Upmix vs synthesize decode

QuadraphonicQuad

Help Support QuadraphonicQuad:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

gvl_guy

1K Club - QQ Shooting Star
QQ Supporter
Joined
Apr 23, 2019
Messages
1,049
Location
Greenville, SC (via NJ, Philly, ATL & SoFL)
I'm still a bit of a newbie when it comes to the more modern surround sound mixes and equipment. Although I've been collecting quad (4-channel) stuff on and off since the 70s, it's been only recently that I bought my first Oppo for modern disc reproduction.

I have a question about "upmixing."

I get the impression that upmixing something is taking the stereo signal and turning it into a fake 4 or 5.1 channel selection. For this, you'd use some kind of encoder? Or synthesizer? It turns it into 4 or 5 discrete channels for playback later.

Wouldn't this be the same as taking a stereo selection and running it through a decoder of some kind to synthesize quad sound? I mean, I guess having the upmix means you don't need any kind of decoder, but wouldn't playing the stereo selection through a QS, Tate or Involve decoder do about the same thing?
 
Different upmixers work differently but in the case of SpecWeb there is no similarity to matrix decoders, etc. whatsoever.

The "Spec" in SpecWeb refers to "Spectral". The upmix algorithm works in the spectral domain vs. the time domain like matrix decoders.

Defiantly also a digital vs. analog technique.

Each "chunk" of music is divided up into spectral bins and the magnitude of those bins, for the left and right channels, is compared to determine where in the stereo field that frequency was panned, at that instant.

That information is used to, yes, you could use the word synthesize, the music re-panned into 5 channels (the 5 in 5.1). There are lots of controls about exactly how that happens, so you can customize it for each song.

The result is not "fake". The result is how the original music was mixed, but spread around 5 channels instead of 2.

and no, it doesn't sound like any push button on any home theatre or Quad receiver.

Also note in the 70's, 80's etc. there wouldn't have been enough (or any in home audio electronics) compute power to do spectral processing in real time, so they used analog phase techniques to try to recover "hidden" channel information in the stereo (or not hidden but encoded, in the case of quad).

Yes, once upmixed you have 6 channels of sound. No need for a decoder.

FYI for going beyond 5.1 or 7.1 I am also using AI based music source separation, in combination with upmixing, to go from stereo all the way to 7.1.4 as in Dolby Atmos, DTS:X, or Auro 3D.

The AI based source separation gives you separate stereo versions ("stems") for vocal, bass, drums, and "everything else". Those stereo stems, and upmixes of those stereo stems, can be placed differently in the 3D 7.1.4 space.

For instance,

vocals upmixed to surround but maybe the rears placed in the rear height speakers (think harmonies come from behind and above you).

Bass comes from the sides (and the subwoofer).

Drums are upmixed into quad and placed in the 4 height speakers.

"Everything else" is upmixed into 5.1 and mapped to 5 of the 7 speakers at ear level (the 2 "side surrounds" having been used for bass).

That's just one possible configuration but it has proven very effective in listening tests with QQ members that have "immersive" systems.

So that is light years beyond QS, Tate, etc. The AI powered music source separation having only become practical in the last 2 years.
 
Different upmixers work differently but in the case of SpecWeb there is no similarity to matrix decoders, etc. whatsoever.

The "Spec" in SpecWeb refers to "Spectral". The upmix algorithm works in the spectral domain vs. the time domain like matrix decoders.

Defiantly also a digital vs. analog technique.

Each "chunk" of music is divided up into spectral bins and the magnitude of those bins, for the left and right channels, is compared to determine where in the stereo field that frequency was panned, at that instant.

That information is used to, yes, you could use the word synthesize, the music re-panned into 5 channels (the 5 in 5.1). There are lots of controls about exactly how that happens, so you can customize it for each song.

The result is not "fake". The result is how the original music was mixed, but spread around 5 channels instead of 2.

and no, it doesn't sound like any push button on any home theatre or Quad receiver.

Also note in the 70's, 80's etc. there wouldn't have been enough (or any in home audio electronics) compute power to do spectral processing in real time, so they used analog phase techniques to try to recover "hidden" channel information in the stereo (or not hidden but encoded, in the case of quad).

Yes, once upmixed you have 6 channels of sound. No need for a decoder.

FYI for going beyond 5.1 or 7.1 I am also using AI based music source separation, in combination with upmixing, to go from stereo all the way to 7.1.4 as in Dolby Atmos, DTS:X, or Auro 3D.

The AI based source separation gives you separate stereo versions ("stems") for vocal, bass, drums, and "everything else". Those stereo stems, and upmixes of those stereo stems, can be placed differently in the 3D 7.1.4 space.

For instance,

vocals upmixed to surround but maybe the rears placed in the rear height speakers (think harmonies come from behind and above you).

Bass comes from the sides (and the subwoofer).

Drums are upmixed into quad and placed in the 4 height speakers.

"Everything else" is upmixed into 5.1 and mapped to 5 of the 7 speakers at ear level (the 2 "side surrounds" having been used for bass).

That's just one possible configuration but it has proven very effective in listening tests with QQ members that have "immersive" systems.

So that is light years beyond QS, Tate, etc. The AI powered music source separation having only become practical in the last 2 years.
WOW. Thank you so much for that explainer. I had no idea (and now I do.) ☺️
 
Back
Top