Best method to downmix surround to stereo? Are LoRo fold downs outdated?

[nickleb474]

I used to be content with downmixing 5.1 surround sound music with the LoRo technique, reducing the gain of the center and rear channels by 3db. Eventually I started to wonder, why do we reduce gain of the rear channels, at all? After trying to find any documentation for a real reason why the rears are not mixed at +0db, I found two main arguments online:

1. To prevent the content in the rears from overpowering or de-emphasizing the front, where the most important sounds are typically located (particularly in films)

2. To prevent phase issues when a sound is panned around the surround field, as when that sound is in-between a pair of front and rear channels, they sum together and incorrectly increase the perceived volume of that sound by 3db

Learning this, I started to try out downmixing with DPLII, but I hate the out of phase sound of the rear channels, so that was a no go for me, personally. So I felt a little defeated and went back to LoRo. But I couldn't shake the fact that these downmixes just sound bad, and worse, they sound wrong. When the most important content happens to be in the rears (examples that come to mind: Lady Gaga singing in the rears on Sweet Sounds of Heaven Atmos, and The Beatles harmonies from Because in the Now And Then Atmos), the balance of the mix is ruined.

Therefore, I'd love to ask this wonderful community, what you think is the best way to downmix surround sound to stereo? Some other techniques I know of are DPLII and simulated surround sound/binaural downmixing. I also developed my own method which I'm quite pleased with, it's a LoRo downmix with no attenuation of the rears, however before that, I reduce the gain of the phantom center between front and rear pairs by 3db, to prevent #2 above (phase issues). I also throw in the LFE at +7db as stated in DD+ LoRo documentation. If you think that my idea sucks please call me out! Just want to find the absolute best way to enjoy Atmos Mixes on the go

Thank you for reading and have a nice day!
Nickle-B

 

[jimfisheye]

I think it's as simple as all the things you pointed out are what you lose going from surround down to stereo or binaural stereo. And some mixes suffer more than others.

Folding a center channel into the front L/R pair as a phantom center is a more matter of fact example of the 3db attenuation needed. You would double up the signal if you put it to two channels instead of one with each channel the same level as the single channel was. The two channels getting -3db adds up to the original level in a single channel.

It gets subjective with the rear (and beyond with 12 channel mixes now) just like you spelled out. Folding down a surround mix is a compromise to begin with. You already know you are missing the main event and listening from bad seats. A compromise like altering levels might work to put the balance closer to the original intent. But some examples will go the other direction.

There's an opinion that surround is more of an extension of stereo and that there is always a proper way to fold it down and it's talked about as though this is some standard thing and not a compromise at all. Dolby is going hard with that kind of brochure speak for their stereo listeners to try to convince them they're hearing surround sound in stereo now with Atmos, for example. Movie mixes are idiosyncratic and mixed with that intention. Most of them are anyway. Maybe some music is too but I want to say music mixes in surround are more unique animals and sometimes don't translate to different speaker configurations as intended.

 

[kfbkfb]

Could try the CD-4 quad disc downmix method (I don't recall any controversy about this downmix method during the 1970s quad era):
L=LF+LB & R=RF+RB & C=CF & C=LFE (trial and error volume level), no additional out of phase content is created by this stereo downmix method but out of phase content between LF&LB & RF&RB is reduced in volume or lost.


Kirk Bayne

 

[nickleb474]

L=LF+LB & R=RF+RB & C=CF & C=LFE (trial and error volume level)

Thank you for your response! Though I'm curious, how does this differ from a LoRo downmix with no attenuation? Maybe I am confusing what F and B stand for. Thanks in advance!

 

[kfbkfb]

I don't know much about downmix methods, mainly the method used in CD-4 and the DD 5.1 to Dolby Surround, F=front speakers in the 1970s quadraphonic speaker square configuration and R=rear speakers.

Due to my marginal CD-4 decoding setup, I nearly always listen to the CD-4 to stereo downmix, I haven't noticed anything really weird in the stereo downmix of my CD-4 discs from a variety of record labels.


Kirk Bayne

 

[nickleb474]

I don't know much about downmix methods, mainly the method used in CD-4 and the DD 5.1 to Dolby Surround, F=front speakers in the 1970s quadraphonic speaker square configuration and R=rear speakers.

Due to my marginal CD-4 decoding setup, I nearly always listen to the CD-4 to stereo downmix, I haven't noticed anything really weird in the stereo downmix of my CD-4 discs from a variety of record labels.


Kirk Bayne

Ah, so you're saying you have a CD-4 decoder, but you typically only listen to the output of that decoder after it's been downmixed by, say, your receiver? If that is the case, then it's probably a LoRo downmix with -3db to the rears, not a straight front + rear downmix. PROBABLY

IGNORE ME. Difference signals be cool as hell

 

[kfbkfb]

https://en.wikipedia.org/wiki/Compatible_Discrete_4

I just listen to my CD-4 discs with my stereo record playing equip, it's the same concept as listening to a stereo FM broadcast on a mono FM radio - L & R are added together at equal volume.


Kirk Bayne

 

[nickleb474]

Yeahhh... My bad!!! Edited my post

 

[par4ken]

I have never heard the term LoRo before. I assume that it is just a front to back stereo fold down?

 

[nickleb474]

I have never heard the term LoRo before. I assume that it is just a front to back stereo fold down?

That's right! It stands for Left Only, Right Only. The rear channels are (99% of the time) reduced by 3db, this is the typical downmix that modern receivers do automatically.

 

[chucky3042]

I remind all that we still have stock of the Intelligent Involve encoder. It's a triband QS compatible encoder with 2 modes one mode uses a variable matrix " constant" encode. Suzanne Ciani' quad demo vinyl used it.

They cost usd$100 to play with it. We claim the encoded stereo sounds just like stereo with no image compression and near perfect decode. I can publish the test results of asked.

 

[barfle]

I would say that any down (or up) mix really can’t be done automatically if you expect certain results. That’s why mixing engineers have jobs.

If you want it done a particular way, or to result in a particular sound, you have to do the work yourself.

 

[par4ken]

Form my point of view the LtRt method makes the most sense. I think that encodes sound better, more richer, fuller than just plain stereo.

I remind all that we still have stock of the Intelligent Involve encoder. It's a triband QS compatible encoder with 2 modes one mode uses a variable matrix " constant" encode. Suzanne Ciani' quad demo vinyl used it.

They cost usd$100 to play with it. We claim the encoded stereo sounds just like stereo with no image compression and near perfect decode. I can publish the test results of asked.

I'm still debating whether or not to get an Involve encoder. It might be fun to play with. The thing is that I have little need for stereo mixdowns anymore. If it is discrete I play it discrete. I never listen to two speaker stereo if I can help it.

 

[nickleb474]

Form my point of view the LtRt method makes the most sense. I think that encodes sound better, more richer, fuller than just plain stereo.

I personally can't stand the sound of the rears being 180 degrees out of phase. I agree it sounds "fuller" as a result of the rears being downmixed louder than with LoRo, but I hatehatehate the fact that content the the center of the rears is completely messed up by being OOP. Sounds awful IMO. Curious if the Involve encoder does this, seeing how it is a matrix downmixer, I'm assuming it does.

 

[par4ken]

I personally can't stand the sound of the rears being 180 degrees out of phase. I agree it sounds "fuller" as a result of the rears being downmixed louder than with LoRo, but I hatehatehate the fact that content the the center of the rears is completely messed up by being OOP. Sounds awful IMO. Curious if the Involve encoder does this, seeing how it is a matrix downmixer, I'm assuming it does.

I love the effect you hate it, fair enough. I would suggest a mixdown where the rear channels (or maybe the front channels) are panned toward the centre a bit. That will add no phasyiness and will prevent the hole in the middle effect of of LoRo. It will also create a mix that can be decoded if desired. You could lower the level of the inward panned rear channels a bit if that sounds better to you.

 

[jimfisheye]

First seeing "LoRo" as well.
I wouldn't over think it too much. The main event is still the unmolested unaltered surround mixes for the speaker array they were created listening with.

After that, we're still hearing the mix in a pinch. Listening to a smaller surround array or stereo earbuds or a car sound system isn't intended to be reference listening. The overall presence in a mix takes over as the most prominent element. As long as we aren't doing a fold down and throwing out the original files! There's no in between for keeping an original surround mix vs making a fold down with that. ie. No way to get the full unaltered surround mix back out of a fold down.

Atmos, for all the comments that can be made about the "brochure speak" and using it like a hardware dongle is at least trying to make a 'single inventory' file that DOES preserve the original mix while letting the stereo listeners hear what's left after folding it down. (If you don't want to come to terms with that, just keep saying "But the screen says Atmos... I'm listening to Atmos!")

Front heights fold into front L/R. Rear heights fold into the rears. Except in a x.x.2 system the heights fold into the L/R center heights. Mixes are being altered quite a bit with just those permutations.

Sometimes quad to stereo can be done with the rear channels hard L/R and the front channels towed in. Then adjust the levels to taste so it doesn't flip the front to back weight of the original.

Back to the phantom image across multiple speakers vs single channel thing again. Folding the C channel into front L/R (yes, with proper 1:1 levels with -3db x2=0db) changes things. When you're behind the mixing board with the ability to follow your ear with any adjustment, the choice of phantom mono across speakers vs single speaker makes a difference. Consider some of the 5.1 mixes that aren't well revered because the phantom center vocal people were used to sounds small and pulled out of the mix put to a single channel instead.