General question on dynamic range and music

[Blumlein 88]

One of the entries in the database with the highest DR is a nature recording by Dallas Simpson - "The Shore of Stones Suite". It sounds phantastic and I am still not convinced (depending on the genre that is) why even higher DR shouldn't be beneficial. Even if the difference would be almost inaudible, 24-bit/192kHz audio is as well almost inaudibly different from 16-bit/44.1kHz and there is still a market for it.

I guess I would like to.

I do a few amateur recordings. When I first did them of some friends with a little band I was happy with my results. I did no compression, just a pair of good microphones in a purist recording style. Sounded good to me late at night on my home system. When I distributed to the people in the group what did they say? Well I cannot hear it good in the car, I wanted to listen to ourselves with the CD in my car. So on and so forth. Even in quieter places than driving along in an automobile everyone complained you had to turn it up too much to hear it and then sometimes it was way too loud. So yes for many purposes you can have too much in the way of variable dynamic levels. You can also compress the music too much and it sounds worse. There is a sweet spot of sorts. And this really is crest factor we are talking about here.

 

[Blumlein 88]

Could you please go into more detail what exactly is flawed on these measurements? Is it the approach in itself ? Is there no way another, better suited approach to be taken?

LUFS is a better way though it too has its own issues.

 

[Jochen]

I've said I slightly simplified what TT meter does. What it should be called more accurately is a crest factor. Short term peaks versus longer term levels. If you will start thinking about and referring to crest factor (TT meter DR values) and dynamic range as two separate ideas which they are it will help clarify things.

Read this article about crest factor, think about it, read it a few times, and it will clear things up. Currently LUFS is a standard for such things. Prior to that DR levels in TT meter sort of provided a similar kind of info (though with more flaws than LUFS) for recordings. There have been other ways to get an insight into similar info. Graphically Audacity shows that if you know how to read it. Crest factor in recordings is related to compression levels and making music listenable in different environments. It is in no way directly related to dynamic range as technically defined. You need to separate those in your mind or it will never make sense.

What Is Crest Factor and Why Is It Important?

Find out what "crest factor" means in the world of audio mastering, how to interpret different readings, and how to use them to inform mastering decisions.

www.izotope.com

Thanks for the link, will read it thoroughly. I just want to remark that not I originally called these measurements "dynamic range", but the database itself. But while the there reported values may only have a loose connection to actual dynamic range (if any), the ordering should have some validity. And furthermore I encountered a lot of good music, hitherto unknown to me, by going through the ordered list.

 

[solderdude]

DR (as in the database) has nothing to do with dynamic range. It is the naming that is the problem here. Not the measurement method.

You are basically thinking the DR in the database is the same thing as the technical term for the available range for signals.
For vinyl it basically means the system noise floor vs the max available amplitude as can be cut in the vinyl.
A 96dB range merely means the softest recorded signal amplitude is 96dB lower in level than the max. amplitude (without clipping).
For 24bit the theoretical dynamic range (softest signal, without dithering and ignoring practical noise levels) is high. Much higher than ever will be possible in practice because it will be limited by practical noise levels.

Dynamic range of the hearing (listening to music) is somewhere around 60-70dB depending on many factors. That would be the difference between peaks and your hearing threshold shortly after the peaks and average levels.
This despite the dynamic range of the hearing being much larger. When one is acclimatized in an acoustically dead room and can hear the blood flow you can actually hear extremely soft sounds below 0dB SPL at frequencies around 2-3kHz (0dB SPL is just a certain SPL not NO sound at all).
You can also endure short peaks of 130dB for instance pistol shot for instance nearby but is not a fun thing. Just like the acoustically dead room is not a fun thing either.
So while the hearing can hear a 'dynamic range' of over 130dB you cannot hear the faint sounds directly after being exposed loud sounds for a short while.

Your hearing threshold is actually a dynamic thing. In the morning and late in the evening your bottom hearing threshold is much lower (can hear faint sounds) than when having been around all-day noises. You can't hear faint noises during the day (and certainly not after visiting a concert or disco) that you can late in the evening after an evening of silence or soft background music.

DR in the database is NOT based on the softest and loudest parts at all but are indicative for found peaks and 'average' levels that are not really average but really peaks that are 'smoothed' over a long period (the song or even album) and can indicative for 'loudness wars' (peaks reduced so the average levels are relatively higher).
It is not about the softest recorded sound that would sit just above the noise level or set gating level vs the loudest peaks.
If dynamic range compression is done extremely well a recording with just DR5 can sound great and low in distortion where DR20 recordings that are clipped might not sound nice. In general the higher DR recordings may sound closer to what one would endure listening to (not sound enforced) live music or live music with really great sound enforcement where no limiters or compressors are used (in order to comply to rules regarding max sound peak exposures)

Totally different things using the same name.

 

[Newman]

Could you please go into more detail what exactly is flawed on these measurements? Is it the approach in itself ? Is there no way another, better suited approach to be taken?

https://www.audiosciencereview.com/.../41843787/&q=Ttdr&c[users]=Newman&o=relevance

Just read the threads around these links.

I see you are making comments about using the database to pick music with the high scores. That’s a really weird way to decide on what music to listen to, because the music itself is a major contributor to the presence and distribution of SPL highs and lows. Hence you are excluding whole swathes of music that aren’t intrinsically going to produce high scores on that tool.

 

[Jochen]

I see you are making comments about using the database to pick music with the high scores. That’s a really weird way to decide on what music to listen to, because the music itself is a major contributor to the presence and distribution of SPL highs and lows. Hence you are excluding whole swathes of music that aren’t intrinsically going to produce high scores on that tool.

It may be a weird way, but as I said, I discovered a bunch of excellent music, often fringe one, I didn't know before. Surely I have other ways too, magazines, indications from friends, proposals from streaming services, whatever, but this is just a complementary method. And it seems that generally I prefer music with high crest factor in comparison to a low one.

 

[Jochen]

DR (as in the database) has nothing to do with dynamic range. It is the naming that is the problem here. Not the measurement method.

You are basically thinking the DR in the database is the same thing as the technical term for the available range for signals.
For vinyl it basically means the system noise floor vs the max available amplitude as can be cut in the vinyl.
A 96dB range merely means the softest recorded signal amplitude is 96dB lower in level than the max. amplitude (without clipping).
For 24bit the theoretical dynamic range (softest signal, without dithering and ignoring practical noise levels) is high. Much higher than ever will be possible in practice because it will be limited by practical noise levels.

Dynamic range of the hearing (listening to music) is somewhere around 60-70dB depending on many factors. That would be the difference between peaks and your hearing threshold shortly after the peaks and average levels.
This despite the dynamic range of the hearing being much larger. When one is acclimatized in an acoustically dead room and can hear the blood flow you can actually hear extremely soft sounds below 0dB SPL at frequencies around 2-3kHz (0dB SPL is just a certain SPL not NO sound at all).
You can also endure short peaks of 130dB for instance pistol shot for instance nearby but is not a fun thing. Just like the acoustically dead room is not a fun thing either.
So while the hearing can hear a 'dynamic range' of over 130dB you cannot hear the faint sounds directly after being exposed loud sounds for a short while.

Your hearing threshold is actually a dynamic thing. In the morning and late in the evening your bottom hearing threshold is much lower (can hear faint sounds) than when having been around all-day noises. You can't hear faint noises during the day (and certainly not after visiting a concert or disco) that you can late in the evening after an evening of silence or soft background music.

DR in the database is NOT based on the softest and loudest parts at all but are indicative for found peaks and 'average' levels that are not really average but really peaks that are 'smoothed' over a long period (the song or even album) and can indicative for 'loudness wars' (peaks reduced so the average levels are relatively higher).
It is not about the softest recorded sound that would sit just above the noise level or set gating level vs the loudest peaks.
If dynamic range compression is done extremely well a recording with just DR5 can sound great and low in distortion where DR20 recordings that are clipped might not sound nice. In general the higher DR recordings may sound closer to what one would endure listening to (not sound enforced) live music or live music with really great sound enforcement where no limiters or compressors are used (in order to comply to rules regarding max sound peak exposures)

Totally different things using the same name.

Yeah, I got it now. Thanks for the basic lecture in psychoacoustics, that wasn't necessary. But not only this database is confusing also many articles like the already cited one: https://hub.yamaha.com/audio/music/what-is-dynamic-range-and-why-does-it-matter/ and even researchers (from wikipedia):

"In 2007, Suhas Sreedhar published an article about the loudness war in the engineering magazine IEEE Spectrum. Sreedhar said that the greater possible dynamic range of CDs was being set aside in favor of maximizing loudness using digital technology. Sreedhar said that the over-compressed modern music was fatiguing, that it did not allow the music to "breathe"."

"In September 2011, Emmanuel Deruty wrote in Sound on Sound, a recording industry magazine, that the loudness war has not led to a decrease in dynamic variability in modern music, possibly because the original digitally recorded source material of modern recordings is more dynamic than analogue material. Deruty and Tardieu analyzed the loudness range (LRA) over a 45-year span of recordings and observed that the crest factor of recorded music diminished significantly between 1985 and 2010, but the LRA remained relatively constant. Deruty and Damien Tardieu criticized Sreedhar's methods in an AES paper, saying that Sreedhar had confused crest factor (peak to RMS) with dynamics in the musical sense (pianissimo to fortissimo)."

 

[solderdude]

Yep, as soon as a term is used that can have several meanings things can get confusing.
In this case the dynamic range is the difference between loudest and a certain level where the 'certain level' part is basically the thing that differs.

 

[Jochen]

I do a few amateur recordings. When I first did them of some friends with a little band I was happy with my results. I did no compression, just a pair of good microphones in a purist recording style. Sounded good to me late at night on my home system. When I distributed to the people in the group what did they say? Well I cannot hear it good in the car, I wanted to listen to ourselves with the CD in my car. So on and so forth. Even in quieter places than driving along in an automobile everyone complained you had to turn it up too much to hear it and then sometimes it was way too loud. So yes for many purposes you can have too much in the way of variable dynamic levels. You can also compress the music too much and it sounds worse. There is a sweet spot of sorts. And this really is crest factor we are talking about here.

Maybe there are recordings that are better listened to at home, not in the car.

 

[Jochen]

I still see comments mixing up concepts like DR rating or LUFS and dynamic range. Two completely different ideas.

To reiterate: one is about average levels vs peak levels. The other is about total dynamic range from quietest to peak levels.

Dynamic range vs needs of ours ears vs needs of a recording/playback medium can be a bit complex. I'm leery of introducing that for fear it will further cloud things rather than clear them up.

Thanks for the elucidation, now it is pretty clear that these are really two separate things.

As an interesting side note, in this loudness war database, which calls itself erroneously "Dynamic Range DB", the second highest entry is the Album "Hama" by Shurayuki-hime. The style of the music is called "harsh noise wall", and this is exactly how it sounds. Mathematically, the crest factor of Gaussian noise is infinite, so making "music" with a high crest factor is easy.

 

[Newman]

...and uninteresting. You make music for musical communication: its dynamic range, crest factor et al simply is what it is.

 

[Adi777]

Others - the loudness war has killed the quality of the music.
Most ASR - DR is not that important.
WTF???

 

[solderdude]

I don't think anyone said DR is not that important.
Poorly applied compression and limiting is.

 

[Sokel]

Think of it practically.
One of the most used song in threads like that are DR's "Private Investigations" exactly because it shows up in the TT site having a DR 19 rating.
And all it has is 3-4 loud bursts.Ok.those bursts sound impressing first time someone listens to it if you crank the volume a little but that's about it,nothing special.

The way I see it is to make sure you have a nice quiet environment (as much as you can,I struggle to 36-38db normally as you can see in REW's measurement) and from then on to have the ability (speakers,power,etc) to hit a reasonable target reserving some for headroom too.

It would be a shame to adjust the music for the system,the goal is the exact opposite.

 

[goat76]

I've said I slightly simplified what TT meter does. What it should be called more accurately is a crest factor. Short term peaks versus longer term levels. If you will start thinking about and referring to crest factor (TT meter DR values) and dynamic range as two separate ideas which they are it will help clarify things.

Read this article about crest factor, think about it, read it a few times, and it will clear things up. Currently LUFS is a standard for such things. Prior to that DR levels in TT meter sort of provided a similar kind of info (though with more flaws than LUFS) for recordings. There have been other ways to get an insight into similar info. Graphically Audacity shows that if you know how to read it. Crest factor in recordings is related to compression levels and making music listenable in different environments. It is in no way directly related to dynamic range as technically defined. You need to separate those in your mind or it will never make sense.

What Is Crest Factor and Why Is It Important?

Find out what "crest factor" means in the world of audio mastering, how to interpret different readings, and how to use them to inform mastering decisions.

www.izotope.com

I'm not sure what the DR numbers are that we see at dr.loudness-war.info, but it's neither the Crest factor nor the Integrated LUFS. So if it's not the dynamic range, what is it?


PJ Harvey - Rid OF Me

 

[little-endian]

I have a general question on dynamic range (DR). Redbook CD offers 16 bit, corresponding to 96dB DR.

According to a more accurate formula, it is even about ~98dB, and I think strictly speaking, that is referring to the SNR, rather than the dynamic range. Although in linear quantisation systems, both refer to more or less the same, in non-linear ones, the dynamic range can be a lot higher if one accepts a varying noise-floor, like the G.711 codec does when mapping 12 bit linear to 8 bit of non-linear PCM.

With dither up to 120dB. High-Res 24 bit audio even 144dB.

Yes, although the term "high resolution" actually already is a stupid marketing term. Being higher "resolved" of what? The visual counterpart of resolution would rather go along with sampling in the case of audio, so higher numbers, given, allow for more bandwidth although any human would require a way better set of ears to make use of it. When it comes to quantisation, nothing is finer "resolved" regardless of the bit-depth, but only the SNR is higher, given proper dither. No stairsteps, they aren't there.

Vinyl, according to different sources up to 70dB and tape machines roughly 65dB.

I'd want to see that on any vinyl, but let's believe that in good optimism. I'd rather opt for tapes or FM-modulated analog audio like the LaserDisc provides it, if it really has to be analog for the ones who don't trust discreet character sets.

So my question would be, why is it not possible (or desirable?) to record the full dynamic range of live music and why is the capacity of the standard CD not taken advantage of?

The sad truth probably is that apparently less and less individuals really want a lot of dynamic range. Even so-called "audiophile" recordings, while often well-produced without a doubt, are rather mediocre in terms of dynamic range. Same partly with movies, where rattling, heavy bass and noise all the time is often confused with high dynamic range, although the opposite is true.

Is the 96dB - 25dB = 71dB difference of capacity minus actual content on CD actually wasted? Not talking about 3dB loudness war abominations or even High-Res. Or am I mixing things up?

The diagnose of being "waste" depends on the desired result, I'd say. Interestingly, it occurs on a technical level as the amount of data is somewhat wasted though if the dynamic range isn't used by the content. Often, 12 bit / sample propably would suffice as well and with nowadays, mostly horribly produced pop music, even 8 bit (properly dithered and noise shaped) would be casting pearls before swine, not to speak of 16, 24 being the definition of absurdity.

Same is true for the sample rate, by the way. If the content never uses frequencies above 16 kHz for example (assuming to start from 0 Hz), then 32 kHz sampling rate would do it as well, taking details such as anti-aliasing filtering aside (which would have to be very steep then, etc.).

If it is a 16bit recording, this is all that's left:

View attachment 283553

Beware of trying to make any point by using Audacity's really lousy rendering of waveforms as this is not what one will get to hear on any DAC's output. Audacity shows the usual stair-steps without any visual reconstruction which happens in reality. Editors like Adobe Audition (formerly "Cool Edit (Pro)") do this way better and show the mathematical reconstruction of the time- and value-continous waveform.

DR (as in the database) has nothing to do with dynamic range.

While I agree with your posting in general, I disagree with the "nothing", especially given such an emphasis as of course, the DR values in the database do have something to do with dynamic range or at least a sub-set of it depending on the definition (or rather the 'scope'). According to some explanation from the DR TT Meter back in the days, this tool deliberately only takes the loudest parts of tracks and effectively measures the average level to peak ratio in order to prevent giving music with quiet parts somewhere else higher results than they "deserve". Sure, the total (macro) dynamic range is the entire span from quietest to loudest, but the DR meter rather takes the micro dynamics approach.

That is the point, redbook CD has about 30dB more DR then vinyl, but at least in that database the DR of the same recording in both formats is basically the same. Why not make use of the additional capacity?

I might be wrong about this point, but I assume that this is also due to the DR meter telling only some part of the dynamic range, here the average to peak relation. Also, most music neither uses the entire range of "so so vinyl" nor that of 16 bit PCM by far of course. Because of vinyl's (well, at least when used in an analog way which is virtually the only established application) higher noise floor, the macro SNR / dynamic range of a given musical piece will be lower than the PCM's counterpart but since the focus is rather on the average to peak - ratio for that measurement, it doesn't change too much.

You may take a 16 bit PCM file, measure the DR value, convert it to a dithered 8 bit one and measure again.


In general, today we have the "screaming irony" that we surpassed the ability to reproduce sound entirely flawlessly since the introduction of the CD with 44.1 kHz / 16 Bit since 40 years now, being enough forever (at least until evolution gives us better ears and with no frustrating aging, please) but instead of using it, we constantly keep barking up the wrong trees* with "high res audio" here, "lossless audio" there, the last dB of SNR even further below the threshold of hearing, considering lossy formats as inferior (god forbid, a streaming service "only" uses AAC or movies nowadays only provide AC3 or DTS) but then praising the worst one of all - scratchy vinyl - which proves pretty well, how bad quality can actually be without necessarily neglecting the enjoyment of music or entirely masking a good production.
Wrong world and a pity for the CD as its marketing claim "perfect sound forever" was closer to the truth than many realise, especially today where anything other than 2.x MHz DSD or 192 kHz / 32 Bit PCM floating point is pure blasphemy.

* for the ones who read German, I can only highly recommend pelmazo's blog where he especially disassembled the usual marketing bullshit and misinformation where "data reduction" and "compression" are thrown together and the oh so evil lossy codecs of course are the root cause of our quality misery. Neither, he is too fond of all the voodoo non-sense which is going on and which drowns real information beyong recognition. Also his other posts are extremely enriching.

 

[danadam]

Where do you see in this database that this definition is used for the measurements?

I'm not sure what the DR numbers are that we see at dr.loudness-war.info, but it's neither the Crest factor nor the Integrated LUFS. So if it's not the dynamic range, what is it?

https://web.archive.org/web/2018091....de/sites/default/files/Measuring DR ENv3.pdf

 

[Jochen]

https://web.archive.org/web/2018091....de/sites/default/files/Measuring DR ENv3.pdf

So it is a slighly modified version of the crest factor.

 

[Cbdb2]

Another problem with these measurements, the DR of the music can change multiple times in one song, intro to verse to chorus for example, while these measurement give one number for the whole piece, sometimes an hour long.

 

[Jochen]

It's this song:

How to measure output Power of an headphone amp with REW

A 30dB difference between average and peak is a bit high unless it is a song that builds up in volume. Or has high DR https://www.audiosciencereview.com/forum/index.php?threads/measuring-a-dac-or-amps-dynamic-properties%E2%80%A6.41936/post-1487975 (with the ones that built up I have seen higher)

www.audiosciencereview.com

Is there a simple toole to measure the real dynamic range (as opposed to the crest factor) of an audio file? What toole is this measured with (I assume REW)? I have seen mentioning that there is a plugin for audacity, but could not find it.