Need advice about mixing 5.1 down to stereo

[Djard]

I demuxed a movie that has too much audio dynamic range for my liking (my surround sound system does not have a "Night Mode") and would like some advice about channel assignment when mixing the channels down to stereo. Correct me if I'm mistaken, but would the following be a logical map? I'm guessing SS Ch-1 is front left, SS CH-2 is front right, etc.

SS CH.3 is the loudest track, with the highest amplitude visible in Audacity that I'm using for the mix down. I don't want the stereo to sound monoaural.

SS CH.....SPEAKER.........DESCRIPTION..............ST EREO MIX
...1.........Front L............EQ of main audio-----> Left
...2.........Front R............EQ of main audio-----> Right
...3.........Center.............Main audio------------> Split to L & R
...4.........Sub-Woofer......Rumble/Ambience----> Right
...5.........Rear L.............EQ of main audio-----> Left
...6.........Rear R.............EQ of main audio-----> Right

[imhh1]

personally i use EAC3to and i select downmix to 2ch... everything is perfect this way.

[budwzr]

There's no mapping. Just flatten everything.

[Cornucopia]

@budwzr, OF COURSE there's mapping. "Flattening everything" is mapping (to mono).

There are a number of ways you can do this, but I would tend to agree with imhh1: standard automated downmix algorithms such as those found in EAC3to, etc. are about as good as any you'll find.

@Djard, yes your mapping does follow the usual 5.1 channel layout. However, you are mistaken: Ch4 does NOT refer to "Subwoofer", it refers to LFE (low-frequency Effects). That is a very specific thing. It is "mono", so should be placed equally into both L+R. Also, its channel & the surround (back/side) channels probably should be at a slightly lower level than the main channels. Say -3dB (good for surrounds) to -6dB (good for the LFE). "Subwoofer" designates an output device that is MANAGED by the player+receiver/amp/card based on the capabilities of your physical system.

Example:
System#1 has no subwoofer, but LFE still goes to L + R, so there is no loss of effects in the program.
System#2 has subwoofer + small satellite mains, so LFE, PLUS low-pass-filtered versions of L, R, & C all go to the subwoofer, while high-pass-filtered versions of L, R & C go to their normal respective speakers.
System#3 has subwoofer + full size mains, so LFE, PLUS low-pass-filtered & reduced versions of L, R & C go to the subwoofer, while the FULL versions of L, R & C go to their normal respective speakers.

So, there's a mapping going on in 5.1 channel assignment, a (possible) mapping going on in 5.1-->2.x downmixing, and a mapping going on in realtime playback management. Goes on all the time.

BTW, the act of downmixing DOES NOT necessarily reduce the dynamic range of the program. Conversely, one can reduce the dynamic range of the program WITHOUT the need for downmixing.

...Oh, maybe that's what budwzr was referring to...

Scott

[Djard]

Yes. I discovered that you can set Audacity to automatically down mix any group of surround sound tracks to 2.0 in the Preferences, which I found refreshing, since EAC3to has no GUI. And I'm not comfortable using command lines (unless they're DOS switches!). Here's what I ended up using that sounds just fine to me, saving the file as an AC3 stereo (320 Kbps bitrate) of only 267,037 KB in size for a 113 min movie.

SS CH...SPEAKER.....STEREO MIX
...1......Front L.........Pan 100% Left
...2......Front R.........Pan 100% Right
...3......Center.........Split & pan 30% L & R
...4......LFE..............Pan 100% Right
...5......Rear L..........Pan 70% Left
...6......Rear R......... Pan 70% Right

[Cornucopia]

I'd do it differently:

1/L-> 100%L
2/R-> 100%R
3/C-> 50%L+50%R (maybe more if you prioritize dialog)
4/LFE-> 30%L+30%R
5/Ls-> 70%L+ -30%R
6/Rs-> 70%R+ -30%L
(that minus means out-of-phase)
...but, to each his own.

Scott

[amethst123]

It is good.

[hello_hello]

Different formats use different channel mappings. There is no "usual channel layout" as such, although programs often display the channels in the same order. When you convert from one format to another, the channels should be remapped as required. There's a list of common channel mappings at the bottom of this page: http://avisynth.nl/index.php/GetChannel

When you open a 5.1ch audio file in Audacity, I'm pretty sure it remaps the channels as required. They're imported/re-mapped using the "5.1ch WAV" mapping. I tested it at one stage by importing a 5.1ch AC3 file, then I imported the same audio after it'd been converted to AAC. The two formats use different channel mappings but they both imported the same way.

As far as I know, the LFE channel never contains anything which isn't already present in the left/right channels (or surround channels). It's just there for extra bass. If you're concerned about too much dynamic range, not including the LFE channel when downmixing might be a good idea. By default eac3to doesn't include the LFE channel when downmixing to stereo. You need to specify -mixlfe in the command line. I'm pretty sure the Dolby Digital specification says players shouldn't include the LFE channel when downmixing to stereo on playback. I've no idea what Audacity does if it imports 5.1ch audio as stereo. Not that it's a big deal either way.

I use a PC for playback so I keep the audio as-is and do the mixing/volume modifying stuff on playback. If you're interested, here's how I downmix and compress audio on playback. You can just as easily achieve the same result when encoding, as long as when you re-encode you can choose to decode via directshow. Here's a few samples.
1. Normal downmixing only.
2. Compressed with RockSteady.
3. Compressed with RockSteady and centre channel boosted (doesn't seem to make much difference when compressing).
4. Levelator (a program for levelling audio volume).

SS CH...SPEAKER.....STEREO MIX
...1......Front L.........Pan 100% Left
...2......Front R.........Pan 100% Right
...3......Center.........Split & pan 30% L & R
...4......LFE..............Pan 100% Right
...5......Rear L..........Pan 70% Left
...6......Rear R......... Pan 70% Right

For the record, I think the standard downmixing formula is to reduce the centre channel by 3dB, then add it equally to the left and right channels (it works out it's volume remains the same relative to everything else that way). In Audacity, if you make the centre channel mono and reduce it's volume by 3dB, the result should be the same. Anything louder than -3dB would be making it louder than it was originally. I could only guess as to what duplicating the centre channel and panning each copy by 30% would do (if that's what you're doing).
If you add the LFE channel, the same 3dB gain reduction would apply. Panning it 100% right doesn't make sense to me.
Adding left and right rear channels is often done without adjusting their volumes. It's probably personal taste there. Some people say you should leave them the same as when the rear channels are in the rear, they're closer to the listener than the front channels. Some prefer to reduce their volume a little anyway. I'm not sure why you wouldn't pan them 100% left and right though (Downmixing to Dolby ProLogic aside).

You also need to be careful of clipping, especially if you boost any channels. It might pay to reduce everything by at least a further 6dB when downmixing. That way, combining the channels shouldn't result in clipped peaks. I'm pretty sure the AC3 spec says it should be an overall 7.5dB reduction, but that's a "worst case" scenario. If the resulting stereo file is a bit low in volume you can always import it into Audacity, normalise it, then export it again.

SS CH...SPEAKER.....STEREO MIX (everything reduced by an additional 6dB to prevent clipping).
...1......Front L.........Pan 100% Left -6dB
...2......Front R.........Pan 100% Right -6dB
...3......Center.........Mono -9dB
...4......LFE..............Mono -9dB (or a bit lower, or don't include it at all)
...5......Rear L..........Pan 100% Left -6dB (or a bit lower)
...6......Rear R......... Pan 100% Right -6dB (or a bit lower)

[hello_hello]

You might also want to look at the DynamicAudioNormalizer. It works on the same principle as the RockSteady plugin I use. If encoding via DirectShow isn't convenient, it's a good alternative. Once I tweaked a few of it's settings the output was very similar to the result I achieve with RockSteady.
http://forum.doom9.org/showthread.php?t=170980

[pandy]

I'm happy with ffmpeg downmixing like this: pan=stereo|FL < FL + 1.41421FC + .5BL + .5SL + 0.25LFE|FR < FR + 1.41421FC + .5BR + .5SR + 0.25LFE

Center channel can be compressed before downmixing (you can also try suggested by hello_hello superb DynamicAudioNormalizer from lordmulder or combine them both), finally after downmixing also additional compander/DynamicAudioNormalizer stage can be used to equalize and normalize perceived loudness/dynamics of audio - perhaps this is far from high end audiophile path but it is perfect for night and mobile usage.

[Cornucopia]

Different formats MAY have used different channel assignments (particularly in the past), but as per Dolby, AES and SMPTE, the STANDARD 5.1 channel assignment is as I mentioned. Broadcasters are required to do it that way, and so expect it that way from producers. Others may be putting out different configurations, but in the eyes of professionals, it would now be considered unprofessional and frowned upon.

Also, you have some misinformation (rather: incomplete info) on the workings of LFE (when it is professionally produced).
It was intended and designed to be used for "low frequency effects" - items which SHOULD NOT appear in standard channels (emphasis on the effects). It is NOT there "for extra bass", and ought to contain mainly items which are not derived from the other channels. You've got to remember the history of surround is tied to theatrical presentation. (decent current) Speakers used in theatres have NO LACK of bass, and the LCRLsRs discreet channels are all supposed to be FULL RANGE (down to 0Hz in theory, ~20Hz in practice). However, in the early days of theatre surround, the main channels would have been overloaded had the bass effects been included in them, so a separate channel was dedicated in order to efficiently utilize its extra power requirements. This is no longer necessary with digital channels (which are able to go evenly from 0-24k+ Hz) and with intelligent bass management (where duplication of bass in both LCR & LFE could easily lead to TOO MUCH bass), and so the LFE is now focused on value-added, specialty sounds - those which may add to the visceral impact but aren't necessary to the overall program. Which is why, in Dolby's bass management downmix algorithms, it is expected to completely drop the LFE channel. Again, LFE does NOT (normally) equate to subwoofer, at least not on a 1:1 mapping.

Couple of links:
http://www.dolby.com/uploadedFiles/Assets/US/Doc/Professional/38_LFE.pdf
http://www.genelec.com/faq/multichannel/101-what-is-the-lfe-1-channel/
Wish I could share with you some more exhaustive info on it, but it's all in physical textbook form.

Btw, your "center channel reduced by 3dB" is exactly what my "50%" referred to, as 50% = -3dB.
The old Dolby ProLogic equivalent mapping for surround channels works out to 50%L + -50%R (so that, should you mix down to mono, it would cancel out and be not included: as it should) for SL and vice-versa for SR. I upped it somewhat to favor their home sides and retain a bit of the ambience effects, even if mixed to mono.

Scott

[hello_hello]

I can't find the link again (I saved the info quite a while ago) but the way I interpret the spec for AC3 (which can store downmix levels for players to use) only 3 values can be set for the centre channel volume reduction when decoding. -3dB, -4.5dB or -6dB. For the rear channels it's either -3dB or -6dB. According to the spec the specified level should be used, or if there's not one specified, then -4.5dB should be used for the centre channel and -6dB for the rear channels.

I had a look at the way of couple of decoders downmix. By default, LAV Audio decreases the centre channel volume by 3dB, the surround channels are reduced by 3dB and it doesn't include the LFE channel at all.
ffdshow mixes the centre channel at -3dB, the rear channels without reducing their volume, and it includes the LFE channel at -3dB.
According to the info here, eac3to mixes the centre channel at -3dB, the rear channels at -3dB and the LFE channel at -3dB if you elect to include it.
I've used the downmix DSP included with foobar2000 for years but I have absolutely no idea which formula it uses.

If you look at the different methods used by the above programs, -3dB for the centre channel seems to be universal, the LFE channel is optional (but -3dB if included) and the rear channels are added with a 0dB or a -3dB reduction, depending on the program. Anything else from there would be personal choice, I guess, aside from mixing the LFE channel into the right stereo channel only. That's just wrong.

[Cornucopia]

Center to -3 is for constant power, -6 is for constant gain (particularly when mixed to mono). -4.5 is a mid-way compromise between the 2. There are a number of options with these "pan laws", as they are called. These are all a matter of taste, but -3 is probably the most commonly used (and is historically what Dolby used in their original 4-ch. Surround method).

Since low frequencies are non-directional, it may not matter acoustically where you place the material, but from a mixing/datamanagement/overhead viewpoint, biasing a single channel with the "power" of the bass samples will skew the L/R balance and my result in distortion in the "weighted" channel. So I agree, it's just wrong.

Scott

[pandy]

Btw, your "center channel reduced by 3dB" is exactly what my "50%" referred to, as 50% = -3dB.

Nope 50% i.e. half (1/2) is equal to 6.0206 dB, 3dB is 1/SQRT(2) i.e. 0.707 (70.7%)

http://en.wikipedia.org/wiki/Decibel

Downmixing equations use audio level not audio power.

Side to this downmixing LFE is not recommended by Dolby but Dolby recommendations are not mandatory and also this is highly subjective area - invoking LFE in downmixing can be partially justified by reduced capabilities of portable devices and reduced perceived loudness in low range. Most of Dolby recommendations are related to RF mode which i found kind of odd nowadays - connecting HD equipment by RF modulator is not a good choice.

[hello_hello]

Different formats MAY have used different channel assignments (particularly in the past), but as per Dolby, AES and SMPTE, the STANDARD 5.1 channel assignment is as I mentioned. Broadcasters are required to do it that way, and so expect it that way from producers. Others may be putting out different configurations, but in the eyes of professionals, it would now be considered unprofessional and frowned upon.

Different formats use different channel assignments. AC3 and DTS use different channel assignments. Which one of them is unprofessional and frowned upon? AAC uses different channel assignments to both wave and the standard 5.1ch assignment you mentioned. Or was the AAC spec for multichannel assignments changed at some stage to be different to what it was in the past? When did that happen?

Also, you have some misinformation (rather: incomplete info) on the workings of LFE (when it is professionally produced).
It was intended and designed to be used for "low frequency effects" - items which SHOULD NOT appear in standard channels (emphasis on the effects). It is NOT there "for extra bass", and ought to contain mainly items which are not derived from the other channels. You've got to remember the history of surround is tied to theatrical presentation. (decent current) Speakers used in theatres have NO LACK of bass, and the LCRLsRs discreet channels are all supposed to be FULL RANGE (down to 0Hz in theory, ~20Hz in practice). However, in the early days of theatre surround, the main channels would have been overloaded had the bass effects been included in them, so a separate channel was dedicated in order to efficiently utilize its extra power requirements. This is no longer necessary with digital channels (which are able to go evenly from 0-24k+ Hz) and with intelligent bass management (where duplication of bass in both LCR & LFE could easily lead to TOO MUCH bass), and so the LFE is now focused on value-added, specialty sounds - those which may add to the visceral impact but aren't necessary to the overall program. Which is why, in Dolby's bass management downmix algorithms, it is expected to completely drop the LFE channel. Again, LFE does NOT (normally) equate to subwoofer, at least not on a 1:1 mapping.
Couple of links:
http://www.dolby.com/uploadedFiles/Assets/US/Doc/Professional/38_LFE.pdf
http://www.genelec.com/faq/multichannel/101-what-is-the-lfe-1-channel/
Wish I could share with you some more exhaustive info on it, but it's all in physical textbook form.

I wish you could too, because there's nothing I'm reading in either of those links which explicitly states the LFE channel contains info not found in the other channels.

I'm reading this:
"Its purpose is to supplement the overall bass content of the program or to ease the burden on the other channels."
and this:
"The .1 channel was originally designed for use in theatres where the main channels could not reproduce the lower frequencies and additional headroom was required at these low frequencies to reproduce high SPL."
Which requires some serious extrapolation, to give you this:
"It was intended and designed to be used for "low frequency effects" - items which SHOULD NOT appear in standard channels (emphasis on the effects). It is NOT there "for extra bass", and ought to contain mainly items which are not derived from the other channels."

Either way though, the principle is probably still the same. If you're worried about too much dynamic range, don't include the LFE channel when downmixing.

Btw, your "center channel reduced by 3dB" is exactly what my "50%" referred to, as 50% = -3dB.
The old Dolby ProLogic equivalent mapping for surround channels works out to 50%L + -50%R (so that, should you mix down to mono, it would cancel out and be not included: as it should)

I never said it wasn't. (Edit: But yes, Pandy's post refreshed my memory. -3dB = about 70%)
Dolby Prologic? I've never decoded a Dolby Prologic or Dolby Prologic 2 stereo file to surround sound in my life so downmixing using a Dolby Prologic formula isn't something I've ever worried about. Plus on ocassion it seems to mess with relative volumes when the audio is decoded as stereo. I noticed that a couple of times when AutoGK downmixed to stereo, as it uses a Dolby Prologic formula. Loud sounds such as gunshots were much quieter (not as a general rule, just something I noticed once or twice). Then again, maybe that's because it was then re-encoded with a lossy encoder such as LAME which is designed to throw away info you can't hear and probably has no concept of Dolby anything.

[hello_hello]

Also, you have some misinformation (rather: incomplete info) on the workings of LFE (when it is professionally produced).
It was intended and designed to be used for "low frequency effects" - items which SHOULD NOT appear in standard channels (emphasis on the effects). It is NOT there "for extra bass", and ought to contain mainly items which are not derived from the other channels.

Here's the LFE channel taken from "professionally produced" AC3 audio (from a TV series). The sample is just the first few minutes.

Here's the left and right channels after applying a low pass filter at 80Hz. So I could hear just the low frequencies. Same few minutes in time.

What's in the LFE channel that's not in the stereo channels (aside maybe some additional higher frequencies due to the low pass frequency I chose)?

After applying a low pass filter to the stereo channels, they even look the same as the LFE channel.
Towards the end of the sample you'll notice the LFE channel contains no audio, while the left and right channels do contain low frequencies, but I've not found a spot where the LFE channel contains low frequencies which aren't in the left and right channels too.

[hello_hello]

Side to this downmixing LFE is not recommended by Dolby but Dolby recommendations are not mandatory and also this is highly subjective area

Because I compress the audio on playback (well.... "effectively" compress it) I stopped including the LFE channel when downmixing on playback. I noticed sometimes the volume of dialogue would drop a bit, always in places with a lot of "bass". The music or effects behind the commentary in documentaries could do it on occasion. Since I stopped including the LFE channel, I've not noticed it happen. Maybe just co-incidence.....

[pandy]

Side to this downmixing LFE is not recommended by Dolby but Dolby recommendations are not mandatory and also this is highly subjective area

Because I compress the audio on playback (well.... "effectively" compress it) I stopped including the LFE channel when downmixing on playback. I noticed sometimes the volume of dialogue would drop a bit, always in places with a lot of "bass". The music or effects behind the commentary in documentaries could do it on occasion. Since I stopped including the LFE channel, I've not noticed it happen. Maybe just co-incidence.....

This is why i recommend to compress before downmixing channels not on downmixed channels - particularly Center can (should) be compressed before (to provide clear dialogue line), LFE also can be compressed before mixing it to R and L to limit levels and to avoid overloading.
We talking about difficult listening scenarios (nigh/mobile devices in presence of the noisy environment etc) - this is special case far from neutral, hifi reproduction scenario and that's why this area is so subjective.

[Cornucopia]

@pandy, by that same link, 50% amplitude is -6dB, 50% POWER is -3dB. I have been referring to power (which is why I used the term). And you're wrong, downmixing equations, as hello_hello has just demonstrated, use a number of different algorithms, some based on amplitude/gain, some on power.

...

I had a nice set of links to SMTPE, EBU, ITU, Dolby, DCI, MPEG2, FLAC, WAV and DTS reference documents (and their dates) showing ALL of those using the channel assignments I mentioned. Lost the paragraphs in a stupid firefox refresh!!! But you could find them yourself, too.
The lone holdouts in this standardization are QT/MOV & it's stepchild MP4/AAC (and possibly WMA/ASF). Doesn't surprise me, knowing their heredity. However, those shouldn't matter anyway, as their structure allows for Named Channels, so there should never be any confusion about assignment when using those formats.

@hello_hello, you're asking me to evaluate a short segment from some show that you found and did something to... for something so amorphous as that I can only say, WHO KNOWS? Maybe the producer(s) of the show only used LFE "properly" in one section (a section not in your sample segment). Maybe they think like you and don't use it properly. Maybe there is no need in this program to make use of LFE, but some tech down the chain didn't want it to be "empty" and so filled it in. I'm not saying it doesn't happen that the recs are ignored or even interpreted differently (even in professional circles), but there are recs and if you want to be professional, you follow them. YOU don't have to be professional if you don't want to. I, however, will continue to strive to be professional.

BTW, one of those recs did explicitly state "LFE to be used only for special effects not put into the standard channels". Not sure now which one (EBU or ITU, I think)...

Scott

<edit>I do think that at some point in the 2000's many of those joined the bandwagon. I know for a fact that prior to DCI, the film ch assignment spec was different, but DCI brought them in lock step with the rest.</edit>

[pandy]

@pandy, by that same link, 50% amplitude is -6dB, 50% POWER is -3dB. I have been referring to power (which is why I used the term). And you're wrong, downmixing equations, as hello_hello has just demonstrated, use a number of different algorithms, some based on amplitude/gain, some on power.

Nope - Dolby always referring to 0dBFS and level not power. - there is no power in digital domain and usually there is no power in audio world.

http://en.wikipedia.org/wiki/Sound_power

BTW, one of those recs did explicitly state "LFE to be used only for special effects not put into the standard channels". Not sure now which one (EBU or ITU, I think)...

This for studio and broadcast usage not personal audio - difference is huge as broadcast need to provide universal method where personal audio don't need to be universal - it can even have frequency equalization to compensate some limitations (for example partial hearing loss).

[Cornucopia]

Nope. There's ALWAYS power in the audio world - it has to get to your ears! That is going to affect how you mix. You don't have new insight about this duality, it has been around for decades.

************

And personal audio (when it's done right) is based on professional/broadcast audio.

Scott

[hello_hello]

Actually, the way I arrived at the figures for downmixing was by looking at the percentage each program uses. That's how they display it. As a percentage. They all use 70.7% for the centre channel. I have no idea how I demonstrated they all use different algorithms, some based on amplitude, some based on power, when they all use the same percentage for the centre channel gain reduction. The only differences were whether the LFE channel was included, and whether the volume of the rear channels was reduced.

Cornucopia,
I see you've decided to ignore the samples I posted as much as possible. Apparently the first "professionally produced" 5.1ch AC3 file I looked at doesn't fit your theory on the contents of LFE channel so by some major co-incidence it's also not normal? All I can say is it fits my experience of what's normal for the LFE channel, although I'd be happy to look at an example which fits your theory. Which doesn't even really make sense to me. I record an explosion, a gunshot, a fart..... I duplicate the low frequencies in the LFE channel for added effect. What then? I remove them from the left and right channels and leave those channels without full range audio? Could you explain the circumstance under which the LFE channel might end up with low frequency effects which aren't in the left and right channels?

I'd look for those documents you referred to but I'd be worried the Firefox might eat my homework too. Did the information you found on channel assignments change after the page refreshed?

The lone holdouts in this standardization are QT/MOV & it's stepchild MP4/AAC (and possibly WMA/ASF).

Oh..... So when you said different formats all use the same channel mappings, you meant all the different formats don't use the same channel mappings?
As per the link I offered originally, WMA does use your beloved "standard" 5.1ch mapping. The same as for WAV. According to the info there, AC3 does not. I wonder how out of date the AFTEN encoder must be?

http://aften.sourceforge.net/longhelp.html
[-chmap #] Channel mapping order of input audio
Some programs create WAVE files which use a channel mapping other than the standard WAVE mapping. This option allows the user to specify if the input file uses WAVE, AC-3, or MPEG channel mapping. The MPEG channel mapping is used by DTS and by MPEG-2/4 formats such as MP2 and AAC.
0 = WAVE mapping (default)
1 = AC-3 mapping
2 = MPEG mapping

However, those shouldn't matter anyway, as their structure allows for Named Channels, so there should never be any confusion about assignment when using those formats.

I wish I'd said something like that. Oh, wait a minute......

I'm not saying it doesn't happen that the recs are ignored or even interpreted differently (even in professional circles), but there are recs and if you want to be professional, you follow them. YOU don't have to be professional if you don't want to. I, however, will continue to strive to be professional.

That's just meaningless rhetoric.

[hello_hello]

Looks like Wikipedia have it wrong again........

Channel identification

Home theater systems

[Cornucopia]

Your info is flawed and possibly out of date.

Yes, I'm going to ignore your incomplete, doctored and anecdotal example from some unnamed source. Unless your experience can be backed up as professional, it IS skewed.

My (recovered) channel assignment links:

DTS in MPEG2(indicative of MPEG2)
DTS in ISO (indicative of ISO MP4)
DCI
AES
EBU
SMPTE
ITU
FLAC
WAV
Dolby

Was able to regurgitate my "eaten homework" after all (not just a figment of my imagination as you would have hoped).

Aften is a reverse-engineered product of a reference encoder/decoder. This means they are doing their best to follow what they think the reference is expecting. Most times they get it right, but not completely. This seems to be one of those times. This is to be expected when they aren't allowed the insider info.

Scott

[Djard]

Being the least experienced, I defer to the expertise kindly shared in this thread. I like the idea of amplifying the weak-volume parts rather than compressing, but I'm kind of fatigued by the need to learn how to use yet another app (I'm already using at least 15 to edit AV). My rationale for directing the LFE to the right channel only was to achieve more separation of sound between the two speakers, countering any mono effect. I now understand why that's "wrong."

Thanks for the idea to place the L & R spkrs out-of-phase. In the past, when working with a mono-sounding audio, I added some high frequencies to ch-1 and reduced the 100Hz band, and vice-versa for ch-2. It helped as the difference kicked in only when those frequencies were elevated. Someone once suggested advancing one channel by 30 ms for additional separation; but--maybe just psychologically--I sense asynchrony when doing that as well as noticing a hint of a choral effect in some parts.

So do I understand correctly that subwoofers play bass from the surround sound channels, through the hardware's bass management system, and that the LFE channel contains only low frequency signals from the other channels, merely supplementing the bass, the info in the LFE channel included in the other channels but not vice-versa?

One more question if I may, and I beg your patience if asking for the complex subject to be ridiculously simplified; but would the following serve as a rough guide to use in an app like Audacity when mixing SS, the ear being the final judge? The "Amplify" column represents perceptible volume. No doubt the percentage equivalents are incorrect.

SS CH....SPEAKER............AMPLIFY
...1..........Front L...........0 Db (100%)
...2..........Front R...........0 Db (100%)
...3..........Center F........-3 dB (50%)
...4..........LFE...............-6 dB (25%)
...5..........Rear L...........-3 dB (50%)
...6..........Rear R...........-3 dB (50%)

[Cornucopia]

I think that's a real good starting point. You may want to adjust to your taste after hearing it, but what you've got listed makes sense and should be balanced.

You are correct: out-of-phase material that is achieved through simple polarity inversion causes less complications than when it's achieved through time or phase delay. Similarly, pseudo-stereo generated by complementary unbalanced EQs are both easy to spot as fake and also generate their own problematic psychoacoustic artifacts. This sort of thing can work well only in a multi-track setting where individual instruments get such tailored treatment.

The LFE channel can contain "specialized" duplications of some of the bass elements already present in the main channels (as hello_hello was suggesting), and it can also contain items unique to its own channel (as I was suggesting). And you are correct: the mains channels would not contain those "specialized" duplicates in their own channel. Doing so could easily cause the bass management system to combine elements TWICE to the subwoofer. (bad)

As far as the dynamic range compression, what model of surround receiver do you have? Maybe you have an equivalent to night mode and just don't know it (as it seems to be a stock feature on all the surround receivers I'm acquainted with).

Scott

[hello_hello]

Your info is flawed and possibly out of date.

Yes, I'm going to ignore your incomplete, doctored and anecdotal example from some unnamed source. Unless your experience can be backed up as professional, it IS skewed.

Given you accused me of doctoring the audio, I'm going to tell you to stick your links..... I won't bother looking at them.

Even assuming what you say is true, it doesn't negate anything I said. If things have changed and most formats use the same channel layout these days, obviously that wasn't always the case. The different formats still need to be remapped when audio is converted. The DVD I bought ten years ago still has the same AC3 audio with the same channel layouts. By the looks of it, the AFTEN encoder still uses the "AC3" channel layout listed on the page I linked to, so if you convert anything to AC3 using it, you're probably use the "old" AC3 channel layout, assuming what you've said is true and the channel layout has changed.

[hello_hello]

Well I confess I weakened and checked out your Dolby pdf. You'll need to explain what I'm missing here. The channel layout looks like the AC3 layout specified on the page I linked to.



Looks like Dolby use the same formula as the rest of us. 70.7% = -3dB.



"Therefore, low-frequency content essential to the program should never be mixed exclusively to the LFE channel".



Seems a bit different to "It is NOT there "for extra bass", and ought to contain mainly items which are not derived from the other channels"
I'll concede it certainly implies the LFE channel can contain low frequencies not found in the other channels. I'm still not seeing a likelihood of that happening much though.

[hello_hello]

Djard,
From the Dolby pdf:

Channel Levels
The following text is in accordance with the ITU-R recommendation and SMPTE standard referred to in the above section.
For consumer and DVD production studios, relative channel levels assume each speaker delivers identical acoustic sound pressure levels to the listener. This excludes the LFE channel, which is intended for reproduction at +10 dB SPL (with respect to the main channels within the same 3 Hz to 120 Hz passband). Assuming that a Surround (S) signal is delivered to a single speaker and two Surround signals (LS, RS) are each delivered to individual speakers, Surround levels should be identical to those for the front channels.

The way I read that, it seems the LFE channel is intended to be amplified far more then the rest. That'd explain why, in the sample I posted above, the bass frequencies in the LFE channel were lower in volume than the same frequencies in the left and right channels. I assume then, if you decide to include the LFE channel in the downmix, there'd be no need to reduce it's volume by too much. -3dB or -6dB. Any more than that, and maybe there's not much point including it at all.

SS CH....SPEAKER............AMPLIFY
...1..........Front L...........0 Db (100%)
...2..........Front R...........0 Db (100%)
...3..........Center F........-3 dB (50%)
...4..........LFE...............-6 dB (25%)
...5..........Rear L...........-3 dB (50%)
...6..........Rear R...........-3 dB (50%)

So what happens if you apply those levels and combining the channels as stereo results in clipped peaks? It can happen. If a left/right channel is at maximum volume and something is added to it at the point where it's already at maximum.... it'll need to get a bit louder, but it can't.
That's why eac3to use 2 passes when re-encoding if necessary. It applies a standard dowmixing formula, but if clipping is detected it downmixes again, applying an appropriate gain reduction to prevent it. Other programs (most of them, probably) apply a "worse case scenario" level reduction, then the volume of the downmixed audio is adjusted until the peaks are at maximum. ffdshow and Avisynth use the same worse case scenario formula. ffdshow uses it if the "normalise matrix" option is checked in it's mixer filter. Left and right channel volumes (front and rear) are all reduced to 29.3% (I think it's 29 point three, it's hard to read). Centre and LFE channels are reduced to 20.7%. I can't remember what that works out to as a reduction in dB. Probably something like 7dB and 10dB. I'll have to work it out later when I can remember the formula for converting percentage to dB.

[hello_hello]

I have a vague recollection of linking to a page earlier which stated the channel order for DTS is C, L, R, SL, SR.
I should go back and re-read my previous posts to check.

Implementation of DTS Audio in MPEG-2
Page 22:



The beginning of the "DTS Audio in MPEG-2" pdf contains the following:

"Do Not Duplicate. Copyright © 2013 DTS, Inc. All Rights Reserved. Unauthorized duplication is a violation of State, Federal, and International laws.
This publication is copyrighted and all rights are reserved by DTS, Inc. Without the express prior written permission of DTS, no part of this publication may be reproduced, photocopied, stored on a retrieval system, translated, or transmitted in any form or by any means, electronic or otherwise."

I guess I should leave it up to a mod to decide if I've violated the world police copyright laws by copying the above copyright warning and pasting it into my post. If I have, I guess a mod can remove it.
Oh..... and the screenshot I posted, taken from the same pdf..... that might upset the world copyright police too...... although I'll vote for "fair use" on that one.