Off topic in Sound Science. the new old moderation.

[sander99]

That signal processing is probably done at the Apple Music Store and the AirPods processing consists of using the internal gyro data to calculate and process the head tracking.

It would be troublesome - or next to impossible(?) - to deliver an already binauralized signal and apply headtracking afterwards. The normal way to do this would be to binauralise the already "counter-rotated" (adapted to the actual position of the head) virtual speaker- and object-positions, because the necessary signal manipulation for adapting to the head position would be different/independent for the different speaker- and object-positions.

 

[Davesrose]

Read the links! Apple music has ALAC lossless albums that are also Dolby Atmos. Apple earphones that have the H1 or W1 chip support spatial audio: which is processed from a surround Atmos track. The Apple Store does nothing but serve up a Dolby Atmos surround format to a computer, Apple TV, iPhone, or iPad.

 

[bigshot]

It would be troublesome - or next to impossible(?) - to deliver an already binauralized signal and apply headtracking afterwards.

Based on my experience with both the AirPods and AirPods Max, it sounds to me like the baked in processing is like an advanced version of Dolby stereo, where individual elements are matrixed into several stock channels… front vocals and dialogue, rear channel, etc. The matrix encoded 2 channel file is sent across Bluetooth as a stereo AAC file, and the chip in the AirPods Max decodes the matrixed elements and places them in virtual space according to the data from the gyro.

The encoded track delivered by the Apple Music Store isn’t anything like a binaural recording. It’s regular stereo with mix elements separated out and matrixed in different phase. Some elements are affected more by head tracking than others, and timing variations are applied differently to infer distance with some. If I had to guess, I’d say there are at least four matrixed elements.

For instance, vocals usually sit in the middle with a distant envelope applied to them. They are affected the most when you turn your head, which places them in front of you. Other elements may have different spatial envelopes and reaction to head turning. This places stuff differently in the soundstage. But it isn’t pinpoint perfect like with speakers. It’s more of a generalized feeling of depth and width. And the placement of elements with AirPods always falls into these stock slots. It isn’t the same as the placement on a Blu-ray version. Not as many options I guess.

Have you heard the AirPods playing spatial audio? My description might make more sense if you have.

 

[gregorio]

The file being played on my AirPod Max when I select Spatial Audio is not the same file you are playing when you select Atmos on your Apple TV.

It’s already been explained to you more than once and the information is readily available from numerous reliable sources including Wikipedia and Dolby themselves, which you’ve also been told. So you cannot claim ignorance as an excuse for make these false assertions and therefore you are now deliberately lying/trolling, which of course you know is unacceptable here!

They may have both started as the same basic master, but they don't sound the same and they don't have the same number of channels.

And how many channels do you think Dolby Atmos has? Of course they don’t sound the same because the file is processed (“rendered”) differently by your device after it’s been received.

Yours is delivered in a proprietary multichannel Atmos file format, and mine is passed across as a lossy two channel file using bluetooth.

Again, just lies! His is not “delivered in a proprietary multichannel Atmos file format” because there is no such thing! Dolby Atmos is NOT a multichannel format, it’s an object based format! Yours is passed across (to your player/device) as exactly the same format but what is delivered to your headphones or speakers is different because that file is rendered differently. How many times do you need that explaining to you?

Spatial Audio for headphones isn't the same thing as Atmos for speakers. I can't figure out how you would argue that it is.

It’s exactly the same ADM file, the ONLY difference is how that file is rendered by your device. We are arguing with you because you are making up false assertions contrary to the actual facts of what Dolby Atmos is and how it works. How can you not figure that out when by your own admission you are guessing?!

I can only guess that you don't have AirPods and have never heard Spatial Audio playback.

Wrong on both counts but don’t let that stop you from trolling!

It's signal processing to synthesize surround sound, just like what Boom 3D does, or those fake stereo gadgets they used to sell in the 70s. It isn't a multichannel format itself.

Again, stop just making up false assertions. Atmos is a discrete surround format, it does NOT use signal processing to synthesise sorround sound and it is NOT even vaguely like Boom 3D does, let alone “just like”!

Enough with the BS already!!

G

 

[gregorio]

Was Dolby Atmos designed purely for a multi speaker set up? so using headphones would be severely limiting the effect?

Bigshot’s “Yes and Yes” answer to your question is incorrect, you really need to ignore him on this matter. For some unknown reason he’s gone off into his own little world of trolling BS regarding Dolby Atmos.

The actual answer to your question is the opposite to Bigshot’s, it’s “No and No”, although the 2nd “No” is qualified.

Atmos is a sophisticated and very clever system with a number of design goals which took Dolby about a decade to develop! To appreciate those design goals you have to understand a bit of history. Up until the advent of the DCP (Digital Cinema Package) Dolby had enjoyed a virtual monopoly on all film sound for a couple of decades. Every theatrical film required a Dolby certified dubbing theatre and of course Dolby charged a fee for certification, plus they also charged a licensing fee to every film mixed in those dubbing theatres. Dolby made a fortune but that all came to an end with digital cinema because the DCP format only allowed film sound using discrete WAVs, it did not support Dolby SR-D (the digital surround used on 35mm film) or Dolby Digital. No more thousands or tens of thousands of dollar licensing fee for each and every single theatrical film made in the world. So, Dolby not only needed a new format that provided a tangible benefit over existing surround formats that was compatible with DCP but also had to be backwards compatible (IE. Support prior formats, 5.1 for example) and be adaptable as a consumer format. This combination of features is what enabled Dolby to dominate/monopolise with both the original Dolby Stereo and then with SR-D/Dolby Digital. Another design goal was to eliminate the problem of filmmakers having to make numerous different film mixes, at least a 5.1 mix, a 7.1 mix and a stereo mix. The result was Dolby Atmos. It is adaptable as a consumer format, IE. A single data stream (ADM file). As well as being compatible with the WAV only requirement of DCP (using Broadcast WAV “chunks” to embedded metadata for location/position/panning). One of the clever parts about this is that a certified dubbing theatre and specific hardware encoder/renderer and decoder is required to create the WAV/DCP files, enabling Dolby to once again charge for certification, charge a licensing fee to film makers and charge the cinemas for a decoder but the ADM file creation only requires modestly priced software, no certification for mixing rooms and no licence fee for content makers, only a license fee for the decoder/renderer chip or software in the receiving device.

So to answer your question: The consumer (ADM) version of Atmos is designed for any consumer speaker setup from a standard stereo (2.0) system up to a 24.1.10 setup, so not “purely a multi speaker setup”. The theatrical version supports setups from 5.1 up to 64 unique speaker feeds. The way this works is fundamentally quite easy to understand, although it requires that you drop the previous way of thinking, which was to associate audio channels with speaker channels. Instead we have audio channels or objects that are assigned co-ordinates in space (in 3 dimensions rather than 2, to accommodate height). These co-ordinates are stored in metadata embedded in the data stream (or embedded in BWAV “chunks” in the case of DCP). The decoder in your device has to be configured for the number and location of your speakers and then maps (“renders”) those coordinates of every audio channel/object to the appropriate speaker or combination of speakers in your particular setup. If you only have two speakers (stereo), then all those coordinates can obviously only be rendered to those two speakers/positions (or anywhere between them).

In the case of headphones though, those 3 dimensional coordinates can be rendered with a binaural algorithm, thereby preserving all that 3d spatial information and indeed certain binaural rendering information can be embedded in the ADM file. So the answer to your 2nd question is also therefore “No”, however there are a couple of caveats:
1. The binaural algorithm in the Dolby renderer (decoder) is based on a single HRTF, so it may or may not work well for you personally, depending on how well your hearing matches (or can acclimatise) to that HRTF. Although it seems possible/likely that Dolby will allow different HRTF options in the future.

2. Dolby’s binaural rendering engine can be bypassed, allowing third parties to implement their own binaural processing. This is a potential caveat because exactly what such a third party binaural processor does is only known to that developer. The developer could take those coordinates and “binauralise” them or might for example instruct the Dolby decoder to render say a 5.1 mix and then “binauralise” that 5.1 mix. Such an implementation would obviously loose the height information and “would be severely limiting the effect” when using headphones.

G

 

[Davesrose]

I think there already are Dolby Atmos headphones that have more individualized HRTF options. Take this pair of JVCs that come with its own receiver that renders Dolby Atmos, Dolby Surround (upconverts to a surround format), and DTS:X (DTS’s object based format). Wether it is able to use Dolby’s binaural engine or their own, reviews seem to indicate that it can simulate a 3d axis for many.

https://www.jvc.com/usa/headphones/home-theater-solutions/xp-ext1/

 

[sander99]

Some remarks about the JVC, also about how it is calibrated to the user (not directly a hrtf or prir/brir like measurement):
https://www.head-fi.org/threads/apple-airpods-max.957525/page-3#post-16397038
Oh, and the JVC has no headtracking.

 

[bigshot]

"Dolby Atmos" is a brand name, like "Frigidaire" or "GE". People tend to think of it as a particular product, and it does fit within a general product category, but there are so many flavors of Dolby Atmos, you can't specifically point to one single way of how it works. The basic idea is sound objects, but the way that is executed uses just about all of the tricks in Dolby's bag of tricks. And sometimes it's built on top of a 5.1 (or more) mix, so the directionality is there even before Dolby adds its magic. There are also different ways of calibrating. Those JVC cans look interesting, they use an entirely different system of calibration than Apple does. (Both pretty primitive in their own way I suspect.)

Dolby Atmos (specifically Apple's Spatial Audio) through headphones doesn't sound anything like Dolby Atmos through speakers. I think the reason for that is that the speaker variety is built on top of a file with multiple discrete channels, and the Apple headphone variety takes those multiple channels and matrixes them into a regular stereo file which is decoded in the cans, much like Dolby Stereo was in theaters back in the day. Another way of doing that would be what Boom 3D does, isolate frequency bands that relate to specific sound elements and process each one separately. In a sense, matrixing does this anyway, because certain frequencies lend themselves to being matrixed better than others.

Without head tracking and custom HRTF calibration, the best that can be achieved universally is a generalized feeling of depth. You can't get pinpoint sound location in front, above, below or behind without HRTF, and actual perceptible distance seems an impossible dream. I've listened to a lot of Apple's Spatial Audio using the recommended AirPods Max and it does subjectively improve the sound and it does at a more open feeling to the sound with better separation. There is a pretty solid channel devoted to voices that is processed to sound just a little distant in front in the phantom center. And you've got the left/right of stereo of course. But rear and above is subject to your own ear's interpretation of it. If you have a visual element representing the sound to look at, your mind forces it into the proper placement. But if there is no visual, it's placed pretty randomly. Rear channel stuff sometimes sounds to me like it's coming from somewhere just below my chin for some reason, or sometimes below me to the side. It almost never sounds like the rear unless it's a movie and the visuals make it clear that the sound is coming from behind. Even then, the directionality flickers around occasionally. It won't stay fixed. Maybe if it was calibrated it would be more consistent. It certainly isn't the "sound of the cinema" with sound coming from all around and above you like they tout in the advertisements. It's still headphone sound, just a little better headphone sound. It still all sounds very close to the head with strong left right and vague indications of the other directions.

When it was first introduced, there was a clear dividing line between Atmos and Spatial Audio in the Apple Music Store. Atmos was for the Apple TV and speakers and Spatial Audio was for the AirPods line. I think that separation was probably due to some sort of contractual agreement with Dolby. That agreement seems to have been revised, because recently Apple has been using the two terms interchangeably. Or maybe the advertising copywriters have just blurred the line. I don't know. In any case, it's best to talk about Atmos with speakers as something different than Atmos with AirPods because it clearly is different.

It would be very interesting if someone could capture a Spatial Audio stream from the Apple Music store and play it on headphones without the Spatial Audio processor built in. I wonder what it would sound like? I know the headphones are responsible for head tracking because you can do that without a Spatial Audio source. But I don't know exactly what kind of decoding the cans are doing of the sound field itself. I suspect that it's like Dolby Stereo in that the file without processing sounds pretty much like normal stereo, and you have to decode it to get the dimensional boost.

 

[gregorio]

People tend to think of it as a particular product,

And in this case, “what people tend to think” is actually correct!

and it does fit within a general product category, but there are so many flavors of Dolby Atmos,

There are not “so many flavours”. It could be argued there are two “flavours”, the DCP and consumer flavours but even that argument is tenuous because they’re both fundamentally the same, just some limitations with the consumer version to accommodate data bandwidth considerations.

you can't specifically point to how it works.

Bullsh*t! I can and have, Dolby themselves have, so have other various sites. And how on earth do you think engineers could even monitor let alone create Atmos mixes if they didn’t understand “how it works” in terms of audio channels, beds, objects, positioning/panning, routings and actually creating the deliverables required by the distributors? Use your brain bigshot, especially as you claim professional experience with audio, not to mention the obvious, that just repeating the same BS does not eventually make it true!! Jeez!

G

 

[castleofargh]

As said before, atmos is atmos, object based format(each sound source has its own spatial coordinate) instead of something based on a fixed number of output channels. If the decoder is told to output 2 channels, it will work out how to best locate all the sound sources with those 2 channels. If we play the same atmos audio on a 5.1, then the decoding will use 5.1 to place the sound sources. Etc.
That’s the benefit of atmos(or the dts version).
If the output is defined as stereo, the decoding will be for 2 speakers and that will sound meh on headphones(like usual stereo on headphone where you need great imagination to think it’s good spatialization).

But if you use something specifically made for headphones with Atmos, then the secret ingredient has to be HRTF based. It doesn’t really matter if the determined output is 2 or more channels as each one can be convolved with an impulse associated with a given direction(matching the selected number of channel in a standardized speaker setup and using the impulse associated with that direction in the HRTF model).
And if we were to add head tracking on top of that, then it’s more impulses that are needed for convolution following the HRTF model for where we look at instant t+little lag for tracker and processing.

Atmos is a pretty good and flexible solution. Any headphone solution extra comes on top of it and is going to be limited by how close the HRTF model is to that of the listener. No miracle for that part. If apple measures the distance between the ears, then it can improve a little bit on the customization of the HRTF, making more people feeling like it’s fine, but I have no idea if they bothered with it?

 

[gregorio]

Atmos is a pretty good and flexible solution. Any headphone solution extra comes on top of it …

Not necessarily. Not sure if you’ll follow this but what happens in practice is that audio channels in the DAW are routed to the Dolby Renderder (encoder), up to 128 of them assigned to either (user definable) beds or to objects. The real-time output from the Renderer is whatever you choose 7.1.4, 9.3.6 or whatever depending on your monitoring setup, one of the options is a stereo binaural output. Typically a number of different formats are output simultaneously, which allows instantaneous switching between formats and for certain requirements, such as loudness metering in 5.1 format, up to 128 channels of these various formats can be output. These real-time output channels are then routed either back into the DAW, directly into the routing matrix of the interface or to both. When the mix is finished, it is printed (either in real time or offline if desired in the case of the renderer software on the same Apple Mac computer). What is printed is a single ADM file (or multiple ADM files in the case of stem mixes required for film/TV sound delivery). That ADM file contains data regarding the binaural headphone mix but I don’t know exactly what data. What you upload to Apple Music is that single ADM file, no different format mixes as stated by bigshot.

I know this for a fact because I was very recently involved in a Dolby Atmos music mix that was delivered to Apple Music for distribution and I know exactly what was done, exactly what Apple required and exactly what was delivered! Just prior to that, I was involved in a Dolby Atmos film mix for distribution on Netflix (Dolby Atmos is now Netflix’s default audio delivery format), so I’ve now got a reasonably good handle on how it works and how to create a Dolby Atmos mix.

G

 

[Davesrose]

I know this for a fact because I was very recently involved in a Dolby Atmos music mix that was delivered to Apple Music for distribution and I know exactly what was done, exactly what Apple required and exactly what was delivered! Just prior to that, I was involved in a Dolby Atmos film mix for distribution on Netflix (Dolby Atmos is now Netflix’s default audio delivery format), so I’ve now got a reasonably good handle on how it works and how to create a Dolby Atmos mix.

I know almost all TV/movie streaming services are sticking with Dolby audio formats (IE carrying Atmos positional data on a 5.1 DD+ stream). For Apple, it seems they like to have some of their own proprietary audio codecs. Apple TV allows streaming content to have Dolby Atmos if it's DD (If I have ripped a UHD disc, I can get TrueHD Atmos on Android or Windows devices-not Apple TV). I've heard from devs that Apple isn't allowing bitstreaming of TrueHD or updating to Dolby MAT v2.0. Though I believe Apple might be utilizing MAT for Apple Music? I've noticed more albums being listed as lossless and Atmos (and Apple says that their lossless audio is ALAC). Perhaps the ADM gets encapsulated in that or Apple's lossy format in different situations?

 

[gregorio]

Perhaps the ADM gets encapsulated in that or Apple's lossy format in different situations?

No idea, the single ADM file is what Apple Music requires for delivery, what they do with it after that is up to them. I do know that the ADM file can be encapsulated in a number of container formats, at least Alac, DD+, TrueHD and MP4 although I’m not sure how it could be lossy encoded or how much data compression you’d gain as obviously you couldn’t lossy encode any of the metadata.

G

 

[Davesrose]

No idea, the single ADM file is what Apple Music requires for delivery, what they do with it after that is up to them. I do know that the ADM file can be encapsulated in a number of container formats, at least Alac, DD+, TrueHD and MP4 although I’m not sure how it could be lossy encoded or how much data compression you’d gain as obviously you couldn’t lossy encode any of the metadata.

G

Yeah, when I look at re-encoding movies, I see there's a DD+ stream and most home tracks get 16 Atmos objects (who's going to have a home theater with 32 speakers?). My iPhone does list lossless and Atmos (which even when I'm on the road, I've got 5G so the full audio track is probably ALAC with Atmos stream). Assume if there's a slower connection, or certain other transmission speeds, Apple is using their AAC (and could put the Atmos metadata stream in like DD+ Atmos).

 

[castleofargh]

When I got to see a demo of SPAT Revolution a few years ago, it did have a binaural output and I could even use "my" impulses. So I get that it can be an output solution or that we could even decide to just make that output the audio file. Where the determination of channel(real or virtual) happens doesn’t change how for headphone playback, we need to first determine a position to use the relevant impulse of that direction from the HRTF model.

Beyond that, I guess that at least in principle we can do anything anywhere. Carry object based data all the way to some AVR, or have each individual sound source convolved with their own impulse for their own directions(from the hrtf model for headphone output). *Cry in Processing*.
Or the same thing with many virtual sound sources could already be turned into say a virtual 7.1 and the file only has those as virtual sound sources.Then a listener would have a software do the convolution just for those 7.1 directions on headphone.
Or the entire process could already be done and the files are really just the final binaural data but it’s called atmos something for headphone because it has the cues for spatial whatever that works great if we have the very average head that probably was used for the HRTF model.
That of course couldn’t allow head tracking as the impulses need to change to match direction. But the first 2 options, and final processing on the Iphone or in Windows, would be possible.
No matter when the processing happens, my point is that the HRTF stuff for spatial sound on headphone, comes after atmos decoding into a discrete number of channels to set a final number of real or virtual speakers. At least I fail to imagine a workaround.