Is lossy outdated in 2019 & onwards?

[Julf]

Only the track you made quieter, the other track, say the drums is still at loud at it was, so the range increases.

The noise floor of the drum track is still the same. You can turn down the other tracks as much as you want, the noise floor is still determined by the loudest track.

 

[Dj7675]

Lossy not outdated for me. I am limited to a 3M DSL connection at our home. I was on the Tidal Hifi plan but could not reliably play lossless. I then took an online test dbt test of lossless vs 320k aac and could not tell the difference. Switched to tidal premium which does 320k aac and have been happy. I think the $9.99/mo for tidal premium @ 320k aac is a really great deal and works great with roon.

 

[solderdude]

Not yet, as there still are a lot more people listening to mp3 than to vinyl.

Yes. The source code for several of the decoders are open source.

They were rhetorical questions to me.
I do suspect vinyl will outlast MP3 though.
There may well be technically superior lossy formats (certainly at very low bitrates) but don't think they'll take over MP3.
It's too wide spread for this.

 

[MRC01]

My Pixel 3 phone doesn't have enough storage for my entire FLAC collection. So when I'm on the go, Opus it is. ...

Same here. On my phone and earbuds, or Bluetooth to the car, I'm not going to hear the difference between the original and MP3 VBR averaging around 190 kbps. And it lets me store 3-4 times as much music in the same space as the FLAC of a CD.

My master library is lossless (FLAC), but I still use lossy (MP3 VBR) for the phone. Lossless compression will eventually be pointless, but due to storage limitations of phones & bandwidth limitations of Bluetooth, we're still a few years away from that day.

 

[Willem]

CD is clearly better than FM, but with DAB+ this is not so obvious since at least in the Netherlands DAB+ is primarily implemented as an in car format at AAC+ at 96 kbs. Conversely 320 kbs is the lossy streaming format adopted by some services such as Spotify, and it is generally accepted that this is so close to CD that it can only distinguished by very careful listening. In short, this will be better than FM. However, at lower bitrates it is a balance between one kind of imperfections against another kind of imperfections. Dutch public radio stations have a bitrate of mostly 192 kbs, which is not as good as BBC Radio 3, but in my subjective judgement nevertheless better than FM. ANd it is important to note that this bitrate is about double of what it was only a few years ago. We are edging towards perfection.

 

[solderdude]

Do they still use compressors like optimod for radio broadcasts even when digital ?
I know it was used a lot for AM radios but could clearly hear it 'pump' on the Dutch FM radiostations as well.

Don't listen to radio any more but the few times I have it on in the car I find the audio quality appalling.

 

[MRC01]

I have an HD radio receiver that lets me toggle digital. That is, with stations that broadcast both I can listen in analog or digital. I find the quality of digital varies quite a bit. Some stations's digital sounds better than analog, some worse. The stations that sound worse are typically using extreme compression & EQ -- which they do on their analog broadcasts too -- like pop/rock.

Yet "better" / "worse" is a subjective evaluation. I'll bet for lots of people, the stations that I think sound worse in digital, sound better to them. At least they don't have any static, which is the only aspect of sound quality that a lot of people notice.

 

[Bounce44.1]

I don't have SACDs or DVD-A source material... but if I did, then sure. Can't do better than the source.

Depending on your system you can use "pitch and catch". Pitch thru the DAC at original sample rate. Catch thru the ADC at the preferred sample rate. Will not be equal but will be better.

 

[Bounce44.1]

Have you found any commercial mainstream recordings that actually use more than 16 bits of dynamic range?

All commercial recordings are mastered at 24 bit. Mostly at 192khz. Part of the mastering process is to downsample the master to what formats the client requires. Each different format in manipulated specifically for that format. There are more and more sites on the interweb which offer HiRes streaming and downloads. How popular this becomes hinges on the big boys who own the masters and what is in their best interests.

 

[Julf]

All commercial recordings are mastered at 24 bit. Mostly at 192khz.

Mastered, yes. usually recorded at 24/44.1 or 24/48 (you don't want to waste processing power and bandwidth when recording 24 or 48 channels). 24 bits is used to give headroom and temporary precision when processing. That doesn't change the fact that the actual recorded audio doesn't have more than 80 dB or so (on a good day) of dynamic range (equivalent to 14-15 bits).

 

[MRC01]

Almost all music uses less than 96 dB of dynamic range. They don't record in 24 bit because they need more dynamic range, but because they don't know exactly where to set the level. They record at 24 bit because this lets them set levels conservatively low to avoid clipping (in case the musicians crescendo louder than they expected based on prior level checks), without losing resolution. After the entire thing is recorded they can shift all samples (no compression, just an amplitude shift) so the peak amplitude is just below zero dB. This shift if properly done is lossless (still 24-bit). Next they can apply any processing they want, in 24-bit for better precision. Finally, they can convert to 16-bit and if the music has < 96 dB of dynamic range this too will be transparent.
That said, with noise-shaped dither, 16-bit can give more than 96 dB of effective dynamic range.

 

[garbulky]

Your subjective personal opinion is noted.

Out of curiosity do you store your files in a lossy format? If so which one? Do you find it adequate for your use?

 

[Bounce44.1]

Almost all music uses less than 96 dB of dynamic range. They don't record in 24 bit because they need more dynamic range, but because they don't know exactly where to set the level. They record at 24 bit because this lets them set levels conservatively low to avoid clipping (in case the musicians crescendo louder than they expected based on prior level checks), without losing resolution. After the entire thing is recorded they can shift all samples (no compression, just an amplitude shift) so the peak amplitude is just below zero dB. This shift if properly done is lossless (still 24-bit). Next they can apply any processing they want, in 24-bit for better precision. Finally, they can convert to 16-bit and if the music has < 96 dB of dynamic range this too will be transparent.
That said, with noise-shaped dither, 16-bit can give more than 96 dB of effective dynamic range.

Very well said. In the analog world pushing the amplitude or level can create a musical distortion and/or clipping. There is no positive outcome in the digital world to push the level only negative results.

The digital world is all math. Everything done to the original wave file has an affect on the signal due to quantization or how the math is rounded. Even a simple level change has some effect. Bit depth minimizes this quantization effect. Most daws process at 32 bit depth with some plugins process at 64 bit. Lots of info on the interweb about this from people who know way more than I.

I am a little confused about the term "dynamic range". When I mix I watch my Dynamic Range Meter pretty close as this shows how much the signal is being compressed or squashed. Depending on the genre I see somewhere between 6db and 12db. Hopefully I am not showing my ignorance

 

[BDWoody]

Almost all music uses less than 96 dB of dynamic range. They don't record in 24 bit because they need more dynamic range, but because they don't know exactly where to set the level. They record at 24 bit because this lets them set levels conservatively low to avoid clipping (in case the musicians crescendo louder than they expected based on prior level checks), without losing resolution. After the entire thing is recorded they can shift all samples (no compression, just an amplitude shift) so the peak amplitude is just below zero dB. This shift if properly done is lossless (still 24-bit). Next they can apply any processing they want, in 24-bit for better precision. Finally, they can convert to 16-bit and if the music has < 96 dB of dynamic range this too will be transparent.
That said, with noise-shaped dither, 16-bit can give more than 96 dB of effective dynamic range.

I would think it becomes pretty hard to listen to if the total DR in the final mix is much above 25dB or so...

Any recording engineers out there? Where do you try to limit it?

 

[MRC01]

... I am a little confused about the term "dynamic range". When I mix I watch my Dynamic Range Meter pretty close as this shows how much the signal is being compressed or squashed. Depending on the genre I see somewhere between 6db and 12db. Hopefully I am not showing my ignorance

Sadly, a lot of recorded music doesn't use more than 12 dB of dynamic range. They apply extreme dynamic compression to make it sound as loud as possible, meanwhile squeezing the life out of the music. Modern Rock & Pop are the worst culprits here.

I would think it becomes pretty hard to listen to if the total DR in the final mix is much above 25dB or so...

That depends on how you measure dynamic range. I have some acoustic music recordings that use 60+ dB of dynamic range. That is, in a few very quiet PPP parts with music playing the digital VU meter is measuring -60 dB. In dynamic peaks FFF it hits 0 dB, or gets within 1 dB of it. So you could say the track uses > 60 dB of dynamic range. These tracks are impossible to listen to in a car, but are wonderfully natural and realistic sounding in a good quiet listening room.

However, the DR14 tool might rate this same track as DR15. That's a different measurement, essentially the ratio of peak to RMS amplitude. It's possible for the RMS value of amplitude to be high, say -9, even when the amplitude hits minimums of -60 dB or lower in a few places.

 

[BDWoody]

Sadly, a lot of recorded music doesn't use more than 12 dB of dynamic range. They apply extreme dynamic compression to make it sound as loud as possible, meanwhile squeezing the life out of the music. Modern Rock & Pop are the worst culprits here.


That depends on how you measure dynamic range. I have some acoustic music recordings that use 60+ dB of dynamic range. That is, in a few very quiet PPP parts with music playing the digital VU meter is measuring -60 dB. In dynamic peaks FFF it hits 0 dB, or gets within 1 dB of it. So you could say the track uses > 60 dB of dynamic range. These tracks are impossible to listen to in a car, but are wonderfully natural and realistic sounding in a good quiet listening room.

However, the DR14 tool might rate this same track as DR15. That's a different measurement, essentially the ratio of peak to RMS amplitude. It's possible for the RMS value of amplitude to be high, say -9, even when the amplitude hits minimums of -60 dB or lower in a few places.

Thank you for explaining. Is there a tool or site that shows the dB range as you describe in terms of actual min/max amplitude rather than the ratio to peak RMS?

 

[MRC01]

It would be nice to have a dynamic range measure that tells the actual min & max, but I've not seen one. I speculate why: It's hard to measure the minimum DB used (quietest part) because even sections of pure quiet won't be digital zero, since the recording studio has ambient noise (no room is perfectly quiet) and digital recordings use dither. The dither is often frequency shaped using different curves, so taking a spectrum and looking for equal energy won't be sufficient. Plus we humans can detect correlated signals (music) even below the noise floor. In short, it's hard (not impossible, but not simple) for computers or math to decide whether the quietest parts contain music.

I think this is why most dynamic range computations avoid this thorny issue by using peak to RMS values. And of course that has its limitations: you don't measure what the actual dynamic range of the recorded music really is. DR14 numbers are a useful guide but they don't tell the whole story and can be misleading for certain types of music.

 

[Blumlein 88]

It would be nice to have a dynamic range measure that tells the actual min & max, but I've not seen one. I speculate why: It's hard to measure the minimum DB used (quietest part) because even sections of pure quiet won't be digital zero, since the recording studio has ambient noise (no room is perfectly quiet) and digital recordings use dither. The dither is often frequency shaped using different curves, so taking a spectrum and looking for equal energy won't be sufficient. Plus we humans can detect correlated signals (music) even below the noise floor. In short, it's hard (not impossible, but not simple) for computers or math to decide whether the quietest parts contain music.

I think this is why most dynamic range computations avoid this thorny issue by using peak to RMS values. And of course that has its limitations: you don't measure what the actual dynamic range of the recorded music really is. DR14 numbers are a useful guide but they don't tell the whole story and can be misleading for certain types of music.

I don't really know how that could work. Firstly I've not found any recordings with silent portions allowing more than 80 db. I exclude purely electronic studio creations.

 

[Bounce44.1]

There are many dynamic range meters which work within a DAW. A rule of thumb is that every increase of 10db doubles the percieved loudness. If the dynamic range of a song is say 30db this results in a range of percieved loudness of 6 times from the quietest to the loudest. This doesn't include the peak or attack portion of the sound. These peaks live in the headroom realm.

 

[MRC01]

Yeah, I find good live acoustic music recordings typically have a noise floor around -60 to -80 dB; that's what digital VU meters read when playing the quiet parts between movements which is just recording the ambient room noise.

Some of these performances have actual music with PPP levels that get down to the ambient room noise level. Others don't. It would be nice to have a measurement that indicates the minimum level of actual music (call it something like "max dynamic range actually used") but it's hard to devise a way to measure that.

To me, incidental sounds like the musicians breathing or clothes rustling as they move is part of the performance, captured by the best recordings and enhances the realism of the experience.