Dynamic range compression of classical music.

Discussion in 'Sound Science' started by xenophon, Nov 24, 2014.
  1. Xenophon
    Perhaps a stupid question but I'm trying to get my head around dynamic range compression of classical music CD's:
     
    If JRiver Media player can be trusted most classical tracks seem to have been mastered with a dynamic range of about 15-20 dB.  Compared to other music that's certainly very decent.  Yet I wonder:  it seems to me -totally subjective as not only did I not measure this but obviously the listening environment is totally different- that when enjoying a live classical concert, the range is a lot larger.  
     
    Is there a set of 'rules' or a validated theory that determine how much compression is optimal for recordings and if so, what are these based on?  And if not, apart from potential volume management problems, why not use, say, 40 dB dynamic range?  
     
    Apples and oranges, I know, but I'd be grateful for any insights.
     
  2. C.C.S.
     
    20 dB is a very large difference. Unless I'm mistaken, 20 dB of dynamic range indicates that the peaks of the track (assuming this is how dynamic range is calculated) are 10x as loud as the quieter sections. It is hard to describe with words how much louder 10 times as loud sounds, but if you listen to most well-recorded and well-mastered classical tracks, you will probably come to the conclusion that this dynamic range is quite adequate for the music. And I'm sure that there are a number of classical and orchestral recordings which make use of more than just 20 dB of dynamic range.
     
  3. esldude
    Compression usually is expressed as a ratio.  2 to 1 compression means a signal that did have 60 db of dynamic range will then have 30 decibel of dynamic range.  Specifying what the dynamic range of music is for a given song is something that usually is fuzzy. 
     
    But how much compression is too much?  Well I have some Wilson recordings that were the mics direct to digital with no limiting, compression or EQ.  They do sound like great recordings and have pretty high dynamic range.  If I listen to them in the car, I simply cannot enjoy them.  Half or more of the music is either gone or so buried in noise you strain to hear it.  Doing some compression of the files myself I settled on 2.5 to 1.  More doesn't really improve things, and I can hear enough of it to enjoy it (though not as much as listening in a quiet room at home).  My vehicle isn't one of the quieter ones so you might get away with 2 to 1 in a quiet car or might still be fine with 2.5 to 1.  I would estimate the effective dynamic range of most of the Wilson recordings to be around 60 db.  So the effective result after my 2.5:1 compression is about 24 db.  On top of 70 db of highway noise in the car it works alright. 
     
    I would expect any good classical recording to at least have 25-30 db of range.  I really think for listening at home or over headphones in a quiet location the better recordings will have more than this.  Over the radio or other source in a car you can't really use more than that. 
     
    I also think most players should have selectable compression.  Maybe 2, 3, 4 or 6 to 1 range. That way you don't need compression as much in the recording, and the listener can choose the appropriate amount for the conditions while listening.  I am hardly the first person to think of this, and it appears not at all likely to happen. 
     
  4. RRod
     
    A dynamic range of 20, as given by the "dynamic range meter" that is used these days, is pretty much an upper limit. I have a few random tracks that manage 21-22 d.r., since they have exactly the right characteristics (a couple of high peaks and no high RMS sections that last > 3 seconds). If we go off dynamic range as measured instead by, say, the difference in RMS between the loudest and softest sections with musical content, I have some tracks in the 50-55 range (e.g. Zander's Mahler 9).
     
    I'm not sure which outcome you're looking at. If it's the dynamic ranger meter, have a look at this:
    http://www.dynamicrange.de/sites/default/files/Measuring%20DR%20ENv3.pdf
     
  5. C.C.S.
     
    Oh, wow. I did have huge misunderstandings, then, about how DR meter performed its calculations. I assumed that it did involve comparing the first peak against the softest section in which music was playing rather than using an RMS average for the highest 20% and the second peak value. While I have a vague understanding of their reasoning for using this methodology, I prefer to think of dynamic range as being measured between the absolute highest and lowest value over the course of the song. I also agree with esldude, above, that user selectable compression would be preferable, but I understand that this is very unlikely to ever become the standard.
     
  6. RRod
    I think most people would agree with you on the definition of dynamic range: a song that goes from ppp to fff has a high range; a song that stays at ff the whole time doesn't. The DR meter measurement focuses on catching compression. So if I test 2 versions of an album, one with compression and one without (or with less), the algorithm is made to give the compressed version a lower score. This is how I tend to think of it, a compression detector rather than a dynamic range calculator.

    With every player coming out these days having all kinds of DSP plugins, and with algorithms like replaygain going around, there is no reason for any compression (not that there ever was). But the simple fact remains that louder sounds better, so as long as people are listening to music on radio-like services (streaming sites included), the tendency will be to ratchet up the loudness to hedge bets against sounding softer than the next guy in the queue.
     
  7. TheSonicTruth
    I own about 20-30 classical CDs. Of them, I ran the tracks on 10 of them through Foobar2000's Dynamic Range snapshot. And like you, but perhaps more so, I was let down by the results. The app I mentions analyzes individual tracks or an entire album, and returns the highest DR value as the final value for a whole album. It analyzes most lossless and lossy formats of audio.

    The classical CDs I ran wavs for all clocked in between DR12-14 - typical for late 1970s pop and classic rock music! Even Kunzel's 1812 on Telarc, cannons and all, returned a measly DR12, leading me to suspect they must have used a heck of a lot of limiting on those guns toward the end! And a lot of compression in general on the actual orchestral stuff. I was expecting at least DR30 for that particular disc, and around DR20 for things like Beethoven's No. 5, etc.

    This is disappointing and deceptive, and probably lends some truth to the conjecture that 16bits really is "not enough" dynamic range for accurate reproduction of hyper-dynamic(20+dB crest factor) content.
     
    Last edited: Oct 5, 2017
  8. danadam
    Um... have you even read this thread? In one of the replies above RRod posted a link to a document describing how DR value is computed by the meter. It is not a crest factor. You can generate a simple sine tone and modify its loudness, so the difference between the first part and the second part is for example 40 dB and still, the meter may show anything, DR2 or DR15, depending on the length of the quiet and the loud part.

    I can rely only on youtube version of Kunzel's 1812, but level meter in Audacity shows around -40 dB in quiet orchestral parts and full scale on canons. Is that not enough? They could easily go even lower with those quiet parts, but I guess they decided that it would be unlistenable then.

    And that's how rumours and false information begin :frowning2:
     
    Last edited: Oct 3, 2017
  9. Strangelove424
    Most of my classical music’s DR is in the teens, as well as the rock, jazz, or soul music I find to be well mastered too. There’s actually a very narrow range of DR that works well. Much of this has to do with the way sound works, and the structural design of the ear. We are only built to deal with certain ranges at a time comfortably before biological protection mechanisms set in, particularly a physical tightening of the ear drum by the tensor tympani muscle with too much stimulus. Which means humans really don’t hear 140db of range at a time the way audiophiles tout. In fact, that would be torturous. Assuming that the noise floor of any environment is ~35db and comfort limit is ~85db (the OHSA’s upper limit for safe working environment, which hopefully you’d apply to your audio listening) that really just leaves about 50db for a safe human hearing dynamic range. The comfortable range will be even lower than that. 16 bit has a maximum range of 120db (or 96db with dither) which is more than enough to enjoy music, in fact it can cause you discomfort or even long term hearing damage with prolonged exposure. And 24 bit’s 144db upper limit is actually enough to be weaponized, and could rob you of long term hearing capability within minutes immediately. So let’s toss the bit stuff to the curb already. It’s not a format limitation, it’s a mastering decision. And if you ask me, keeping the DR below 30db, or even better below 20db, makes it more enjoyable to actual human listeners who have their own biological limitations and comfort factors to think about.
     
    Last edited: Oct 3, 2017
  10. danadam
    Didn't you mean it the other way around, i.e. 120 dB with dither?
     
  11. TheSonicTruth

    " It’s not a format limitation, it’s a masteringdecision.
    And if you ask me, keeping the DR below 30db, or
    even better below 20db, makes it more enjoyable to
    actual human listeners who have theirown biological
    limitations and comfort factors to think about."


    But if actual symphonies are capable of such wide ranges in loudness, why not preserve, if not all, at least more of that on CD and other digital formats? When I load a classical CD into a player, I want to experience as much of that performance or session as it would have sounded in person.

    Guess it's akin to what Henry Ford once said about assembly lines for cars back in his day: 'You can have any amount of dynamic range on CD, as long as it's DR12.' Mastering Engineer Ian Shephard advocates DR8 as a 'reasonable' value nowadays. Shall we lower the bar some more?


    What a waste of a great format....
     
    Last edited: Oct 3, 2017
  12. Strangelove424
    I thought dither reduced DR because of a higher noise floor, but I could be mixed up between actual vs perceived. But even with 96db, there’s still plenty of range left to damage hearing with, or perhaps even enjoy.

    Have you ever had to put your hands to your head to protect yourself from the orchestra? Probably not. An orchestra is not a rock concert (~100db peak) so to assume you need any more than 96db to depict an orchestra isn’t correct. I’d assume a typical orchestra peaks around 80db. Factor in the ambient noise floor of every shuffle, cough, breath, and rustle of paper that must atleast make it to 30db, you’re back at a 50db range again between noise floor and max. As has already been point out, the DR doesn’t necessarily mean lowest SPL vs highest SPL, so you could very well be experiencing close to a 40 or 50db range on some classical CDs with much lower indicated DR. The problem is that there are so many different factors you are trying to compare with live vs reproduced music, that pinning it all down on DR isn't accurate. Of course a live symphony is better than reproduced, but reproducing the DR isn’t really the limiting factor. For one thing, most music halls have incredible acoustics and reproducing that, especially with mic limitations, I think is the hard part. Not to mention reproducing entire sections of instruments vs highly localized woofers. But I don’t find myself hating the DR of most classical music albums. I find I very much enjoying them actually. Reproduced music will never match live music, but I do not think of redbook standard as a wasted format.

    Personally, I can enjoy a DR as low as 8 or so, but really like to stay in the teens. Anything too close to twenty is too much I think. Some movies mastering with DR of 30db I just hate. I could go on for days about how annoying that is, but just Google “always have to change volume on movies” and you’ll find other pissed off people like me. At a certain point, more DR really doesn’t help. Like I said before, it’s a really narrow range that's optimal. And once you have to go sprinting to your remote with your ears on your head for protection, I think the mastering engineer failed.

    Edit: I think it's relevant to point out that you were only disappointed by the DR of your classical albums once you looked at a number. They sounded like 20 or 30db to you, and only until you looked at that number did your feelings change. Isn't listening the important part? If it sounded good, and dynamic, what meaning does the number hold on its own? The DR meter is not a benchmarking utility.
     
    Last edited: Oct 3, 2017
    JaeYoon likes this.
  13. castleofargh Contributor
    while you're concerned about what can exist, most albums are produced trying to offer a somehow pleasing experience. of course excessive dynamic compression is simply not a pleasing experience, and quite frankly it can be really tiring to my ears. but the other way around with no compression at all can often be even less enjoyable to me. the moment a track forces me to pick between hardly hearing the quiet passages because they're too low, and blowing my ears up the rest of the track, I consider it a bad master. how the real thing had crazy loud passages only tells me that I would have left midway or used earplugs, which is another kind of annoyance. I see nothing desirable in having to listen to music at really high volume.
     
  14. TheSonicTruth

    The difference between a typical rock concert and a symphony is that the symphony doesn't *stay* at those high levels for very long. Its average SPL is perhaps not even half that of a rock concert. Also it's not only about how loud either can get, but how low either goes.

    Except maybe for a ballad or two, the rocker will stay up there, in average volume terms, for most of its duration. Average SPL might be as high as 100dB, and rarely dip below 85.

    Conversely, a classical symphony might crescendo at 90dBSPL, but average around 50-60. It's that *difference* in loudness I'm talking about, and the difference in loudness within a live symphony performance is normally much greater than that at a live rock performance. Accordingly, the range of loudness in a classical recording should not be reduced to the range of loudness within a rock or pop recording, as measured in DR, on a format that can handle both loudness ranges very well.

    I just find it interesting that among the older discs(1990 and earlier release) in my CD collection, DR12 seems to be the typical result that I get via Foobar2000, irrespective of genre(!) There are a couple outliers: "Brothers in Arms"(DR15), and a Chic compilation(DR14) - both of which clock higher than my classical CDs(most of which released in the 1980s). That would seem to indicate that more compression was used during transfer of that legacy material to CD than we might have suspected, even before the late 1990s "remastering" era.


    Yes, DR is only a number, a snapshot, but it seems to hint that the symphonies in my collection might have had their 'wings clipped' so to speak, during mastering, rather aggressively compared to what is done with typically less dynamic pop or rock albums on CD.


    In other words, if a symphony or jazz ensemble is more dynamic than a rock or pop concert, in person, then that should be, as much as is reasonable, the case with CDs of those genres. That's all I'm trying to say. Of course it's important that these classical performances "translate" reasonably on a wide range of listening devices(from iPods to 100watt per channel home rigs), but my final, ultimate priority is realism.
     
    Last edited: Oct 4, 2017
  15. danadam
    Sigh... you still don't seem to understand what the foobar's dr meter is actually measuring. Roughly speaking it takes only the loudest 20% of the track into account and shows the difference between the second highest peak and the RMS of those 20%. The rest of the track may be the same loudness as that 20% or it may be very quiet, it can even be a total silence, it doesn't matter, it doesn't affect the DR value.

    Here, see at those 2 files. Let's say the first one is your rock concert and the second one is a classical concert. Guess what are their DR values:
    dr14.png
    Surprise, surprise, they are both DR12:
    Code:
    foobar2000 1.3.10 / Dynamic Range Meter 1.1.1
    log date: 2017-10-04 16:33:29
    
    --------------------------------------------------------------------------------
    Analyzed: ? / ?
    --------------------------------------------------------------------------------
    
    DR         Peak         RMS     Duration Track
    --------------------------------------------------------------------------------
    DR12      -2.89 dB   -14.84 dB      2:00 ?-fileA
    DR12      -2.90 dB   -21.72 dB      2:00 ?-fileB
    --------------------------------------------------------------------------------
    
    Number of tracks:  2
    Official DR value: DR12
    
    Samplerate:        44100 Hz
    Channels:          1
    Bits per sample:   16
    Bitrate:           153 kbps
    Codec:             FLAC
    ================================================================================
    
    testfiles.zip
     
    Last edited: Oct 4, 2017
    JaeYoon likes this.

Share This Page