The Fundamentals of Atmos (1 Viewer)

[JohnRice]

1) My current understanding, based on both that video and the official Dolby documentation (from the professionalsupport.dolby site, not their pretty aimed-at-consumers site) are that bed channels and objects are two distinctly different things.

I can understand that it might seem I said otherwise, but that was never my intention. What I was explaining, maybe badly, is that the 128 objects in a theatrical soundtrack can be divided between bed channels and dynamic objects. There is a total of 128, so if 10 are used for bed channels, that leaves 118 for dynamic objects. If fewer are used for bed channels, there are more for dynamic objects. It is in that way I meant they are the interchangeable. How they operate is different. That's all I was trying to convey. Until I see reputable information that, specifically, there are 128 discrete objects in, specifically, home Atmos, I will remain skeptical. That number is always and only expressed regarding the mixing of the original, theatrical mix. In fact, I recall that video early on mentioning the factors considered in "mixing down" the theatrical soundtrack for home use.

In Dolby's own documentation they mention the home version can recreate all the objects of the theatrical mix.

That's where the word play comes in. Yes, all the objects of the original soundtrack can be recreated, but that does not mean they remain as 128 discrete objects. If that was the case, they would make a point of specifically stating that all 128 discrete objects of the theatrical soundtrack still exist as 128 discrete objects in home use. Instead, my understanding is they are combined (into stems?) of far fewer discrete channels/objects. There remains the reality that the configuration home Atmos systems are all over the place. A sound mixer can never have the slightest idea how the soundtrack is played back. I would expect this would often result in a "less is more" attitude to mixing home Atmos.

In any case, this seems to have become more about nitpicking my comments than anything else. So I will move along. I do admit I got a little worked up about what I learned a couple years ago.

EDIT: I do want to add that when I was learning this a couple years ago I had no trouble finding references to the 16 channel limit, but when I did a very quick search a few months ago I didn't find any of those.

Maybe most important is, as Josh is referring to below, not just the system itself, but how it is actually used in the various mixes. Does it matter how many dynamic channels there could be, if they are rarely used? Also, as re alluded to below, and I swear I saw examples of this with Disney in particular, that there can be 7.1.4 bed channels. It's all very confusing, and I suspect that might be intentional.

 

[Josh Dial]

Very interesting video. Thanks, Carlo!

It doesn't change my overall position on Atmos: broadly speaking for UHD there are passive and active Atmos tracks. Those are my own labels. This is my "conspiracy". Not all Atmos tracks are created equal.

In passive mixes, the engineer simply and only uses the x.y.2 bed layer to play whatever sounds they assign to the height speakers. In these mixes if doesn't really matter how many height speakers you have. The "panning" will almost always be phantom/matrix, or, at the very most, a clever engineer will tinker with the content to create a panning effect in the same way they would with a car driving from back to front or left to centre to right front speakers.

In passive mixes there are no discrete objects floating in space (your height speakers). You may still hear panning but it's the "regular" old kind we've had for years back on the "ground".

In active mixes the engineer uses discrete objects with or instead of the x.y.2 bed layers. In the most clever active mixes, the sounds that need to be moved will be placed into a discrete object. That object moves through the soundscape however the engineer desires. So when the sound pans it actually is moving from speaker to speaker.

It's very difficult to actually determine by listening alone whether the track is passive or active. There are "good" passive mixes. Indeed, I've read reviews on this forum where members have praised Atmos tracks for having good "dynamic motion" even though the tracks don't actually use discrete objects. Ultimately one of the only reliable ways to confirm a mix is active is to use something like the Trinnov visualizer. If you take a look at the Dune UHD in the Trinnov visualizer, you'll see that it actually uses discrete objects. Moreover--and I'm far from an expert in interpreting the visualizer--it seems the engineer has an x.y.4 (not .2) bed layer. Why do I say this? Because at times the front left height speaker is "on", the back left speaker height speaker is "off", the front right height speaker is "off", and the back right height speaker is "on". So two diagonal speakers are "on" and two diagonal speakers are "off". At the same time, there are 7+ discrete objects floating around. I'm not sure how this is made, but it's what is happening in the visualizer.

For what it's worth, I don't have a Trinnov. A co-worker (actually retired now) has one so my above experience with Dune is from memory when my co-worker was gracious enough to let me "audition" the Trinnov 16 at his house. It's my next purchase, though.

 

[Carlo_M]

Until I see reputable information that, specifically, there are 128 discrete objects in, specifically, home Atmos, I will remain skeptical. That number is always and only expressed regarding the mixing of the original, theatrical mix. In fact, I recall that video early on mentioning the factors considered in "mixing down" the theatrical soundtrack for home use.

Your skepticism is both correct and, in a small way, misses the mark. There absolutely has to be a downmix for home theater. That is indisputable. Their theatrical implementation can and does have more discrete overhead channels than home theater. Absolutely agree. By its very definition, objects will need to be folded down as opposed to remaining discrete. This has been the practice since the beginning of recording. Everything was mixed down to mono. Then stereo. Then surround. Then 5.1. Then 7.1. Think of the average movie soundtrack. Orchestra can have 100+ instruments. Multiple dialogue tracks. Dozens or hundreds of sound effects. But it all needs to go down into 5.1, or 7.1 or 7.1.4 or whatever. Everything is mixed down.

The same for objects. Even for theatrical mixes. You can have up to "118 objects" (or whatever) but how many Atmos theaters do you think have 118 discrete overhead channels? I'm going to guess exactly zero. Most Atmos installations in your cineplex probably have a dozen or less (Dolby is cagey about this, they don't publish stats) but also think of the mixing rooms that use Atmos. My friend's husband works in a professional recording facility in Hollywood (they mix major Hollywood movies and he's worked with talent ranging from Tom Cruise to Tom Hanks to Rian Johnson) and had her ask him what their Atmos mixing room overhead speaker situation is.

His answer: 4.

Now they're a medium sized facility, so I'm sure Lucasfilm/ILM/Skywalker etc. maybe have 6 or 8, but I'd be surprised if they had more. So by definition, if you have an aggressive overhead sound mix with 118 overhead objects, they're already being downmixed at the point of creation of the theatrical mix.

It would be tedious and time consuming for a mixer to map out all 118 sounds to the maximum number of possible overhead channels, and in fact if you look at my screencap of Logic Pro (or watch the Pro Tools video) it doesn't ask you how many overhead speakers there are, or in which precise speaker to place an object. It asks "how do you want the object to move in this 3D overhead space" and you use the object panner to indicate how you want the sound to move around the room. That motion is encoded as positional metadata. Then it is the renderer's job to encode that movement per object, and then the decoder's job to recognize "okay I'm in Theater X with Y number of discrete speakers, how do I recreate this?"

The bottom line is: no mixer is encoding 118 discrete things and mapping them out to the maximum number discrete overhead channels because 1) there is no theater (outside of maybe Dolby's flagship) capable of that, and 2) as I showed in the Pro Tools and Logic workflow, it doesn't work that way (the mixer indicates motion using the panning plug in, not by mapping to a speaker location). See that right there is the essential difference for a bed channel vs. an object. A bed channel is something you assign a sound(s) to. Left channel. Right channel. etc. An object is given positional metadata for the decoder to then recognize what the playback system is capable of (2 overhead? 4? None?) and then use its internal algorithms to place it based on the system capability.

Think of it in this overly simplistic way. A traditional stereo bed channels: left and right. What happens if you only plug in a left speaker? You only get the left channel. Because that's how it worked pre-Atmos. Put sounds in these channels. But if every sound was object based, the decoder would know "well the only way to play this sound is to ignore the positional data and just play it" (assuming that's how the decoding algorithm was programmed) and then you'd get all the sounds out of that one speaker.

Even if you could theoretically map 118 discrete objects to whatever the maximum overhead discrete channels is (again, it is not how it works, but just for argument's sake) that's a metric crap-ton of work that no mixer in their right mind would spend time doing. Listening to Traunwieser, he mentions soundtracks are essentially the end of the chain and often everyone involved in making the film procrastinates so mixers have much less time than they'd like to make the final mix. Everything is already mixed down for the theatrical Atmos experience and it's mixed down a bit further for home.

 

[Carlo_M]

@Josh Dial

I completely agree that the quality and aggressiveness of Atmos mixes (especially upmixes of older non-Atmos soundtracks) are all over the place. Those that are "flagship" titles where money and time was spent (e.g. Star Wars Ep4-6) will sound good. Others will be nearly indistinguishable from their original soundtrack, or worse, sound gimmicky if sounds are put in places that they originally weren't meant for (some Atmos mixes of older music is guilty of this).

All I'm trying to do is 1) illustrate the differences between bed channels and objects, and 2) show that there isn't some behind-the-scenes conspiracy that the home version of Atmos is hobbled so badly that it's not worth going to a 7.x.4 system.

If it's an aggressive Atmos mix, with discrete use of objects placed in the .4s (instead of, as you say, a passive mirroring of already existing sounds in the bed channels) then a home .4 Atmos system will shine. If it's a mailed-in upmix, you'll wonder why you spent the money.

But at that point, the "blame" (if you even want to call it that) lies in the authoring of the Atmos track as opposed to Dolby trying to hide some fatal flaw in the home implementation. And again, given how John T in the video talks at length about the time crunch mixers are often in, being given the work (and notes/alterations from test screenings) just weeks before release date, I would say "blame" is too strong a word. Sometimes mixes are what they are because the mixers did the best they could in the time they had and had to kowtow to many competing/conflicting voices.

My personal rationale is this: I do not expect my .4 Atmos system to be extremely active in most movies I watch, especially if they're dialogue driven (dramas, rom-coms, comedies, etc.). I don't expect a Blazing Saddles 4K Atmos mix to suddenly have bullets whizzing everywhere when they invade Rock Ridge. But the new Dune at home sounds great in 7.1.4, and that's what I paid the $$$ for.

 

[JohnRice]

2) show that there isn't some behind-the-scenes conspiracy that the home version of Atmos is hobbled so badly that it's not worth going to a 7.x.4 system.

I don't know who you claim is saying this, but I have REPEATEDLY expressed that my concern is with the value of spending the money to expand BEYOND 7.x.4. Repeatedly.

The bottom line is, it‘s not remotely the same as theatrical Atmos, and its debatable whether there’s a real benefit to setting up a home system larger than 7.1.4.

Anyway, the takeaway is that anyone wanting to spend the significant $ to expand beyond 7.1.4 really needs to know the serious limitations they will face.

My goal was to explain the complexities and limitations of home Atmos in systems beyond 7.1.4.

My point is that, due to the limitations of the current HOME Atmos system, it's probably not worth going beyond the standard four surround and four overhead channels.

That's basically my point. How many HTs would get any benefit from a system larger than 7.x.4 anyway? Home Atmos processors can theoretically produce many more channels than that. The problem is, home Atmos systems work a certain way up to 7.x.4, but things change once you go beyond that.

 

[Josh Dial]

I completely agree that the quality and aggressiveness of Atmos mixes (especially upmixes of older non-Atmos soundtracks) are all over the place. Those that are "flagship" titles where money and time was spent (e.g. Star Wars Ep4-6) will sound good. Others will be nearly indistinguishable from their original soundtrack, or worse, sound gimmicky if sounds are put in places that they originally weren't meant for (some Atmos mixes of older music is guilty of this).

...

If it's an aggressive Atmos mix, with discrete use of objects placed in the .4s (instead of, as you say, a passive mirroring of already existing sounds in the bed channels) then a home .4 Atmos system will shine. If it's a mailed-in upmix, you'll wonder why you spent the money.
...

My personal rationale is this: I do not expect my .4 Atmos system to be extremely active in most movies I watch, especially if they're dialogue driven (dramas, rom-coms, comedies, etc.). I don't expect a Blazing Saddles 4K Atmos mix to suddenly have bullets whizzing everywhere when they invade Rock Ridge. But the new Dune at home sounds great in 7.1.4, and that's what I paid the $$$ for.

Total agreement.

One thing to add is that I've actually been pleasantly surprised by the DTS NEO: X and Dolby Surround upmixers. I've been re-watching every episode of every Star Trek series. I'm up to season 5 of Voyager. DS9 and Voyage have both sounded decent! Like nothing mind-blowing or anything, but whatever the software is trying to do it's doing fairly well. When the Voyager whooshes past in the opening credits it actually travels from my rears to the heights and then to the fronts. And the music isn't just merely shunted to the heights. Frankly the upscaled audio for Voyager sounds better than some of the "Atmos" tracks from Disney...

 

[Carlo_M]

Frankly the upscaled audio for Voyager sounds better than some of the "Atmos" tracks from Disney...

Although Traunwieser doesn't work on MCU films, a lot of what he said in the video may go a little ways in explaining the lackluster audio on recent MCU films.

We know that now, more than ever, MCU films are on a factory production type schedule. They have release dates announced years in advance. They also are prioritizing their own streaming service over physical release by weeks. Given the seemingly standard practice of getting the work to the mixers at the last step, often close to release date, and given how the mixers probably won't have the luxury to spend weeks on both the theatrical and then the nearfield mix, they may be cutting mixes designed to "work in both environments" which inevitably leads to compromises in dynamic range.

John T also mentions how when he does home mixes he has to be mindful to dial down the bass from the theatrical mix so that it doesn't blow out speakers in lesser home systems (like soundbars, inexpensive HTiBs, etc.). If Disney is shortchanging the amount of time their mixers can spend doing dedicated theatrical and home mixes, they may just be "playing it safe" and using conservative settings for bass and dynamic range for the overall mix.

Keep in mind everything I just wrote there is pure conjecture. But John T was pretty forthcoming about the challenges he and his peers face in modern mixing, and given how quickly these MCU films are churned out, I wouldn't be surprised if that played no small part in the declining quality of MCU soundtracks over the last few years. That and a lot of mixers working remotely over the pandemic, who had home setups created but maybe aren't as well calibrated as the studio rooms they were mixing in before COVID forced us to work from home (John mentioned he had a home rig built specifically so he could work during lockdown).

 

[Josh Dial]

John T also mentions how when he does home mixes he has to be mindful to dial down the bass from the theatrical mix so that it doesn't blow out speakers in lesser home systems (like soundbars, inexpensive HTiBs, etc.). If Disney is shortchanging the amount of time their mixers can spend doing dedicated theatrical and home mixes, they may just be "playing it safe" and using conservative settings for bass and dynamic range for the overall mix.

If I had to guess, I think this is definitely the case. However, Disney's releases have been lackluster for quite some time now. At least as far bad as at least Thor Ragnarok in 2017 and Cars 3 also in 2017. That was the year I first noticed having give what I call the "Disney Bump".

If I had to make a second guess, I think it's possible Disney engineers are, because of cost/time-saving measures, applying a simple curve to short-cut the job.

 

[Carlo_M]

Yeah I agree it was around the time of Ragnarok that I found their soundtracks to be of lesser dynamic range and bass impact. Good lord was that 2017? Feels like yesterday.

What's odd is, after posting here, I decided to pull out the UHD of Shang Chi, something which I also felt had compressed DR and lessened bass impact. Sure enough the first 30+ minutes of the movie was what I remembered, very little bass impact where I expected it (opening fight scene, couple of "needledrop" soundtrack moments where I thought the music should be "bumping") and of course the need for my receiver to be about +7db above most other discs (+10dB above where I play Christopher Nolan UHDs).

But then it got to the Bus Fight...and suddenly that big bass soundtrack thump kicks in and my living room is alive with sound. I stopped the film not long after (had to prep for dinner and also some more work from home to finish up) but I was shocked at how dynamic and impactful just that bus fight scene was. Why isn't the rest of the soundtrack like this?

 

[Doug2000]

This is one of the better videos I've seen explaining Atmos

Thanks Carlo for this. I've been traveling so havn't had time to dig into it yet. But wanted to say thanks.

 

[Sam Posten]

Welcome to the party, pal

Dolby Atmos Wants You to Listen Up. (And Down. And Sideways.)

True believers in the immersive audio format say it could restore a musical appreciation lost to a generation that has come up during the streaming era.

www.nytimes.com

 

[VonMagnum]

There seems to be one fundamental misunderstanding of how home Atmos works.

There are not "12" bed channels in home Atmos (supposedly limiting it to only 4 objects for 22 more speakers). In fact, there is only ONE bed channel in home Atmos, the LFE channel.

The 5.1 or 7.1 "bed" channels are combined with objects in the spatial encoding phase. Bed OBJECTS are recreated by the renderer. How many channels are present in the TrueHD or DD+ downmix have no bearing on the final Atmos channel output.

For example, if you play the Dolby "Conductor" demo from TrueHD with base 7.1 or the MP4 version with a base 5.1, you get the same fundamental output from the Atmos decoder (other than lossy versus lossless). The rear channels are in no way "missing" from the MP4 version. I have both here and they sound identical to my ears other than volume dialnorm differences.

So, if you want to count objects available for other speakers, you have to count ALL of them since other than LFE, they are all objects as far as home Atmos is concerned.

Overheads are NOT "bed channels" in home Atmos so they don't use up any of the 16 streams. Cinema Atmos includes 2 overhead bed channels. Any content in them plays over all overhead speakers in an Atmos Cinema.

Ready Player One is an example of a cinematic Atmos movie that got improperly converted to the home version. Most know it only had 2 overhead channels used in the home version (using lower speakers to guide any overhead movement illusory effects). The cinema version used two overhead bed channels and played those sounds in all overhead speakers. The home version oy plays in Top Middle (or split in Height/Tops as a phantom Top Middle if an actual TM isn't available). That was incorrect. It should have played in all overheads as an array (up to 10 in home Atmos).

If all 7.1 "standard" ear level streams/objects are in use in a home mix, you'd still 8 streams left for objects and those objects are spatial coded on every frame to premix/recreate the larger soundtrack as best it can on up to 34 speakers.

Speakers and objects have no direct relationship. Objects can vary in size, for example. One object can play over as few as one speaker or as many as all 34. So the idea that you won't get much/any use of speaker counts above 9.1.6 is simply not true.

The only real enemy of speakers not getting used is the "fixed object" method that some studios (Disney in particular) sometimes use (less common with more recent soundtracks than older ones, but still a potential issue).

That's one area DTS:X Pro has a huge advantage. Even a stereo soundtrack can be expanded to 30-channels. Sadly, DTS:X seems to be losing the war, largely because of streaming where it can't seem to get any traction, possibly due to the higher minimum streaming rates (data is money). Even Apple only streams Atmos at 768kbps lossy and DTS is less efficient.