Technical explanation of Atmos and DTS:X?

[FriedChicken]

Once again I'm confronted with the dichotomy of understanding. In the one corner: the layman explanation provided for by the dingest of dong youtubers. In the other corner, the math heavy white papers of concepts that take years of industry specific knowledge to understand.

Then here I am, smack in the middle. Decent understanding of engineering and math and sound and physics and whatever else I might need to know.

Can someone provide, or at least point to a decent explanation of what exactly Atmos and DTS:X are, and how they differ from existing surround formats?

 

[Dunring]

If you do a Google search of
Atmos and DTS:X
There's several excellent comparisons and detailed explanations.

 

[FriedChicken]

If you do a Google search of
Atmos and DTS:X
There's several excellent comparisons and detailed explanations.

If there is I haven't found it, I've found a bunch of crap "13 speakers and ceiling speakers"... well gee willikers, ya think?

 

[FlyingFreak]

Found one great talk about atmos on YouTube!

its in French though …

 

[xpistofep]

Once again I'm confronted with the dichotomy of understanding. In the one corner: the layman explanation provided for by the dingest of dong youtubers. In the other corner, the math heavy white papers of concepts that take years of industry specific knowledge to understand.

Then here I am, smack in the middle. Decent understanding of engineering and math and sound and physics and whatever else I might need to know.

Can someone provide, or at least point to a decent explanation of what exactly Atmos and DTS:X are, and how they differ from existing surround formats?

Two main differences from “surround formats”:

1) “immersive” formats have an additional height layer in addition to the surround layer (now renamed the bed layer)

2) surround formats are based upon playing static sounds in particular channels/speakers; immersive formats add ‘objects’ which allow the engineers to place/move the sounds in a more seamless (read less labor intensive) fashion in any of the speakers.

LFE (the.1/subwoofer) hasn’t really changed much.

Unless you want to go back the white papers, the differences between how Dolby & DTS implement immersive are not important enough for a high level discussion.

Where to discuss more detail?

 

[BJL]

Once again I'm confronted with the dichotomy of understanding. In the one corner: the layman explanation provided for by the dingest of dong youtubers. In the other corner, the math heavy white papers of concepts that take years of industry specific knowledge to understand.

Then here I am, smack in the middle. Decent understanding of engineering and math and sound and physics and whatever else I might need to know.

Can someone provide, or at least point to a decent explanation of what exactly Atmos and DTS:X are, and how they differ from existing surround formats?

There is a practical difference between Atmos and :X vs. conventional multichannel audio. In conventional multichannel audio the number of speakers used at playback is defined (and limited) by the mix (putting aside post-processing such as Neural:X). 4.0, 5.1, whatever, the number of channels determines the number of speakers that will be used. In an Atmos recording, you are not limited in that way. If you add speakers, Atmos will take advantage of the additional speakers. The recording is not a limitation.

This is how Dolby describes it: "Dolby Atmos content is mixed as audio objects instead of traditional channels. This means the content is not tied to any specific playback configuration. Further, the technology automatically adapts the object audio to take full advantage of the number and placement of your speakers, from systems with five speakers on the floor and two speakers producing overhead sound up to a Dolby Atmos system with as many as 24 speakers on the floor and 10 overhead speakers." (from Atmos Installation Guidelines at page 3, available from the Dolby website).

One advantage is that you can start with a more simple set up, and as you expand your system, the music (or movies) that you own will take advantage of your evolved system, unlike the old days of buying a recording in stereo, and the getting the 5.1 version, etc. (if you are inclined to get the newer mix), the audio mix in an Atmos recording adapts to your system, as it is treating audio as objects, not as channels.

I'm not familiar with dts:X, but I have read that it is conceptually similar.

 

[FriedChicken]

There is a practical difference between Atmos and :X vs. conventional multichannel audio. In conventional multichannel audio the number of speakers used at playback is defined (and limited) by the mix (putting aside post-processing such as Neural:X). 4.0, 5.1, whatever, the number of channels determines the number of speakers that will be used. In an Atmos recording, you are not limited in that way. If you add speakers, Atmos will take advantage of the additional speakers. The recording is not a limitation.

This is how Dolby describes it: "Dolby Atmos content is mixed as audio objects instead of traditional channels. This means the content is not tied to any specific playback configuration. Further, the technology automatically adapts the object audio to take full advantage of the number and placement of your speakers, from systems with five speakers on the floor and two speakers producing overhead sound up to a Dolby Atmos system with as many as 24 speakers on the floor and 10 overhead speakers." (from Atmos Installation Guidelines at page 3, available from the Dolby website).

One advantage is that you can start with a more simple set up, and as you expand your system, the music (or movies) that you own will take advantage of your evolved system, unlike the old days of buying a recording in stereo, and the getting the 5.1 version, etc. (if you are inclined to get the newer mix), the audio mix in an Atmos recording adapts to your system, as it is treating audio as objects, not as channels.

I'm not familiar with dts:X, but I have read that it is conceptually similar.

This is in line with everything else I have already read/seen, but says nothing about how atmos actually works.





I vaguely remember someone describing atmos as creating voxels of the listening environment, and then placing the audio objects accordingly. This was years ago. Basically rather than having discrete channels of audio, the "channel" of audio would be say front left, and the audio processor would then "place" the audio in that "position" using whatever appropriate speaker configurations and auditory tricks; Relying heavily on processig, and divorcing itself from pure stereo listening.

Unfortunately I don't remember where I got that explanation, and I haven't found any other sources with this explanation.

 

[amirm]

This is in line with everything else I have already read/seen, but says nothing about how atmos actually works.

I haven't looked at the details of implementation. But the basic idea is that at authoring time, you pick X number of objects. The authoring tools then let you locate and move them in 3-D space. These sound sources are compressed and stored with sideband data that expresses the movement. At playback, the "renderer" looks at how many speakers it has at its disposal and attempts to present that movement using them. As noted, some of the channels are stationary and others move.

 

[BJL]

This is in line with everything else I have already read/seen, but says nothing about how atmos actually works.

There is a brief video on using Logic Pro X to create Atmos mixes, maybe this (in addition to the helpful comment by Amir) will be of interest to you, and of course, there is much more on this site. https://professionalsupport.dolby.c...ntegrated-Dolby-Atmos-Workflow?language=en_US

 

[Chrispy]

This is in line with everything else I have already read/seen, but says nothing about how atmos actually works.





I vaguely remember someone describing atmos as creating voxels of the listening environment, and then placing the audio objects accordingly. This was years ago. Basically rather than having discrete channels of audio, the "channel" of audio would be say front left, and the audio processor would then "place" the audio in that "position" using whatever appropriate speaker configurations and auditory tricks; Relying heavily on processig, and divorcing itself from pure stereo listening.

Unfortunately I don't remember where I got that explanation, and I haven't found any other sources with this explanation.

What is "pure" about "stereo" modes to begin with?

 

[FriedChicken]

What is "pure" about "stereo" modes to begin with?

Unaltered by processing

 

[Reverend Slim]

Once again I'm confronted with the dichotomy of understanding. In the one corner: the layman explanation provided for by the dingest of dong youtubers. In the other corner, the math heavy white papers of concepts that take years of industry specific knowledge to understand.

Then here I am, smack in the middle. Decent understanding of engineering and math and sound and physics and whatever else I might need to know.

Can someone provide, or at least point to a decent explanation of what exactly Atmos and DTS:X are, and how they differ from existing surround formats?

DTS:X (and native Auro, while we're on the subject) are essentially based on hemispherical panners. So the mixer can steer sounds around a dome and the levels in each speaker are essentially a straight amplitude adjustment between known pre-defined speaker positions. DTS:X can also provide sound to speakers in addition to those known positions by using its upmixer. Pretty straight forward.

Atmos lets mixers steer sounds using XYZ coordinates in 3-D space. So unlike a straight hemispherical panner around the room boundaries where speakers are located, the mixer can also move sounds through the room. You have 10 locked objects that are static and located at the 7 expected ear-level speaker locations as well as 2 height-level locations (at the top mid speaker position). These are the "beds" and are essentially used to contain the channel-based sound in the mix. The mixer then has an additional 118 objects that they can steer dynamically around the space. The renderer can then take all this data and use any speakers available to it to try to place sounds at that same XYZ location in the listening space.

In the theatrical version, those 128 objects remain discrete and the renderer can scale to specific speaker locations which are defined when the initial setup is done for the cinema decoder. In the home version, those 128 objects get grouped down to about a maximum of 15, depending on the needs of the mix, and are more limited to pre-defined speaker locations. But since we don't need the same spatial resolution in a home space compared to a theatrical space, it translates well without losing any of the object movement perceptually. Any questions?

 

[Chrispy]

Unaltered by processing

Except that made during the processing of the "stereo" recording?

 

[FriedChicken]

Except that made during the processing of the "stereo" recording?

Yes, the professional, and not all recorded audio is processed.

 

[FriedChicken]

So it sounds like what Atmos essentially does is leave the processing up to the DSP, moreso than has been in the past, where the DSP does, if anything, only room correction. Basically it's building on top of the room correction and turning it up to 11. This makes sense for stuff like sound bars that can respond to their specific geometry.

So the next question becomes: who/how is this being implemented? Is there a single Atmos chip, or do different receivers have their own implementations of spatial audio? That would open a whole entire can of worms for testing... and be very subjective... "who has the best implementation of Atmos" kind of thing.

 

[Chrispy]

So it sounds like what Atmos essentially does is leave the processing up to the DSP, moreso than has been in the past, where the DSP does, if anything, only room correction. Basically it's building on top of the room correction and turning it up to 11. This makes sense for stuff like sound bars that can respond to their specific geometry.

So the next question becomes: who/how is this being implemented? Is there a single Atmos chip, or do different receivers have their own implementations of spatial audio? That would open a whole entire can of worms for testing... and be very subjective... "who has the best implementation of Atmos" kind of thing.

Atmos is a recording format/codec. The avr/avp isn't particularly processing it in a direct mode, but can tweak a bit in various sound modes. Atmos is controlled by Dolby with licensing, whether it's a particular chip I don't know (or whether DTS does similar for its codec/licensing).

 

[bungle]

next question becomes: who/how is this being implemented? Is there a single Atmos chip, or do different receivers have their own implementations of spatial audio?

I am pretty sure it is software that is licensed. Similar to Dirac Live.

 

[Reverend Slim]

So it sounds like what Atmos essentially does is leave the processing up to the DSP, moreso than has been in the past, where the DSP does, if anything, only room correction. Basically it's building on top of the room correction and turning it up to 11. This makes sense for stuff like sound bars that can respond to their specific geometry.

So the next question becomes: who/how is this being implemented? Is there a single Atmos chip, or do different receivers have their own implementations of spatial audio? That would open a whole entire can of worms for testing... and be very subjective... "who has the best implementation of Atmos" kind of thing.

It isn't really about DSP or room correction, as what it's essentially doing is just steering the sounds based on the data and your given speaker layout. For instance, if an audio object is located directly between two known speaker locations, the renderer knows to output that sound at equal amplitude in both speakers with a corresponding reduction to account for summation so that there's no level shift. Think of it as a realtime panner located in your AVR, and the Atmos track is a set of instructions on where to pan things and when.

The Atmos renderer is the same in every device, and follows the same rules. Higher end stuff like Trinnovs can do some remapping of speaker positions, but other than that, how Atmos performs is going to be consistent across all devices. There are no different implementations because they all have to use the Atmos renderer for output. What they do with that output after decoding (i.e. room correction) is another question.

 

[FriedChicken]

It isn't really about DSP or room correction, as what it's essentially doing is just steering the sounds based on the data and your given speaker layout. For instance, if an audio object is located directly between two known speaker locations, the renderer knows to output that sound at equal amplitude in both speakers with a corresponding reduction to account for summation so that there's no level shift. Think of it as a realtime panner located in your AVR, and the Atmos track is a set of instructions on where to pan things and when.

How is it not? It sounds like it’s turning YPAO and other room correction algorithms on it’s head, in a way removing a processing layer.

 

[-Matt-]

... In the other corner, the math heavy white papers of concepts that take years of industry specific knowledge to understand.

Please can you provide a link if you have found such?