So it seems we are talking about matrix encoded stereo, then no SpecWeb wouldn't separate any of that. What happens is we move the audio into the spectral domain (magnitude in frequency bins in time vs amplitude in time) using a "phase vocoder", then the magnitudes, in each frequency bin, are compared L vs R to determine what channels to output the magnitudes in. The Frequency values (which could be converted to phase values but aren't in this case) are just passed on to the five inverse phase vocoder processes.
I've messed around some with "quad" decoding but others have put so much into perfecting quad decoders that I didn't take it through to production.
I guess another way to say all this is IMO 98% plus of stereo is produced with what is called ILD panning, tricking our brains to locating sounds based on different levels (but the same phase) coming from the left and right speakers, as apposed to ITD, or tricking our brains into locating sounds based on the time or phase difference between signals coming from the left and right speakers.
Mixing consoles/software have pan pots that use ILD, in constant power (sin cosine), constant amplitude, or linear panning rules.
So Spec/SpecWeb exploits ILD to upmix stereo to surround.
I'm told there are some producers / mixing boards that do ITD, but I can't honestly say I've ever identified any music actually mixed that way, but still I think that is different from matrix encoding stereo.
This is all in reference to sounds in the horizontal plane, by the way. It gets way more complicated when talking about how we perceive sound direction vertically.
https://en.wikipedia.org/wiki/Sound_localization
Hello Z!
Thank you for the comprehensive reply. Starting with an explanation: I have tried multiple times to use Plouge Bidule. The methodology & terminology just leaves my mind comatose. I’m not saying it’s wrong as others smarter than me on this forum have had good results from it. It’s just a way of doing things I cannot wrap my head around. So when SpecWeb came out I was very interested. A chance to use the legacy of Plouge Bidule in a way I can understand. Drag & click enter! Your post has illuminated aspects of SpecWeb that I either did not know or wrongly assumed.
I guess another way to say all this is IMO 98% plus of stereo is produced with what is called ILD panning, tricking our brains to locating sounds based on different levels (but the same phase) coming from the left and right speakers, as apposed to ITD, or tricking our brains into locating sounds based on the time or phase difference between signals coming from the left and right speakers.
Mixing consoles/software have pan pots that use ILD, in constant power (sin cosine), constant amplitude, or linear panning rules.
So Spec/SpecWeb exploits ILD to upmix stereo to surround.
I agree that a large part of constructing a stereo multi-track soundstage is based on amplitude panning. But ever since Les Paul discovered tape slap back echo (maybe earlier) artists have been using much more than just panning to make a soundstage. Echo whether plate, spring or digital, chorus, flanging, vocoder all produce an intricate array of not just level but phase relationships as well. I am probably stating the obvious. But more specifically any 2 ch stereo recording can have a level relationship between one and zero, any phase relationship between matching and opposite phase.
Using only ILD panning a stereo recording will sound wide but flat. Producers / engineers put echo & out of phase effects to make the stereo sound stage seem deeper. Problem with this is it still is “depth” bunched up around the front speakers. A good decoder, hardware or software, will take advantage of both these level & phase differences to create a full sound field around the listener.
I have used SpecWeb to make some very enjoyable upmixes, in fact sharing them with J Pupster. Based on your post I did what I should have done earlier & create a test signal that would give me a better idea of what SpecWeb was doing. I created an 80 second test in AA3 that is a clock sweep starting at center front. For front L/R blend it uses in phase blending -7.65 dB. For rear L/R it uses opposite phase blending -7.65 dB L/R. Center front is L=R in phase, center back is L=R opposite phase & center left = Left only, center right = right only. These are very well known quad encoding points. I fully realize SpecWeb was not created for quad decoding. I am using this because these are also very likely amplitude/phase points to crop up in conventional stereo recording.
Begin to 10 secs= center front
10 > 20 secs = right front
20 > 30 secs = right center
30 > 40 secs = right back
40 > 50 secs = center back
50 > 60 secs = left back
60 > 70 secs = left center
70 > 80 secs = left front
After decoding through SpecWeb audible impressions were: center front was strong & right where it should be. Right front was a vague blur of sound off to the right maybe a bit upfront. Center right came beautifully out of right back speaker. Right back was vaguely from the rear & to the right. The real surprise is that center back popped up right out of center front. This is obviously a stereo to surround decoder that doesn’t know front from back. No need to relate what happened on the left side as this was symmetrical.
Some of this is exactly how you described it that SpecWeb only pays attention to level differences between channels. I postulate that the most successful stereo to surround decoders pay attention to both amplitude & phase in a recording.
Anyone wishing to utilize my original 2 ch clock sweep test can find the file here:
Stereo Clock Sweep
Anyone wanting to hear the SpecWeb decoding can find it here:
SpecWeb Clock Sweep
Opening up the SpecWeb multi-ch flac in Audition yielded this:
In retrospect I think that my efforts at successfully pre-synthesizing for SpecWeb resulted from 2 things.
First is that out of phase blending before decoding widens the sound stage & nudges sounds panned to the sides even further out so they pop up in left back, right back. Secondly out of phase blending also reduces the amount of center front content in the original stereo. A common comment about SpecWeb is that the center front output is just too prominent & this would help reduce it.
