and R before filters: Result has stable center channel

3) PEQ enabled, filtered speakers on L and R, filtered speakers on Center and Sub

     No center channel mixing: Result has no center channel

     Center channel added to L and R before filters
     a) L and R copied to Center and Sub
     b) L and R filtered
     c) Center and Sub filtered
     Result has mostly stable center (this may be normal)

4) PEQ enabled, full filtering setup

     No center channel mixing: Result has no center channel
     Center channel added to L and R before filtering: Result has unstable center

Please keep in mind that adding the center channel in the PEQ is not normal. I expect that when the mixing is repaired the floating center will stabilize.

[dvogel1]

Please review attached PEQ

[mwillems]

Please review attached PEQ

I don't think this is likely to be causing the center channel issue, but just out of curiosity, how do you offset all that boost?  You've got quite a bit of net gain, and not enough offsetting level reductions to avoid clipping.  For example, those two 50Hz shelving filters seem to be adding 20dB of gain that is only partially offset by the reductions at the top.  

With some volume settings that would cause clip protection to engage pretty frequently. Do you use internal volume with a maximum volume setting or something else?

[dvogel1]

I don't think this is likely to be causing the center channel issue, but just out of curiosity, how do you offset all that boost?

3) and 4) almost lead me to think it is a process update issue because the floating center got considerably worse with the full PEQ active.

Yes, I use internal volume with a limit and I'm working on some modifications to address the gain.

[6233638]

The mixing will be fixed in the upcoming build.

I took another look at the downmixing again tonight after seeing some discussion on the subject in the LAV Filters topic over at Doom9, and it looks like the LFE channel is 3dB too loud in a regular stereo downmix (equivalent to 2.24 or +7dB rather than 1.59 or +4dB in LAV Audio) and the surround channels are being mixed at 1.0 rather than 0.7071 (which is valid but not standard)

The mixing matrix is actually normalized, however its actually smart and uses the per-channel True Peak information if analyzed, so that the normalization is only applied as strong as it needs to be to not clip, to preserve as much volume as possible.
In short, it should never clip due to mixing.

You can probably see the difference if you clear the peak level field.

Of course mixing also works smart with volume leveling, and any volume it removes in mixing is accounted for in volume leveling and not removed again.

That seems like a very good solution.

[dvogel1]

I have tested the 19.0.124 update with a 5.1 video. The center channel is present. That's good. However, the center channel still floats. For example, the voice of a person speaking at the center of the video image sounds like it is properly centered for a second or two and then shifts left or right for a couple of seconds and then back to center, etc. Also, the overall volume seems to be attenuated by several dB.

[6233638]

However, the center channel still floats. For example, the voice of a person speaking at the center of the video image sounds like it is properly centered for a second or two and then shifts left or right for a couple of seconds and then back to center, etc.

I think this is the video you're watching rather than anything MC is doing.

Also, the overall volume seems to be attenuated by several dB.

The more channels you mix into a stereo signal, the quieter it has to be to prevent clipping. (distortion)

[dvogel1]

I think this is the video you're watching rather than anything MC is doing.

Please realize that I'm not going asking unsubstantiated questions.
It's the same for all. You identified a 3 second rotation of some channel (center?) in a previous post although I can't find the post. Has that been fixed with this release?

Perhaps I should back up and ask a simple question. Does this format produce a simple, stable stereo output from a multichannel source? If it contains some remnants of 5.1 surround sound them I'm wasting everyone's time.

The more channels you mix into a stereo signal, the quieter it has to be to prevent clipping. (distortion)

Why is the volume lower than is was previously when it was essentially the same as music playback?

[mwillems]

Perhaps I should back up and ask a simple question. Does this format produce a simple, stable stereo output from a multichannel source? If it contains some remnants of 5.1 surround sound them I'm wasting everyone's time.

It does for me (although I'm not sure what you mean by "remnants" of surround sound).  I get a stereo output with any of the 2 channel in a container options: the sound pans appropriately when the source of the sound in the film is moving, so in that sense it retains a "remnant" of surround sound, but my own phantom center is stable when the persons talking are stationary in the center of the screen.  That means that one of two things is the case:

1) It's possible that there's something goofed up in the channel mixing and that your system is more revealing of stereo flakiness than mine is.  

2) There is something else causing variations in volume or timing in your system.  Your earlier investigations showed that manually adding in the center to the lower frequency channels created a stable center, but that the center destabilized as you added more channels in.  That to me suggests that there may be something odd about the interaction of some of the stages in your multi-way (the last two pairs).  

A suggestion: with output format set to 2 channel in a 7.1 channel container, and JRSS mixing enabled, and everything else set up the way you normally use it (i.e. full PEQ filtering), try muting L,R, sub, and center and see if the phantom center still floats with just the upper channels (assuming there's enough information left there for spatialization).  Then maybe try muting a few more combinations of pairs. If it floats with some combinations but not others, you might be closer to diagnosing what's going on.  If it is stable, or mostly stable, no matter which two pairs you pick, then that points back to alternative 1) above.

I'm sorry I'm not more help; I can't reproduce the problem you're seeing myself.

Why is the volume lower than is was previously when it was essentially the same as music playback?

The likely reason for the volume change is that some adjustments were made to the way downmixing adjusts the volume level to better account for potential clipping that would otherwise be caused by downmixing.  Those changes should result in increased volume in most cases, though, so I'm not sure what's happening there.  Maybe Hendrik can weigh in on the changes to volume.

[dvogel1]

Muting combinations of channels was revealing.

Mute L/R/C/Sub
    Stable center

Mute L/R/RL/RR
     Unstable center

Mute C/Sub
     Stable center

getting to the culprit...

Mute C
     Stable center

I have to set the volume to 100% to achieve what is normally about 1/2 volume.

[mwillems]

Muting combinations of channels was revealing.

Mute L/R/C/Sub
    Stable center

Mute L/R/RL/RR
     Unstable center

Mute C/Sub
     Stable center

getting to the culprit...

Mute C
     Stable center

Interesting!  So even with just one channel muted (Center) the center stabilizes. What happens if you only mute "sub"?  If muting sub also stabilizes the center that suggests that your center and sub channels may be reversed (i.e. the left channel info is on the right, etc.).  Otherwise, is there anything special about the element hooked to the center output or the signal chain leading to it?

I have to set the volume to 100% to achieve what is normally about 1/2 volume.

Do you have volume leveling enabled?  What does audio path show with regard to volume adjustments?  That's definitely an unusual change.

[dvogel1]

So even with just one channel muted (Center) the center stabilizes.

That is correct. The "center" is stable when the Center channel is muted.

What happens if you only mute "sub"?

With only the Sub channel muted the "center" is almost stable, that is, some of the Center channel is present on the side that it shouldn't be.

The only time that the "image" is rock solid is when the Center channel is muted.

The Center and Sub channels feed left and right midranges respectively. Please keep in mind that the filtering works correctly when playing 2 channels sources.

For reference (stereo setup):
L/R - woofers
Center/Sub - lower midranges
RL/RR - upper midranges
SL/SR - tweeters

[6233638]

Why is the volume lower than is was previously when it was essentially the same as music playback?

When you are mixing information from multiple channels, you have to reduce the volume to prevent clipping.

If you mix a stereo signal down to mono for example, you need to reduce the volume on each channel by 6dB.
If you did not do that, if both channels play anything louder than -6dB at the same time, the result would sum up to be above 0dB which would be clipped. (distorted)

If you mix a 5.1 signal down to stereo, you now have six channels of information being mixed into two, so you need to reduce the volume even further to prevent clipping.

A fully normalized downmix should use:

Left: -12.1 dB
Right: -12.1 dB
Center: -15.1 dB (3dB lower than L/R because it is being sent to both)
Sub: -8.0 dB (4dB louder than L/R because LFE is normally +10dB and summing bass frequencies among two channels is perceived as +6dB)
Left Surround: -15.1 dB
Right Surround: -15.1 dB

With this downmix, if all channels have an 0dB signal played at the same time, the total in the stereo downmix would be exactly 0 dB.
If any of the channels were louder than this, the signal would be driven above 0 dB and clip.


With 7.1 you have to reduce the volume even further:

Left: -13.5 dB
Right: -13.5 dB
Center: -16.5 dB
Sub: -9.4 dB
Left Surround: -16.5 dB
Right Surround: -16.5 dB
Left Rear: -16.5 dB
Right Rear: -16.5 dB

 
With build 124, what Media Center is supposed to do, is use the peak values from analyzed audio to avoid reducing the volume as much.

If the loudest channel is -6dB, then the values for everything in the mix could be pushed up by as much as 6dB without clipping.
If one channel is -6dB and the others are even lower than that, you would be able to push up the volume of the downmix even further than 6dB while still avoiding clipping.

However, the downmix currently used does not appear to be correct (too much LFE) and the compensation for peak levels does not seem to work correctly in this build. (clipping protection is being enabled)
 
 
 
Something I would also like to see, is the option to normalize everything relative to the volume of a 7.1 signal. (reduce stereo playback by 13.5dB, and 5.1 playback by 1.4 dB)

It would reduce the volume of all video playback, but mean that volume is consistent regardless of the source format.
Something like that could theoretically be done with zones and zone switch, but what a pain that would be.

[Hendrik]

Something I would also like to see, is the option to normalize everything relative to the volume of a 7.1 signal. (reduce stereo playback by 13.5dB, and 5.1 playback by 1.4 dB)

Volume Leveling should make this irrelevant, as it is told about the volume reduction due to mixing and should compensate to create a even volume level, even for downmixes.

Regarding the other issues, I reduced the LFE for an upcoming build to be energy neutral in a stereo downmix, and also reverted the peak level change, as it didn't account for other factors in the mixing, which resulted in driving it into clipping.