5.1 bed creates 7.1 signal in internal Atmos renderer

[MAS]

The title describes the problem:
When I create a 5.1 bed within an Atmos project, the signal from the rear speakers (Ls/Rs) is also routed to the side speakers (Sl/Sr). It doesn’t matter whether I define an audio track or a folder track as a 5.1 bed: I always get a 7.1 signal!
Even if I route a mono object to the rear left speaker, for example, I still get a signal from the side speaker (see picture). (Even when I select only that speaker.)

This is a serious bug! Can anyone confirm this?

I tested it in Nuendo 13.0.20 and Nuendo 12.0.70. Both versions have the same bug.

[MAS]

Thank you for your reply, but that is not relevant here. The signal is positioned (in the virtual room) so that it should only come out of the left or right rear surround speaker. Any “re-routing” would only be relevant during playback. (Or when I use the downmix in the renderer.)
When mixing in Nuendo, however, only the coordinates in the room are relevant. Or the channel layout of the bed. This is why the signal in 9.1.6 also has to come from the desired direction. This applies to objects and beds.

In addition, beds are ultimately there to address a specific loudspeaker layout. But if a 5.1 signal becomes a 7.1 signal, then a 5.1 bed is useless.

By the way: A 7.1 bed works as it should.

The problem also occurs with other renderer configurations (e.g. 7.1.4 or 7.1).

[Dietz]

It’s also funny that the “9.1.6” we see in Nuendo doesn’t follow the layout suggested by Dolby:

[MAS]

Because you picked the wrong 9.1.6 layout. Instead of “9.1.6 ITU” you have to select “9.1.6” (without “ITU”). I made the same mistake initially. Then YOU pointed out that there are two different layouts.

[Dietz]

Yes, but this naming confusion is just asking for mistakes, that’s my point.

… especially if you like to mix channel-oriented rather than Object-oriented (like me), because without the Dolby renderer active you just get “9.1.6”.

[MAS]

If Steinberg can add “ITU” or “Auro-3D” after the name of a layout, I don’t see why it shouldn’t work with “Dolby”?!
I posted a picture of PT in the other thread. I think Steinberg should adopt this and add “Dolby” to the appropriate layouts:

@Dietz

(…) because without the Dolby renderer active you just get “9.1.6”.

And this is the “ITU” layout?! (Except it no longer says “ITU”.)

I don’t have this problem. Even with the internal renderer disabled, I see both 9.1.6 layouts. In Windows, however, the list looks different. Instead of “10.2 Experimental” there is a “5.2.5” layout.

[MAS]

I ran the exported ADM file through the Media Encoder and then played it in the Reference Player to make sure it wasn’t just a display error. Unfortunately, this is not a display bug in the renderer. The “doubled” signal is also written to the ADM file.

[MattiasNYC]

@MAS

Thank you for your reply, but that is not relevant here. The signal is positioned (in the virtual room) so that it should only come out of the left or right rear surround speaker. Any “re-routing” would only be relevant during playback. (Or when I use the downmix in the renderer.)
When mixing in Nuendo, however, only the coordinates in the room are relevant. Or the channel layout of the bed. This is why the signal in 9.1.6 also has to come from the desired direction.

Actually Luis might have been right though.

When you pan your source signal that goes into the 5.1 bed the panning is correct, right? In other words it shows up correctly within the 5.1 bed group track.

The difference is what happens in the renderer, and when you write it would only “re-route” when playing back or downmixing well that’s exactly what you’re doing though - you are playing back through the renderer and the renderer upmixes that 5.1 signal into your output path that is larger than just 5.1. If you start switching between any 5.x.x format and 7.x.x or 9.x.x the latter two will always do what you show.

So the question really is if that’s the intended upmixing behavior of the Dolby Atmos system.

You could maybe test an ADM export and then re-import that to see if the actual 5.1 bed retains proper panning. I’m suspecting it does.

[MAS]

I was able to solve part of the puzzle: The documentation for the Dolby Reference Player (DRP) or Media Encoder (DME) gives the following layouts for 5.1 and 7.1:

5.1: L, R, C, LFE, Ls, Rs.
7.1: L, R, C, LFE, Ls, Rs, Lrs, Rrs.

According to this overview, both layouts use the same surround speakers (namely Ls/Rs). Internally, however, the renderer uses a different layout:

5.1: L, R, C, LFE, Ls, Rs.
7.1: L, R, C, LFE, Lss, Rss, Lrs, Rrs.

This means that a 5.1 signal in a 7.x.x or 9.x.x bed will be processed differently than in a 5.x.x bed. It would be helpful if this was documented somewhere, as it has a big impact on the mix.
For example, if you route your 5.1 music stem to a 5.x.x bed, you will get a free “upmix” from the renderer when playing back with a layout equal to or greater than 7.x.x. This is also problematic because renderers in consumer devices offer additional settings to further “process” the signal.

This explains why a 5.1 signal is processed differently in a 7.x.x bed than in a 5.x.x bed. But it does not explain why the renderer in a 7.x.x bed sends the 5.1 signal for the two surround speakers to the rear speakers as well. According to the Dolby Atmos Master ADM profile, Ls/Rs and Lrs/Rrs have the same coordinates (x, y, z).

My conclusion so far: You only have full control over the signal in a 7.x.x bed.
It would also be helpful if Steinberg didn’t just use “5.1” for a 5.1 layout. A more precise description like 5.1 Dolby (L, R, C, LFE, Ls, Rs) would be better.