3D mixdown to 5.1.2 (or 5.0.2)

[Dewdman42]

Alright, guess I'm going to try to make a MP3D user preset to handle 3D-to-5.1.2 mix down..since one from the factory isn't included.  Unless Dietz has a better idea how to do it...

From what I can see I would need to create a coiefficents file that is similar as the one used for 7.1.2, but with different values for the matrix mixer.  That would involve using third party utility to create that file I guess?

Or perhaps I could use the 3D-to-7.1.2 factory preset to start out and adjust the matrix mixer a bit to make it more appropriate for 5.1.2.  I think I would need to edit that user preset while MP3D itself is in 7.1.2 mode in order to see all available matrix faders and to reorganize the fading without the 0 degree sides present.  Then save it as a new user preset that way and try it with 5.1.2 channel configuration.

open to suggestions.....

[Dewdman42]

So here's my first attempt, I did the following to create a 5.1.2 set of coefficients (see attached image).

When I imported this into a 5.1.2 MP3D mix down preset, for some reason that saved JSON assigned channel 2 to CENTER instead of RIGHT and so on..I don't know why it skipped channel 2.  Very possible I didn't have some of the settings correct when I exported the coefficient file.. In any case, at least it had the right set of virtual loudspeaker placements to represent 5.1.2.  So Then I just adjusted the matrix faders so each one was going to the right output for 5.1.2....and basically it seems to work and sound pretty good...so I must have done something right...but my question for Dietz...is whether I should fine tune this a little more, I notice some of the factory presets have some cases where coefficients are sending to more then one Matrix fader for some reason, and there were numerous settings on the coefficient editor that I didn't know what to do with..

Lastly if there is some easier way to handle 5.1.2 (5.0.2), would appreciate any insight..  This helps me a lot as my LogicPro session on my 12 year old MacPro is able to mix 100 channels of 3D audio without dropouts and it sounds really great through binaural headphones...

[Dewdman42]

So here's round two.

I added some imaginary loudspeakers to the coefficients for two reasons:

1. I read on the AllRADecoder website about using imaginary loudspeakers to create symmetrically balanced triangles, which must have something to do with the way ambisonics processing works. So anyway, I added a few for that purpose.
   
   
2. I read something in there about needing an imaginary loudspeaker below ear level set to gain=0, for some reason that is not completely clear to me right now as to why that would be needed, but what the heck I added that one too.

I also managed to configure it to skip channel 4 (LFE)... I didn't mess around with changing any of the matrix faders, I just have each non-imaginary loudspeaker going to the relevant channel output, fader set to full on each one, until I hear otherwise.

The attached image shows the configuration... json is also attached.

Note I also added some imaginary loudspeakers at the sides, which may not have been necessary or desirable I'm not sure, but it had the effect of facing the side triangles slightly forward and back...dunno...just experimenting now..

Unknowns:

Not sure what the "weights" parameter is for or if specifying the order as 1st, 2nd or 3rd makes any difference in this use case as I suspect not all of the JSON decoding configuration is being used by MP3D.

dewdman-5.1.2-b.json.zip-1696259641396-o1sg6.zip dewdman-5.1.2.vmo.zip-1696259641397-m18xv.zip

[Dewdman42]

(never mind)

[Dietz]

Hi,

 

VSL has deliberately chosen _not_ to implement a full-featured coefficient designer for spherical Ambisonics decoding directly in MIR 3D. Such an advanced tool requires some time to get used to and would have overloaded an already very feature-rich application like MIR 3D by far. Please understand that we therefore cannot and will not provide detailed support for AllRADecoder - it is an application released and supported by the smart people at IEM Graz.

Nevertheless, I will try to create a dedicated 5.(1).2 JSON file for MIR 3D during the next days. I must admit that I've never heard of anyone working in this format (... in fact, none of the DAWs I work with support it), but your reasoning about CPU consumption makes sense. You would still need to open MIR 3D in a format with a higher channel count for it to work, and you might need to re-route the channels derived from MIR 3D to match Dear VR Monitor's input configuration.

Best,

[Dietz]

There's a quick approach for a 5.(1).2 spherical setup in a 5.1.4 frame (see ZIP in the attachment). You'll have to open in a 5.1.4 channel. The .2 top layer channels will be routed to top front and rears, but you can easily change that yourself if your DAW support 5.1.2 directly (Nuendo doesn't). I included a screenshot to illustrate the idea.

The VMO-file (for Vienna MIR Output) in the ZIP goes to MIR 3D's "Output Formats" folder (C:\Users\[YOURNAME]\AppData\Local\VSL\Vienna MIR Pro 3D\OutputFormats under Windows, or Library/Application Support/VS/Vienna MIR Pro/Output Formats on MacOs).

Personally I don't like this format a lot, to be honest. It's quite remarkable how big the loss of spaciousness is compared to 5.1.4. 😊

5.(1).2-MIR_3D_HOA_(5.1.4frame).vmo.zip-1696259642033-z57nv.zip

[Dietz]

PS: You might be better off in general by a coefficient-based 5.1 plus two capsules pointing towards the ceiling.

.... oh, and here's the JSON with the coefficients I used to create the Output Format I posted above (s. attachment).

5.(1).2_DT_8U002722_canonical_basic.json.zip-1696259642226-e1kbx.zip

[Dewdman42]

I did try 5.1.4 and it killed my CPU also.  

The 5.1.2 one I made works fine though in terms of CPU, unless there is some sonic problem with it because I don't know anything about ambisonics...  

The LogicPro CPU is getting hit by having too many channels per plugin being processed.  With 100 tracks of that, it turns into 1200 audio channels at 7.1.4.  Regardless of the fact that using a downmix reduces how much audio is actually being transmitted on a subset of those channels, it brought my system to its knees.

Its nearly cut in half with 5.1.2.  So it makes a big difference for me here.  My home atmos system is also 5.1.2, so I don't find that format to be that odd at all!

Of course I can also just use 5.1.  Or even Quad!  In the quad case, you didn't provide a 3D-to-quad downmix...there is just the original mic array for quad...but may be ok if I wanted to do that, or 5.1 if I want to use your improved downmixing.  Eventually I will setup quad loudspeakers in my studio, but I'm not likely to add center channel or LFE in there.

But I quite like the sound I am getting from 5.1.2 through binaural headphones, which is the best my machine can handle.

I didn't find it that difficult to use the AllRADecoder...  Since you are unable to comment on what I did, I will just have to be ok with it and stop wondering if I did anything stupid.  ;-)

[Dewdman42]

@Another User said:

.... oh, and here's the JSON with the coefficients I used to create the Output Format I posted above (s. attachment).

 

Thanks I will check that out too.  My initial impression after reading what you did is that this would not have as good of a sonic impression as a true ambisonic 5.1.2 because I think the ambisonics engine would take what normally might end up in the 5.1.4 front-ceiling, and with 5.1.2 spread it out to both the side-ceiling and front...such that it will sound somewhat in between.  Obviously more speakers in the array means more precision, but with lack of such speakers...then that is the better option.   Mearly downmixing the 5.1.4 virtual speaker array to 5.1.2 of course blurs the true fore-aft spatiality.

[Dietz]

@Another User said:

My home atmos system is also 5.1.2, so I don't find that format to be that odd at all!

No doubt about that - it's just not supported by the DAWs mentioned above, sorry. 😊 Makes it quite a bit harder for me to do preset work in that format.

[Dewdman42]

@Another User said:

My home atmos system is also 5.1.2, so I don't find that format to be that odd at all!

No doubt about that - it's just not supported by the DAWs mentioned above, sorry. 😊 Makes it quite a bit harder for me to do preset work in that format.

makes sense!  B-)

[Dewdman42]

@Dietz said:

.... oh, and here's the JSON with the coefficients I used to create the Output Format I posted above (s. attachment).

Thanks again for making that coefficients file and posting it!  That is very enlightening.  Plays back on my system and sounds great.  I think your coefficients has a bit better stereo separation then the one I did last night, due to more triangles 😉  Other subtle changes you made to the gain settings and radius of each vortex...  Thank you.

I guess the reason you have the TRR and TRL faders duped is because your DAW only supports 5.1.4 or something of that nature...so you're having to approximate a 5.1.2 playback on that daw channel configuration.  On my system, I will be using 5.1.2 channel configuration, so I shouldn't have to dupe any audio to two faders in the matrix.  I actually won't even see the TRR or TRL matrix faders in my case.  it just labels the two ceiling speakers as TSR and TSL,  intended to be at 90 degrees azimuth  rather then 45.  

Thanks again for the follow up Dietz!

[Dewdman42]

ps I wish I could figure out how to make screenshots show up on the actual posts like you did above...  

[Dewdman42]

@Dietz said:

PS: You might be better off in general by a coefficient-based 5.1 plus two capsules pointing towards the ceiling. 

I wanted to ask you about this too though.  Are you suggesting that I should consider taking the above VMO and add two virtual microphones pointing towards the ceiling, in combination with the 5.1.2 coefficients file for virtual loudspeakers?

[Dietz]

No, this would be a different approach. 

[Dewdman42]

@Dietz said:

No, this would be a different approach. 

When I get my actual studio setup with some rears for "quad" it sounds like I might want to look into something similar like that though, will cross that bridge later... Really loving the sound of 5.1.2 in binaural headphones for now!