Impact of lossy compression when streaming Atmos

[riskylogic]

How do streaming Atmos mixes compare to the physical product (Blu Ray etc)? Do they lose anything or are they virtually identical?

https://www.quadraphonicquad.com/fo...lossy-compression-when-streaming-atmos.32330/

 

[LuvMyQuad]

Do you have a favorite example where the differences are clearest to you?

The release where I noticed it most was George Harrison: All Things Must Pass. I only have a handful of releases where I have a physical Atmos copy.

 

[popshop]

@BKarloff I think the differences are very similar to that of a 320 mp3 vs CD. Most people can't tell the difference - but in a good listening environment, listening to a good mix with the right instruments...you can hear and feel a difference. I always hear it in the cymbals.
The new Tears for Fears is a 10/10 release, and yes the bluray sounds more full and dynamic, with less compression.
I still listen to the apple music atmos anytime I don't want to wait 2 minutes for the bluray to load and go through the stupid menus. hahaha.

 

[harync]

https://www.quadraphonicquad.com/fo...lossy-compression-when-streaming-atmos.32330/

I think that example speaks to different choices when mixing for 5.1 vs. Atmos, and not the impact of compression.

I think Popshop's 320 Mbps vs. CD example is a good comparison. A mix with greater complexity would be expected to suffer more in a streaming version. But not enough to make me dig out the blu-ray most times.

 

[riskylogic]

I think that example speaks to different choices when mixing for 5.1 vs. Atmos, and not the impact of compression.

I think Popshop's 320 Mbps vs. CD example is a good comparison. A mix with greater complexity would be expected to suffer more in a streaming version. But not enough to make me dig out the blu-ray most times.

I suspect compression may have an impact on the mixing. If something is being mixed primarily for streaming then perhaps it is better to not load up the rear and ceiling speakers as much as you would if the mix target is targeted for a bluray release. As a related example, Soord says he tries not to put too much in the center or rear of his 5,1 mixes because the small speakers many people are listening won't support the dynamic range. I think Steven Wilson primarily mixes for bluray (or at least he was) and I think the nonlossy version sounds of his stuff much better.

I rip everything to my hard drive - playing it from there is as easy as streaming.

 

[harync]

I suspect compression may have an impact on the mixing. If something is being mixed primarily for streaming then perhaps it is better to not load up the rear and ceiling speakers as much as you would if the mix target is targeted for a bluray release. As a related example, Soord says he tries not to put too much in the center or rear of his 5,1 mixes because the small speakers many people are listening won't support the dynamic range. I think Steven Wilson primarily mixes for bluray (or at least he was) and I think the nonlossy version sounds of his stuff much better.

I rip everything to my hard drive - playing it from there is as easy as streaming.

I think there are multiple things going on...

1. Atmos mixes are not based off of the concept of channels of sound, but rather as a dome of sounds containing discrete objects. While somebody could approach an Atmos mix like quad/ 5.1 mixes (e.g., a guitar is only in the Left Surround), most go for ambience. Unfortunately this often leads to Front-heavy mixes.
2. There is no benefit to avoiding the use of the ceiling since the sound is derived/subtracted from the 7.1 mix. If you listen to the 7.1 non-Atmos version of that mix, you will hear those objects in one or more speakers. Again, this leads to have mixes that are generally more ambient in construction.
3. I assume most mixers are simply tasked with making an Atmos mix, and aren't thinking if it's a blu-ray they'll go one direction, and if it's streaming they'll go another.
4. Mixing in surround and Atmos is a learning curve, and there seem to be a lot more people trying their hand at it than ever before with mixed results.
5. Apple Music uses it's own Spatial Audio algorithm for binaural listening. Somebody can make the perfect Atmos mix for a proper Atmos system, but it will sound like crap on airpods. I suspect a couple Atmos mixes have been pulled and remixed for that reason (I think this is the most likely explanation for the extreme toning down of John Mayer's Sob Rock).

 

[dobyblue]

That pretty much seems to be a universal problem with streaming Atmos. The dynamic range and depth just isn't there. I have Yello - Point and Steven Wilson - The Future Bites on bluray and they sound much better than streaming from Tidal. Furthermore, whenever I switch from streaming to my hard drive just about everything sounds like a sonic masterpiece. I'm sure it's because the bitrate just can't handle it. Streaming has about the same bitrate as Dolby Digital and there's more channels. It's very lossy.

These are two titles I’ve compared also, to similar dismay.

 

[Stupy]

Given this long running discussion, I sat here and compared R.E.M.'s Automatic For The People - the bluray atmos from the boxset vs the Tidal streaming atmos (note: all played through 5.1).

Two things I note:
1. The streaming atmos is better than expected for ~782kbps; considering the atmos is ~9100kbps on the bluray.
2. The streaming atmos cannot compete with the bluray; I am most hearing weaknesses at higher (eg cymbals) and lower (eg bass) frequencies.

The streaming atmos lacks the clarity/depth of the bluray atmos. When you play the bluray version, even though the studio is all an illusion, it sounds like those instruments are being played right here in front of you - the detail is incredible. That's how a bass guitar sounds. That's how a drum sounds etc. In the streaming form it does not feel right - but not bad enough to turn it off or cry in your beer.

I recognise the future is streaming; but yes, there's no substitute for the blurays yet. And thank you Pink Floyd for your stand alone Animals... about to give that a good crack right now...

 

[humprof]

I wasn't quite sure where to post this two-part article on French mixing & mastering engineer Poussin (he apparently comes from Gilbert Gottfried's land of the one-named people), since it covers so much ground. The history of studio tape recorders (with detailed specs); tape preservation, restoration, digitization, and playback; the ins and outs of Serge Gainsbourg's studio albums and the challenges of remixing them in surround (it was a post by @haikubass about Gainsbourg in Atmos that led me to Poussin); a tour of DES Studio; a primer on Atmos mixing and the specifics of Poussin's approach . . . does this belong in Surround Artists/Engineers/Producers? Inside Atmos Equipped Studios? I finally decided to drop it here because of Poussin's answer to the question are there any major differences between a [compressed] DD+ file [for streaming] and the ADM file? Nothing new, really, but nice to have it laid out once again. Emphasis in the original, English occasionally a little creaky:

We compare in the studio, between the master and the version broadcast by Apple Music. The difference is obvious, the result is good, but we lose a little in dynamics, in precision of spatialization, in nuance in the musical return. How to explain this difference?

It’s quite simple to understand. Let’s take an example of a song, for 4 minutes of music an ADM file weighs more than 1 Gb, the ADM file is an uncompressed file containing the bed and the objects in 24 bits 48 khz. The corresponding DD+ file is encoded in 768 kb/s and obtains a file of 20 Mb. There is a ratio of more than 50 between the two files. There is no secret, the encoding in DD+ compresses the sound with an important loss of information to obtain this compression ratio. It is normal to have differences in the listening. To reduce the impact on the sound rendering, Poussin optimizes the use of the BED and the objects to obtain the best possible rendering after compression in DD+.

The alternative would be to use DD TrueHD with compression is lossy. The bit rate of Dolby True HD is not comparable to that of DD+, which easily goes up to 6000kb/s and even 10000kb/s. With DD TrueHD we find a rendering closer to the studio master.

https://magicvinyldigital.net/2022/04/22/from-tape-to-dolby-atmos-sacd-visit-of-des-studio-part1/https://magicvinyldigital.net/2022/...n-sacd-from-a-tape-visit-of-des-studio-part2/

 

[riskylogic]

I wasn't quite sure where to post this two-part article on French mixing & mastering engineer Poussin (he apparently comes from Gilbert Gottfried's land of the one-named people), since it covers so much ground. The history of studio tape recorders (with detailed specs); tape preservation, restoration, digitization, and playback; the ins and outs of Serge Gainsbourg's studio albums and the challenges of remixing them in surround (it was a post by @haikubass about Gainsbourg in Atmos that led me to Poussin); a tour of DES Studio; a primer on Atmos mixing and the specifics of Poussin's approach . . . does this belong in Surround Artists/Engineers/Producers? Inside Atmos Equipped Studios? I finally decided to drop it here because of Poussin's answer to the question are there any major differences between a [compressed] DD+ file [for streaming] and the ADM file? Nothing new, really, but nice to have it laid out once again. Emphasis in the original, English occasionally a little creaky:



https://magicvinyldigital.net/2022/04/22/from-tape-to-dolby-atmos-sacd-visit-of-des-studio-part1/https://magicvinyldigital.net/2022/...n-sacd-from-a-tape-visit-of-des-studio-part2/

"The difference is obvious, the result is good, but we lose a little in dynamics, in precision of spatialization, in nuance in the musical return."

Yep. That's why I still want the blurays.

 

[himey]

The front channels are drums & bass predominantly on left, rhythm guitar on right, vocals in middle. The center has isolated vocals and snare at a lower level than the fronts, plus the bulk of the guitar solo.

The side speakers have just the rhythm guitar and percussion (cowbell left, tambourine right) duplicated at around the same level as they appear upfront, with the net result being they image a bit further out into the room.

The rear speakers have just vocals (ADT'd lead + backing) again duplicated at around the same level they appear in the front speakers. However, the "Mr. Wilson, Mr. Heath" vocals are completely isolated in the rears. It's hard to tell exactly what's happening in the height channels because of the low bitrate, but they seem to be a low-level reflection of what's happening at ear level. You could probably mute them and not notice anything

Have you isolated the top channels of a streaming title Atmos mix, and compared that to the tops of the same Blu-ray title?

 

[sjcorne]

Have you isolated the top channels of a streaming title Atmos mix, and compared that to the tops of the same Blu-ray title?

Yes, the streaming version usually has artifacts while the Blu-Ray is clean (though it varies per title). That said, with all channels going it's usually hard to notice the artifacts.

 

[himey]

Yes, the streaming version usually has artifacts while the Blu-Ray is clean (though it varies per title). That said, with all channels going it's usually hard to notice the artifacts.

Which titles did you compare?

 

[sjcorne]

Which titles did you compare?

Shakespears Sister - Hormonally Yours
Tears For Fears - The Tipping Point
Steven Wilson - The Future Bites
Kiss - Destroyer
REM - Automatic For The People

 

[himey]

Shakespears Sister - Hormonally Yours
Tears For Fears - The Tipping Point
Steven Wilson - The Future Bites
Kiss - Destroyer
REM - Automatic For The People

Cool. Thanks. I will check out the ones I have.

Are you really saying that the "artifacts" in the top channels, due to lossy compression, is disproportionate to the artifacts in the bed channels? I can't wrap my head around that. Or are you saying something different?

Is your test decoder an Atmos receiver, or have you dug deeper with isolating the tops through software? Thanks again.

 

[sjcorne]

Are you really saying that the "artifacts" in the top channels, due to lossy compression, is disproportionate to the artifacts in the bed channels? I can't wrap my head around that. Or are you saying something different?

The rear surrounds and heights usually have the most artifacting, while the fronts and sides are relatively clean. But again, these things don't really stand out unless you solo specific channels.

 

[himey]

The rear surrounds and heights usually have the most artifacting, while the fronts and sides are relatively clean. But again, these things don't really stand out unless you solo specific channels.

I always thought that was the mix itself and not the (lossy) format.

Your post about the top channels of Taxman and how you can't get a true evaluation of it because of the lossy format, is an interesting subject. I think some audio samples of top channels, in a quad format, could go a long way to sort this out. I still don't get how lossy compression could affect one channel, or sets of channels more than the others, unless it was built into the compression, which I don't think it is.

 

[humprof]

I always thought that was the mix itself and not the (lossy) format.

Your post about the top channels of Taxman and how you can't get a true evaluation of it because of the lossy format, is an interesting subject. I think some audio samples of top channels, in a quad format, could go a long way to sort this out. I still don't get how lossy compression could affect one channel, or sets of channels more than the others, unless it was built into the compression, which I don't think it is.

Must have something to do with what French mixing engineer Poussin said in that story I posted yesterday, no? "We lose a little in dynamics, in precision of spatialization, in nuance in the musical return." The information lost in compression that's being sent to the height channels would be metadata, I would think. The question is: how does "loss of precision in spatialization" translate to audible "artefacts"?

 

[sjcorne]

I still don't get how lossy compression could affect one channel, or sets of channels over the others, unless it was built into the compression, which I don't think it is.

Is it possible that whatever lossy compression scheme they are using prioritizes certain channel sets over others? Maybe someone more familiar with how these things work can answer that question, but I can only report based on what I'm hearing.

I think posting samples of audio pulled from the streaming services on the open forum would be ill-advised.

 

[himey]

Must have something to do with what French mixing engineer Poussin said in that story I posted yesterday, no? "We lose a little in dynamics, in precision of spatialization, in nuance in the musical return." The information lost in compression that's being sent to the height channels would be metadata, I would think. The question is: how does "loss of precision in spatialization" translate to audible "artefacts"?

Losing a little dynamics is a bit different than "not being able to tell what is happening in the height channels due to the low bitrate".