Stereo expanding & binaural tricks
So I realize I’ve been mostly learning the synthesis part of sound and mixing and sequencing grooves, but I have been neglecting stereo placement.
With the new advances of binaural recording, how could I implement some of it’s features in drambo? (fake, more precisely)
What I mean, is how do I place sounds in a 3D field?
PAN, delay, reverb, of course. Reverb is possibly height, while delay is possibly distance. But binaural also plays with filters.
Is this documented somewhere? What freqs or envelopes I should use to simulate a sound traveling around my head?
Any recommendation on stuff I could read to understand this better? There are a few papers, I’m checking on those but most goes over my head... (edit: unintentional pun, I swear)
My dream is to have an XY (or for 3D: gyroscope!) controlled FX rack that takes a stereo signal and moves it around like I’m a fire-dancer.
Any help appreciated.
Comments
Would something like extracting the difference form the L&R channels, applying filter/amps with careful selected envelopes (an LPG of sorts, maybe), and then summing back to the L and R bands start to do the trick? Or am I way off basis?
If you have in-ear microphones, you could record a typical impulse like what's commonly used in recording IRs and have a good loudspeaker send that. ideally, you (or a model head wearing them) would sit in a room at a steady place, while you record multiple impulse responses with the speaker changing its position.
That's some work for sure but since the impulse responses are likely quite short, you could run multiple convolutors simultaneously and switch or crossfade between them for virtually positioning the sound in the room.
The results should sound quite authentic because you're recording the real responses instead of trying to model them with more or less success.
If you're happy, you could repeat the process in different rooms and have your own library of 3D spaces.
You could have 3 processing streams, for example:
If you need help for crossfading between all these, just ask ☺️
" Reverb is possibly height, while delay is possibly distance."
nope, both are distance ...
the better simulations scan your ears to do the dolby atoms thing
as this seems essential to how you experience refections of sounds around you ...
logic on desktop has surround panners and shit
I guess they are in the iPad version too?
if you can export as whatever the dolby format is - you sure can export as binaural too ... ;)
Hmm, I see where you’re heading. I’ve been considering getting a pair of those binaural headphones.
But what I’m after right now is faking it. I would settle for a Suzanne Ciani’s 4 speaker type of simulation where you could just make the drums (or anything) appear to jump between corners of the room
I’ve had this argument before 😅
Let's say you want to emulate a church bell on the distance. Delay isn't enough, you need reverb to make it sound from above.
Or maybe it's just me :)
stereo expanding
for starters use true stereo panning and not the "balance" on the mixer
Intringuing... I'll sure take a look later, thanks!
Yeah, I guess this is the way if I want some kind of decent results. But convolution is not one of my priorities right now (so much to learn, so little time). I was hoping some clever time FX and filter modulation of the mid-side and/or L+R mix would do the trick.
In the meantime I found a lot of reading material, which I probably should have gone through before starting this thread.
big daddy
Interesting read, are you hinting at the “pan pot”? But as revolutionary as that must had been back then, now I want a pan-pot for front and back :)
did you read it all?
imagine true stereo panning over 5 speakers - front left, front center, front right, - back left, back right
it had it all already ...
another hint
you dont need the back speakers much ;)
the action is happening infront of you ...
so we are basically working with a stereo soundstage (with extra center speaker) ... 😁
its a lot to fold your head around,
let it sink
im not sure why everybody wants to sit in the middle of the orchestra now because its possible? 🤷🏻♂️
its not a good idea,
the orchestra was infront of you since forever and nobody ever complained "the music doesn't sound 3D enough" I need to sit in-between these musicians 🤷🏻♂️ ...
I was thinking more of having the orchestra swirling around me, than being in the center.
Like a kick that beats once in front and once behind you. Or some hihats that spin around your head
Im reading on the subject to have a better grasp, since I may be thinking thiis is easier that it really is
Try using phase inversion, it's an old technique that is used a lot in film foley and sound effects.
You'll have to experiment a bit but it can be done,
Aside note, high frequencies travel faster than low freq's so close up sounds have
more high-frequency content than sounds that are further way so maybe try using
eq's and cutoff filters to create distance as well as depth alongside reverb and panning.
Thinking about it you could include using MS-decoder's with sounds that have stereo content.
drums dont like that 🤷🏻♂️
keep in mind if you are messing with the pan you also mess with the level
try pad sounds ...
experiment and take your time
dont overdo it
there are records that are famos for almost being unable to listen to because of the wild panning ;)
white album - the Beatles , cough
(stereo records where new and they didnt know how to do it right yet, so they were hm, "very creative" 😣 with the pan ^^, if you try to listen to it with headphones your ears are about to fall off 😂 )
@lala
“(stereo records where new and they didnt know how to do it right yet, so they were hm, "very creative" 😣 with the pan ^^, if you try to listen to it with headphones your ears are about to fall off 😂 )”
Lololol….so you noticed that as well. 😂
You can do a lot of stereo tricks with this filter bank I made. Simple panning of different bands, but also static and echo delays, or sending bands to different effect chains from the outputs.
😁
That's very cool, I'll be sure to take a good look, I'm sure I'll learn something 😅
Meanwhile I tried a simple proof of concept using @rs2000 suggestion.
I only used 7 HRIRs corresponding to elevation 0º to 90º (those are included in the patch, since they are free).
I took the IRs from here: http://recherche.ircam.fr/equipes/salles/listen/ ( I took the first subject -1002)
The patch is simply 7 convolvers each with the respective IR, x-faded by a N-1 switch. I tried with a variety of input sounds, but the effect is not very noticeable.
I haven't tried the "compensated" version of the IRs yet. I'll see if it makes a difference. The are all taken at a 195cm radius, so no variation there.
Her it is:
But one can see already how expanding this technique for the remaining positions, would be a very tedious process. This particular database has 187 IRs per subject and all at the same radius (so two dimentions left azimuth/elevation).
So even forgetting raius, imagine creating 187 convolvers, not to mention the clever switching that would have to be implemented between them to have independent sliders for azimuth and elevation select the appropriate convolver (but I'm sure it's doable). Multiply that by several different radius and it would be a nightmare to build.
I might play with these concepts for a while longer (I'm learning convolution in the process) but I'm not very enthusiastic.
For starters, does anyone knows of a better database? There are a few free ones online, but I don't what to go through hundreds of files to find the best ones, if anyone can spare me the trouble.
" The are all taken at a 195cm radius,"
these won't work ;)
its to much of the same thing ...
what you want is 1meter away 5 meters away 20 meters away ...
going around in a circle of 2m in some way ... won't do much, its not different enough
its not only position, distance is key here ...
@pedro yes, that's exactly the way to go!
Don't give up. Elevation is the least noticeable of the 3D movements.
If you use IRs from the horizontal plane (with the sound source moving around you and recorded with good microphones placed in your ears or in a head model, the effect would become much more noticeable.
@lala
Thanks, but I realize that. I mainly just wanted to know if it would be as simple as grabbing some HRIR wav files, loading in a convolver and that's it. Well, apparently it is, but there's also a lot more to it..
I'm gonna try with better IRs because these are kinda old, I just grabbed them for convenience.
@rs2000
l have been studying on the subject of SOFA files which are the modern version, if I understand correctly. SOFA files have the whole headspace encoded, but with some scripting tools you can extract specific IRs, like extract a wav IR for 45º left 60º up and 100cm radius. That could be interesting to generate just some "points" in the xyz plane, and x-fade between them. It wouldn't take an enormous number of IRs with some careful planning.
I don't know if I want to devote much time to this project, but there are a lot of already converted to wav HRIRs online, for instance this thread has several links:
But these are 7 channel wav files, what do I do with them?
Also I have been thinking how to implement the cross-fade when more than one dimension is involved (you offered to help 😅).
Let's say, I have my elevation x-fade going like in the patch I posted, but I also want an azimuth of 90 (which I understand is 45º left), should I have the stereo audio going through the 2 convolver banks in series? Like apply the IR for the elevation, then apply the IR for azimuth? (and then eventually a third one for radius)
I'm feeling it must require some left/right separation, not exactly like this but this type of L/R separtaion:
source: https://www.mattmontag.com/auralization/headphones/
left and right channel switched and maybe phase shifted ...mixed in after processing ... works well to blow up any sound
Roland dimension D (chorus) does that too ...
all this advanced stereo stage stuff is a nightmare if try to do it with mixer balance,
get used to the true stereo panner, its the only way out of this 😜
Here’s an MS-Decoder that you can use between mono and full stereo without mono.
It’s good on reverbs, delays, chorus etc, basically anything that has a stereo image.
I use it all the time on reverbs especially or if I’m finding difficulty in placing a sound in the stereo field.
If you open up the patch and tap into the mono and stereo signals directly you can add effects to those
and then merge them back again.
Quite fun.
Waves does a plug-in based on that principle using compression and eq.
@pedro Different from the diagram you've posted, I'd start with IRs taken from mono sources.
Have you found a set of IRs captured at different azimuth angles?
I should have made this two threads: one for stereo expansion and one for binaural tricks.
But once I’m back on my desktop I have a few points to add. Just need to do some checking first