LOUDNESS WAR - If you can't beat them join them.

[Neuro]

Thoughts in the head of an audiophile.

It is important to respect the musicians' artistic creation. Music production is a work of art.

Music today is digital. Why not initially mix the music to an optimal HIFI standard according to the musicians' wishes. The files are then converted using a special standard MASTER ALGORITHM that optimizes the sound for mobile phone listening. The algorithm must be so flexible with different publicly given input values that different productions get a different sound image on the phone. Can't be particularly complicated to produce. The end product provides optimal sound in mobiles.
In HIFI installations another standard ANTI-MASTER ALGORITHM is run which reconverts the digital music to optimal HIFI quality with the right dynamics, LUFS etc. The final product will probably be somewhat larger, which is probably not a problem today. Anyone who wants optimal sound must buy a new gadget - liked by the industry.

MASTER/ANTI-MASTER ALGORITHM gives the musicians the respect for their artistic work they deserve.
The problem is that the industry needs to agree on new standard algorithms for a small market - the audiophiles and the musicians. The cost in the mastering step should be negligible with a standard algorithm.

In summary - one music product for everyone. For the audio lover, the sound is optimized with additional gadgets/software.

Neuro

 

[Blumlein 88]

You grossly underestimate the complexity of mastering. You grossly underestimate what compression and limiting software does. So undoing that can create something so totally different it makes no sense.

You also would be requiring twice the work (or more) on the part of those people creating product. Not going to happen.

In theory something like that could be done, but it is going to be well beyond mastering, anti-mastering. Don't feel bad I had the same or similar ideas for a long time.

 

[somebodyelse]

If you could swing that sort of cross-industry change you could also swing mastering to consistent reference level rather than having replaygain do it, and introduce appropriate endpoint processing for the local environment, like compression if background noise is high, or adding harmonics for phantom bass with small speakers. Artists could get on with making the best recording they can, and leave optimising for specific environments to endpoints made for that environment. This is broadly similar to the standardisation on the movie side, but unlike that industry, the audio industry doesn't have the consensus necessary to adopt an equivalent.

 

[Blumlein 88]

Again I think you underestimate what all modern software for compression is doing. For instance, you would chop up a signal into maybe 3 to 5 different frequency segments ( can use more segments if you like). Then for each segment apply different levels of compression with different knees, different attack and release times different gain levels. I might do this one complex thing differently to several tracks which eventually get mixed together. These can sound dramatically different in the end result by varying those parameters. It takes skill and experience not to make a mess of it, but done well can be helpful for particular sounds. I might duck different tracks after or during all this. Ducking is when you let the level of one track control the level of another track. Like when a drum hits you reduce other stuff for more impact. Then quickly bring other stuff back up. It is a variation on gating. Now there is no doing something like this to get the sound you and/or the band want at a low level of compression and letting some overall broad compression do more to make it useful in the car, and not do it for home listening etc etc. Yes, you can do the broad single compression to get something, but the general gestalt of the result is not predictable for the end use of it.

Maybe you could ask the mastering/mixing people to provide crushed and uncrushed, but it is twice the work. Also some of the style of the sound cannot really be duplicated in both versions and the people making the music might reject it. So again, not happening.

 

[raindance]

Just don't compress it in the mastering and let the playback device add compression on playback.

 

[Blumlein 88]

Just don't compress it in the mastering and let the playback device add compression on playback.

Is no one listening? It is NOT THAT SIMPLE!

I'm not defending the loudness wars, but it does no good to offer pie in the sky solutions due to not knowing how compression is applied these days. Even the simplest one step compression you select threshold, attack and release times. That can alter the sound of one instrument in one track, and more so when more than one thing is going on. There is no universal choice for those values. Whether doing or undoing them. Then in cases where one chooses moderate compression for track A, and mixes it with only one other track B with heavy compression there is no way to differentially uncompress that on playback. Now multiply that by 8 tracks or 24 tracks and the complexity of undoing such a thing is way beyond doing.

 

[Cecill]

Compression? I can’t understand how Mr. Kiwanuka could listen back to his album 'Small Changes' and be satisfied with this level of compression. It’s so excessive that I’ve had to stop listening altogether.

 

[Blumlein 88]

Okay super simple example. I created a Risset drum sound in Audacity. I used the compression effect there which is about as simple as it gets. It is just a few seconds. You'll hear the Risset drum hit, then again with one compression, and a third using different compression. It repeats that way a few seconds. The two compression versions only differ in attack, release and ratio. If this were a real drum the differences would be much greater. Imagine for effect I alternated between these each drum beat. There is no single decompression that would fix it. And this is as simple as it can get.

Flac file attached in a zip download.

 

[GD Fan]

Okay super simple example. I created a Risset drum sound in Audacity. I used the compression effect there which is about as simple as it gets. It is just a few seconds. You'll hear the Risset drum hit, then again with one compression, and a third using different compression. It repeats that way a few seconds. The two compression versions only differ in attack, release and ratio. If this were a real drum the differences would be much greater. Imagine for effect I alternated between these each drum beat. There is no single decompression that would fix it. And this is as simple as it can get.

Flac file attached in a zip download.

I'm curious how different the compression settings were? It's evident but fairly subtle in the second sound but the third is quite significantly different from the original.

 

[Blumlein 88]

I'm curious how different the compression settings were? It's evident but fairly subtle in the second sound but the third is quite significantly different from the original.

I didn't write it down. The differences were large, but not extreme in any direction.

 

[Sacha]

When people with no clue about producing audio opine on how audio should be produced...

 

[tmtomh]

Thoughts in the head of an audiophile.

It is important to respect the musicians' artistic creation. Music production is a work of art.

Music today is digital. Why not initially mix the music to an optimal HIFI standard according to the musicians' wishes. The files are then converted using a special standard MASTER ALGORITHM that optimizes the sound for mobile phone listening. The algorithm must be so flexible with different publicly given input values that different productions get a different sound image on the phone. Can't be particularly complicated to produce. The end product provides optimal sound in mobiles.
In HIFI installations another standard ANTI-MASTER ALGORITHM is run which reconverts the digital music to optimal HIFI quality with the right dynamics, LUFS etc. The final product will probably be somewhat larger, which is probably not a problem today. Anyone who wants optimal sound must buy a new gadget - liked by the industry.

MASTER/ANTI-MASTER ALGORITHM gives the musicians the respect for their artistic work they deserve.
The problem is that the industry needs to agree on new standard algorithms for a small market - the audiophiles and the musicians. The cost in the mastering step should be negligible with a standard algorithm.

In summary - one music product for everyone. For the audio lover, the sound is optimized with additional gadgets/software.

Neuro

View attachment 409382

I totally get what you are saying and I think some folks here are being needlessly snide and dismissive in their responses.

I think @Blumlein 88 (who I do not include in my above remark) has the right answer for why it's not that simple and likely not feasible.

However, with that said, I think it's worth noting that a simple version of the technology you are talking about here was released almost 30 years ago: HDCD.

HDCD is a standard/protocol that includes a number of digital tools, but two of them are more or less precisely what you are describing: Low-level Range Extend and Peak Extend. Basically what both of these tools do is enable the mastering engineer to master in 20 bits, and then encode the extra 4 bits in the least-significant bit of the 16-bit CD standard.

So the result is a CD with, effectively, 15 bits' worth of dynamic range, but which in an HDCD-capable player can be decoded to have 20 bits' worth of dynamic range.

When an HDCD with Peak Extend is decoded that way, the overall volume level of the music is turned down 4dB, and the full peaks are restored, up to 4dB louder than the undecoded version. (Low-level Range Extend works similarly, just on the quiet end of the volume spectrum.)

HDCD died out for a few reasons. One is that it required proprietary hardware. The other is that there really wasn't much point in it, because you didn't need more than 16 bits to master highly dynamic music. It was more of a gimmick, and was partially marketed in exactly the way you are discussing: a special "high dynamic range" format that could "preserve" and reproduce music "better" than standard CD.

So in the end it sort of returned to Blumlein's point: no real demand and more complicated than it was worth.

 

[kemmler3D]

The problem is that the industry needs to agree on new standard algorithms for a small market - the audiophiles and the musicians.

Good luck with that!

The cost in the mastering step should be negligible with a standard algorithm.

As others have said, you're dramatically underestimating how complex the "mastering step" is in the studio.

You're talking about having a low- and high-dynamic range version of the recording where compression is selectively applied on playback for the low-dynamic range version.

While most DSP you encounter (EQ, FIR, etc) is very computationally light these days, the software they use for dynamic compression is pretty demanding, you need a modern PC to run it.

Check out iZotope Ozone for an example of a mainstream mixing / mastering dynamics suite. It would be in the running for a "standard algorithm" if there was one, but as you can see there are dozens of controls, and you might use one instance of this plugin per stem, so the CPU demands are material on complex recordings.

Your WiiM or phone or TV or DAP can't handle that.

It would be much, much simpler to just ship two masters of every recording, the high dynamics and low dynamics one, and let the user pick which one to listen to. Two audio files is infinitely simpler to deal with than an entire mix/mastering FX chain running on your phone or TV or whatever.

I understand the logic and appeal of "just do the compression on playback" but this is a bit like saying car companies should leave cars unpainted so the buyer can "just paint them at home"... possible in principle, seems simple enough if you haven't gone into the details, but not practical in reality.

 

[kchap]

It would be much, much simpler to just ship two masters of every recording, the high dynamics and low dynamics one, and let the user pick which one to listen to. Two audio files is infinitely simpler to deal with than an entire mix/mastering FX chain running on your phone or TV or whatever.

How about perceptual coding for mobile phone, e.g. mp3 and a lossless for HI-FI, e.g FLAC.

 

[EERecordist]

Thoughts in the head of an audiophile.

It is important to respect the musicians' artistic creation. Music production is a work of art.

Music today is digital. Why not initially mix the music to an optimal HIFI standard according to the musicians' wishes. The files are then converted using a special standard MASTER ALGORITHM that optimizes the sound for mobile phone listening. The algorithm must be so flexible with different publicly given input values that different productions get a different sound image on the phone. Can't be particularly complicated to produce. The end product provides optimal sound in mobiles.
In HIFI installations another standard ANTI-MASTER ALGORITHM is run which reconverts the digital music to optimal HIFI quality with the right dynamics, LUFS etc. The final product will probably be somewhat larger, which is probably not a problem today. Anyone who wants optimal sound must buy a new gadget - liked by the industry.

MASTER/ANTI-MASTER ALGORITHM gives the musicians the respect for their artistic work they deserve.
The problem is that the industry needs to agree on new standard algorithms for a small market - the audiophiles and the musicians. The cost in the mastering step should be negligible with a standard algorithm.

In summary - one music product for everyone. For the audio lover, the sound is optimized with additional gadgets/software.

Neuro

View attachment 409382

To try out your idea, set up a free DAW and find some expansion plugins. Most plugins have a 30 day free trial.

In the old days of magnetic tape, Dolby and DBX had pre-tape compression, and post tape expansion to reduce tape hiss. Dolby was level calibrated and DBX was, I believe, a straight 2:1 encode and 1:2 decode, level independent. You could probably find a used vintage DBX decoder.

Music listening use cases include concert halls, the recording/mastering studio, clubs, stadiums/arenas, home hi-fi, home medium/low-fi, outdoors on bluetooth speakers, and in cars. Each is going to have a noise floor and an acceptable playing volume. Cars are just about the worst situation, because of the noise floor. Over the air AM & FM broadcast had compressors between the control room and the transmitters. Vinyl mastering usually includes compression/limiting.

I haven't looked into it but I know Apple has mastering requirements, so other streaming services will have variations just for loudness normalization. They probably process the files using their own compression software before they go into the streaming library. You could look at their requirements and see if they are algorithmically reversible.

Mobile wired earbuds and closed back headphones are probably a good listening environment, second to concert halls and quiet homes.

The music may pass through Dolby or MP3 compression based on psychoacoustic masking. It would be hard to reverse that.

 

[AnalogSteph]

While most DSP you encounter (EQ, FIR, etc) is very computationally light these days, the software they use for dynamic compression is pretty demanding, you need a modern PC to run it.

The Rockbox compressor always ran fine on my old ClipV2/Clip+ without causing insane battery drain. So that's like what, a few dozen MHz worth of ARM9 core (if that)?

Obviously, it's just a basic compressor (hang-type AGC), not an oversampling brickwall limiter. It's generally gotten the job done for me though, particularly since the non-zero attack time was implemented. In an environment of compromised dynamic range you're generally happy if you can listen to your stuff half-decently, it doesn't have to be perfect. Clearly, if a decade-old tiny DAP can pull it off, it shouldn't be a challenge for a modern player or a car radio.

It would be much, much simpler to just ship two masters of every recording, the high dynamics and low dynamics one, and let the user pick which one to listen to.

Quite arguably so, yeah. That being said, this is likely to remain a niche. Shipping just one version for an uncompromised listening environment and letting the playback equipment take care of any compression is not the worst alternative. Do you really want to accommodate the restrictions of all of the many possible compromised listening environments individually at mastering? That'll be quite the list.

How about perceptual coding for mobile phone, e.g. mp3 and a lossless for HI-FI, e.g FLAC.

Anyone can transcode lossless to lossy at home, been doing that for many years.

 

[kemmler3D]

letting the playback equipment take care of any compression is not the worst alternative.

I would just argue it's not a realistic alternative. First of all, compression is typically applied per instrument (stem), AND on the overall track, so you'd actually need to ship stems, which I don't see happening for a variety of reasons. Or the industry just decides that compressed music will sound dramatically worse from now on? Doubt it...

Second of all, I don't see artists, mixers, or mastering engineers trusting their sound to a low-CPU simple compressor out there in the wild.

Do you really want to accommodate the restrictions of all of the many possible compromised listening environments individually at mastering? That'll be quite the list.

Similar to above, nobody wants to do that...

 

[AnalogSteph]

I would just argue it's not a realistic alternative. First of all, compression is typically applied per instrument (stem), AND on the overall track, so you'd actually need to ship stems, which I don't see happening for a variety of reasons.

Going that far would be silly, that's true.

No, just ship something that sounds good in a typical (largely) uncompromised home listening environment, in whatever way that particular sausage may be made. I'd be fine with 1994 era mastering levels. Then have the resident DSP sort things out in compromised environments. In environments like a car the sound system should know best what FR tweaks etc. to apply depending on ambient noise level anyway.

 

[kemmler3D]

Then have the resident DSP sort things out in compromised environments. In environments like a car the sound system should know best what FR tweaks etc. to apply depending on ambient noise level anyway.

Loudness compensation based on ambient noise would be one thing, basic compression based on ambient noise might be OK, but realistically there is no one-size-fits-all compression setting that actually works in real life.

I have direct experience with this - at my old job I had to fight with the engineering team over global dynamic compression settings in an onboard DSP, as they were trying to limit current to avoid stressing the battery in a portable speaker. Not only did most of the settings sound worse, many of them didn't limit RMS or even peak current very well. It was quite a hassle. I believe the quest for a universally acceptable dynamic compression scheme is mostly futile, better to leave it in the hands of the studio folks.