Tools for Analyzing the Quality of Mastering

Discussion in 'Sound Science' started by strangelove424, Sep 20, 2014.
1 2 3 4 5 6 7 8 9 10
12 13 14
  1. gregorio
    1. First of all you explained that as the loudness war increased, so did the sales of vinyl. That was incorrect as in practice the exact opposite occurred and even if there were a direct correlation, a correlation does not prove a causation anyway!

    2. "Over" (over-driven) compression has been a musical tool since the 1960's and is desirable/essential to many genres and musicians (Hendrix for example). Also, "over-driven" is relatively easy to define, as non-linear distortion caused by a deliberate overload condition. "Over" is therefore both definable and musically neutral, whereas "hyper" is neither!

    3. See #6.

    4. You stated you'd read the article I linked to but obviously that statement is either false or you did read it but somehow failed to understand any of it! You do realise there was a section of the article specifically dedicated to "Death Magnetic"?

    5. How can you ask that question if you have listened to it? Of course "over-compression" has been used, if over-compression had not been used it would be a different genre! And the same is true for Death Magnetic, the Metallica fans would have liked "Death Magnetic" even less if there hadn't been any over-compression because it wouldn't even have been trash metal anymore.
    5a. "Varying his distance to the microphone" is compression and so is dynamic amplification! The only difference is that it's manual compression and in the case of dynamic amplification, typically called "riding the gain" but both are still compression. The problem with static amplification is that you're limited to how much amplification you can add by the highest peak value in the whole track/channel.
    5b. To achieve an RMS of -13.3 requires heavy compression and again, very obvious "over-compression" has been used. And also again, you are relying on a tool which is at best an indicator and at worse, nonsense. It is referenced against pink noise correct? I'm sure whatever metrics it is using of pink noise it would be possible (and probably trivially easy) to record something which exceeds those metrics, and then what, how would ClippingAnalyser show compression of less than 0%? Or, to put it another way, it's almost certainly possible to create a recording with compression which ClippingAnalyser registers as 0% compression and another recording without compression which registers as some percent higher than 0%. The reference of pink noise is abstract and compression cannot be accurately or reliably measured.
    5c. But that's the whole problem! What amount of compression is "inappropriate over-compression"? What maybe an entirely desirable, even essential amount of compression (or over-compression) for one piece of music might be an entirely inappropriate amount for another! That's the problem with Death Magnetic (as obviously you did not understand the article): An entirely appropriate amount of compression was applied to Death Magnetic, unfortunately, it was an appropriate amount for EDM and some other contemporary genres but NOT thrash metal! The problem facing Metallica was that the whole ethos of thrash metal is to be loud and distorted, which was somewhat difficult to achieve in the 1980's but entirely possible. However, in 2008 Death Magnetic was in competition with a new generation of genres which employed far greater compression/loudness than the genres thrash metal was competing against in the 1980's and they either therefore had to apply completely inappropriate amounts of compression or not achieve the relative loudness demanded by the genre.

    6. Yes I do! Go to the DR Database and in the album box type "Schubert". You'll see the DR figure is pretty much the same in the most recent recordings as those from the '70's and '80's. You'll see the same if you type in "Tchaik", "Beeth" or other big name composers. Where's the war?
    6a. You're free to think whatever you like but you have made a claim/assertion with no supporting evidence and worst still, the best evidence we do have, actually contradicts your assertion!

    The rest of your post, and indeed most of the above as well, is based on arbitrary percentage levels of compression or hypercompression and distortion which you cannot define except in terms of what you personally like or dislike but in an attempt to justify/validate your likes/dislikes you are missing and sometimes misrepresenting the facts. A couple of other examples to hopefully put this nonsense to rest:

    9. Yes, that would be the logical conclusion. However, as clipping is NOT a form of limiting and in fact the whole point of limiting is to avoid clipping, then your conclusion is nonsense!
    9b. That has ABSOLUTELY nothing whatsoever to do with engineers, let alone how open minded they are! That decision is purely down to the artist (and/or record label).
    9c. OK, now you're just being ridiculous! You think maybe I should ignore "genre dogmas" and mix or master classical impressionist music the same as I mix and master thrash metal or vice versa? I wouldn't be long for this job if I took your advice, just one such mix or master would get me the reputation of utter incompetence!

    Last edited: Mar 4, 2018
  2. Husky
    1. First of all I reported my personal experiences and what I've read e.g. on forums discussing different releases, where people often advise to prefer the LP because it suffers less from the loudness war

    2. So, respecting your #9, we would have to use "inappropriate over-compression and/or clipping" to describe the means of the loudness war, right? Would be unwieldy, wouldn't it?
    3. -
    4. Your very absolute and undifferentiated verdict that I wouldn't have read something, that in fact I've read, or wouldn't understand anything, where it is obvious that I understood at least quite a lot, gets really annoying.

      Previously you stated:
      "Remove the extreme loudness and severe distortion/compression from heavy metal and what you're left with is no longer heavy metal!
      (...) If a thrash metal band comes to me with a dynamic, slow, sweet sounding track, that's absolutely fine but by definition, it is NOT trash metal!"
      As the Death Magnetic CD has extreme loudness, severe distortion and compression, by your previous definition it's perfectly trash metal, whereas the Guitar hero version isn't, as it has very low loudness (compared to what is common these days) and no distortion from clipping. See 5.c) for further discussion.

      Do you think there is over-compression in the GH3 version? If yes, how much would you think there is? Would be great if you could tell me some more or less corresponding settings in Audacity.

    5. b)
      There is a reason it's only called a compression indicator. Indeed, quiet uncompressed music shows lower Z-values than pink noise, so maybe it should show a negative compression indicator instead of a lower limit of 0%. All the exact values can be viewed in the ClippingAnalyzer result sheet, and there it shows with e.g. -14% on such music. But maybe it would be irritating to see negative values. Perhaps it would be better to rename it to e.g. "compression comparator" or something alike.
      Would be great if you could upload some examples with some different and known levels and types of compression, so we can actually test what it shows.
      When I analyze I don't have the requirement that the values are absolutely exact, so e.g. +/- 5% is no problem for me.
      You argument that each genre has a certain amount of compression (what about digital clipping?). What amount exactly and how do you measure it? Shall it be absolute or relative to other genres at release date?
      So for Death Magnetic, you seem to prefer an absolute amount, that would have been lower than it is on the CD. This contradicts with your demand that the genre must have higher loudness than later louder genres. So now it would be impossible to release new thrash metal, because you can't fulfill the absolute and relative requirement at the same time. Would you have advised Metallica to either lower the loudness, leaving thrash metal (going to what genre?) or keep their loudness and change their music (in terms of instruments, arrangement and so on) to EDM?
      I think this shows again how absurd it is to think that music must have a certain amount of compression to be allowed to be classified into a certain genre.

    6. There are very view classic albums in the DR database. Let's look at Tchaikowsky: 12 entries for 8 different releases, where the latest relase has the lowest DR. Schubert has even less (and I think it may be difficult to compare a symphony to a piano trio). Beethoven has some more (38 entries, 30 releases) with a tendency to get louder. So you don't have the statistics either, and what we have doesn't proof that classical music would be completely unaffected.
      (lower is louder) (30 releases)
    9. From Wikipedia: "Clipping is an extreme version of limiting." You may discuss it there. And you may discuss with e.g. Audacity developers (the Limiter has hard clip as an option).
    9. c) Yes, I know you could scandalise and that's deplorable. In my view as a listener, that's a big part of the problem with the loudness race, it's a herd like behavior. It seems people in the industry were afraid of achieving good sound quality because the result could be less loud than others.

    10. I'd like to heed bigshot's advice and learn from you. And I'd like to get the discussion more constructive. So let me ask:
    a) What can be done to end the loudness war and set the goal on better sound quality?
    b) What have you done and could you do personally as an engineer?
    c) What can consumers do? (Other than to restrict oneself to classical music and hope it doesn't get more affected...)
    Last edited: Mar 6, 2018
  3. pinnahertz
    I'm afraid I'll have to agree here. Clipping is a form of limiting, and is actually used that way in every modern broadcast audio processor. However, that does not mean deliberate clipping is audibly bad. There are many ways to use a clipper, and by "clipper" I'm not referring to deliberately overdriving a system, but rather a device or processor made to accomplish precision clipping. The reason for the differentiation is that many systems, when overdriven, don't simply clip, they can do some unpredictable things. A clipper in a processing chain serves a definite purpose, though. One example is when processing for media with a hard maximum limit, like analog broadcast systems. If you try to control instantaneous peaks with a peak limiter you have problems because a peak limiter must alter gain quickly, with an "attack'" time that results in full limiting of a 1/2 cycle of the highest frequency in the audio pass band, and then "release" back to normal gain at some rate that doesn't create an undesired artifact. The result is quickly elevated intermodulation distortion because of the rapid gain modulation of the peak limiter. An alternative solution is to slow the peak limiter's attack and instead control overshoots with a clipper. If time constants and thresholds are correctly chosen, the result is inaudible clipping, and less audible IMD. Some broadcast processors go a step farther and apply a form of distortion reduction post-clipping. The advantages can be several dB increase in loudness without as much gain change side-effect and IMD. The peak limiter and clipper are tightly calibrated with each other. But that's not something that will be obvious on any dynamics analyzer, and will not correlate to audible result.

    Clipping is not always bad, not always audible, and mostly cannot be well differentiated by analysis of the final result.

    However, the use of a clipper as the Only form of limiting is just stupid, sounds terrible, and pretty much is not what's going on, even in Loudness War Processing. That's why you wouldn't consider a clipper as a "limiter", it doesn't work by itself, it must have other processing ahead of it, and interacting with it.
    Spot on. But it doesn't matter.
    First step is awareness. Have fun with that one. Loudness war processing has been with us for 60+ years in one form or other. And, as the 2014 AES paper "Hyper-compression in Music Production: Listener Preferences on Dynamic Range Reduction." - (Robert W. Taylor and William L. Martens) (cited in my earlier post) clearly mentions as an unstudied area of potential bias, the "influence of the cultural and creative models of the individual genres of music" may have a powerful effect on what is perceived as acceptable or even "normal" for any given genre. The polar comparison might be classical (where the loudness war doesn't exist) vs EDM. That makes "awareness" difficult because those favoring a particular genre will also be pre-conditioned as to what is normal, or even part of the definition (as Greg has stated). There's no easy answer for awareness on the consumer side.

    On the production side, you have a different problem, genre-linked processing expectations not withstanding, that of relative loudness competition. It's partly an ego-game, partly misconception of the importance of loudness in the first place. How do you get the first guy to back off? And how much do you back off without redefining what's necessary for the definition of the genre? Very difficult. I would hope that the more independent labels and artists might take the leap, but I'm very skeptical. One electronic artist I'm familiar with re-released several of his tracks "in High Quality Audio", but what he did was 24 bit and a slightly higher sample rate. Nothing was done to the loudness processing. At least he's aware that quality could be improved upon, even if he didn't understand how.
    Very little. We're asked to do something and get paid for it. We can educate to the extend a client will accept that education, but in the end, if someone wants a road built with a bump in it, that's what they'll get, even if we know how to build a smooth road.
    I've always said "Vote with your wallet", but frankly, that's silly because it's been long proven that for the masses, quality doesn't mean crap. It's all about content. People will listen through poor quality audio if the content is good. This has been proven in broadcast audience data for years. In markets with dozens of signals, audio processing and audio quality have no correlation with popularity and listener numbers. It's why an AM station can still beat an FM station in ratings, and why the least processed, least loud station in any major market will also have the lowest listenership. It's not the processing, it's the content, which will likely be classical, and doesn't command popular numbers.

    But, back to voting with your wallet. It actually does have an effect when refunds are given, because pay-outs have to be accounted for, and far differently than sales. If you like an artist and the latest release is smashed to death, demand a refund because the quality is bad. If one person, just one person does it they may think he's really just a nut and they'll ignore him. And if two people...two people do it, in harmony, They may thing they're both idiots and They wont pay attention to either of them. And three people do it, three, can you imagine, three people buying music then demanding a refund for bad loudness war quality, well They may think it's an
    Organization. And can you, can you imagine fifty people a day, I said, Fifty people a day buying a track, then returning it and demanding a refund because of bad audio quality! Well, friends, They may thinks it's a movement!

    And that's just what it is , the Alice's Restaurant Anti-Loudness War Processing Movement! Sign up here, and I'll send a royalty check to Arlo Guthrie. (OK, kids...go google that one)
  4. Zapp_Fan
    I'd just like to say that intentionally using clipping as a form of limiting in music *production* is extremely rare as far as I know, but it's very interesting to hear that it's useful in broadcast! I guess latency is more of an issue in a live environment, imagine that.

    As far as the way to win the loudness war, Pinnahertz is pretty much right IMO. Labels and artists that cater to the mainstream aren't going to back off their master limiter thresholds until 1) they start losing money over it and 2) lots of people very clearly, explicitly, and repeatedly ask them for more dynamic range. Pinnahertz is getting at something interesting, which is that it doesn't take millions of people to effect this change - just a considerable handful of people being very loud (haha) about it. The vast majority of people don't care either way, so if the labels think they can squeeze some more revenue by giving the opinionated few what they want, without losing any of the indifferent masses, they'll do it.
  5. pinnahertz
    Oh yeah, it's used in production! But if done "well" (subjectively), you might not know it's used. Been used in broadcast since the 1960s, and much more as tech got better. Latency has nothing to do with it though.

    If labels thought they could sell more copies by making a "lite" version, they'd do it, but they may (correctly) have analyzed that it doesn't work like that. When beer companies introduced Lite Beer, beer sales didn't increase at all, but now they have an entirely new production line to support. It will be the same for music.

    Yeah, I know "you" would buy more music if you could get the less processed version, but "you" don't bump the stats at all.
  6. bigshot
    Clipping in the analogue era was a lot more common than clipping today. You could burn into a 24 track and get a really nice dynamic swell without a lot of noise. I wouldn't use clipping on digital myself. There are better ways to skin that cat.
  7. pinnahertz
    Clipping in the analog era? Hmmm....there were pretty much zero actual analog clippers in the analog recording world. You're probably thinking of analog tape saturation, which was definitely used but nothing like clipping. Tape has a soft and mushy frequency dependent saturation characteristic, the resulting distortion and IMD products are quite different from that of a clipper.
  8. bigshot
    They called it "burning in" back when I was working on 24 track.
  9. pinnahertz
    Yup, that's the stuff. "Burning" was slamming into the tape as hard as you like. The thing about tape is distortion goes up very slowly above reference level (185nWb/m, 200, 250, 320, take your pick), and gets progressively less linear the higher you go. And the distortion goes up faster at higher frequencies. Intermodulation distortion goes up too, even faster. The resulting sound is a sort of soft overload, junky without being too crackly (how's that for descriptive?). Tape formulation and speed changes how this works too.

    Clipping is a hard, absolute maximum that happens very quickly, with distortion going up quite fast above clipping. "Soft Clipping" is essentially a more gradual onset of distortion caused by nonlinearity, the degree of which can be adjusted in some cases, from very, very soft to hard. Everything affects how audible the results are. If you're familiar with how a silicon diode works, it begins to conduct somewhat gradually as the applied voltage reaches its cut-in voltage when forward biased. That effect can actually be used as a soft clipper. Then, if you reverse-bias the diode above 0 volts, the cut-in effect occurs at a higher voltage, but over a smaller percent of the total applied voltage, making clipping harder. Pretty simple and basic, but useful to an extent. There are better means to do hard clipping, of course. How hard vs soft the clip is determines how firm the clipping threshold is, now rich the resulting harmonic content is, and how audible the clipper is given the same amount of clipping for the same time duration.

    Ok, more than anyone wanted to know. Sorry, the old melon is filled with useless trivial info that just leaks out sometimes.
  10. vertical

    Another business approach could be something modeled after what’s happened in the movie industry with 3D or IMAX. Overall sales quantity may not change, but the companies could charge more for the additional version.

    Yet, they’ve probably thought of that, explored it, and are holding off...maybe waiting for the release that someone else funds that alters the landscape like Avatar did for movies.
  11. castleofargh Contributor
    as far as I can see, the industry has fully adopted that idea with anything they could already. stuff like formats and resolutions where it's easy for people to understand the good old "more is better" marketing. even how vinyls still exist is IMO evidence that if something can bring any money, clever guys will provide it.
    but how do you market an idea of better mastering? you can do it for the lolz like MQA and Pono, where you claim to sell quality in general. it costs nothing because it's a subjective idea anyway and comes with zero legal or practical requirements. but how do you do that for day to day albums with different masters? or in the specific case of compression, what tool, who's opinion is going to become law and tell what is deluxe and what is for normies? very risky idea IMO. take DR database as reference and the same day you'll have people abusing the tool to make an extra buck from the same initial product.
    if we based the value of a song on the look of the signal like many do, what would stop anyone to take some random album, run them through a gate to make it look pretty and dynamic on the visualization, and voila! a version worth more money because of a flawed classification method.
    then what's next? people starting to play as loud as they can so the record can have even more dynamic?

    even those in the industry who really want to focus on some personal standards of quality like a few small record labels do, have to face how difficult it is to control and define the idea of good sound. in the end we like an album or we don't, that's still very much how an album is judge IMO. as a complete piece where individual variables are mostly irrelevant on their own.

    some of the highres albums are the way you want them to be, some SACD and DVD too. with that idea that those will be purchased by guys with particular expectations like more dynamic. but as we've seen from the start with highres, nothing is ever for sure. some end up with even worst masters and even to this day, some simply aren't highres at all. so what they sell for more money is the one thing they can control for sure, the damn container. :'(
    it's a dysfunctional system but nobody seems to have found a solution that satisfies consumer confidence and the money guys at the same time.
  12. pinnahertz
    What the movie industry has done is not comparable at all, though. First, the theatrical experience is established already as a go-out special event. Second iMax and 3D have not been universally accepted at all. In fact, a significant population segment views both of those as strongly negative. Third, 3D and IMAX are both clearly and obviously different from standard, easily perceived by everyone.

    Not likely the case with "lite" versions of heavily processed music. It's presented in the same experience environment as the standard version, would probably cost more, and be largely unnoticed as different (except by us of course). And, it would no doubt cost more. Content drives the market, cost anchors it. Especially in the USA, people in general will not pay for higher quality, especially if they can't even tell it's different. There's always a niche, but that's not enough to steer the ship.
  13. Zapp_Fan
    I basically agree, the average person would have trouble grasping what dynamic range is without a lengthy (by the standards of advertising) explanation, and so trying to sell them more of it won't work. Likewise, even most audiophiles, let alone "normies" don't really seem to get how compression works, so trying to sell people less compression only works if they already have an opinion on it.

    However, one thing I can tell you with certainty is that while "hearing is believing", believing is also hearing. If people are successfully instructed to believe that a certain style of mastering is better (call it "throwback" the way pepsi puts out "throwback" flavors) they'll be quite certain the sound is better, regardless of what it actually sounds like. Now, they won't necessarily pay more for it, but it could be a way for one label or another to steal market share, which is maybe even better than charging higher prices. This scheme would probably require celebrity endorsement (like Dr. Dre or something) and at least a few years of multi-million-dollar ad budgets, but it could be done.

    Now, I don't know if label execs really think in terms of market share, but I am sure that they don't mind selling the same old recording in new formats 3 or 4 times over. So if hi-res and vinyl start to peter out they *could* turn to "throwback" digital/streaming remasters.
    Last edited: Mar 8, 2018
  14. pinnahertz
    The highlighted part is pretty much impossible. Who would do it? Record companies already are selling a product just fine, and have no need to segment their market like the lite beer guys did.
    Sure, reselling old recordings is good, as we've already seen, remasters are not always done well, not always an improvement, and "high-res" versions that are just up-sampled are just money grabs, may even be done under license by a third party. The bigger issue is to what material do you have access to create the "lite" version? The final master will have processing baked in. So where do you go in the production chain, who retains artistic control at that point, and to what subjective level do you make your "lite" version? It becomes an artistic decision someone has to make, and many have to agree on. It's a messy process, far worse than just up-sampling a 16/44 master and releasing it as HD Audio. And what do you do about the new stuff that drives most of the market, and the final decisions have been made?

    My point is that creating a "lite" version is anything but simple, and for it to be a market success is an uphill climb with minimal returns. Remember, the listener preference data favoring less processing does not exist, but data to support that processing doesn't matter does (broadcast audience research for the last 60-70 years).

    Sorry to be so pessimistic, but I think we're stuck with what we have until there's a popular trend away from the loudness war. Something a few popular bands come out about making some sort of neo-political statement about the loudness war, and presenting their art differently. You need a movement about 10X as big as the so-called vinyl resurgence (so-called because the stats just don't paint vinyl as a significant market share yet, all the big numbers relate to percent of change of very small numbers). That's the "education" of listeners you need to bend the market trend, but it's still all about content first, then price, then quality dead last. So a Radiohead needs to self-release their "High Quality" version for free, then successive releases in high quality only while vocally educating their audience as to the benefits of their new version.

    Sorry I don't have those kind of connections.
  15. Zapp_Fan
    I sort of agree, sort of don't. My point is that cleverly branding "throwback" or "lite" masters as something desirable can get people to buy them regardless of how they sound. Beats headphones sold very well even when they sounded like absolute trash, because they were fashionable and branded well. I'm far from the first person to suggest that marketing and placebo is as powerful (or more so) than actual sound quality. I know maybe as much about marketing as I do about audio, I am confident it could be done with a sufficient budget. (millions for sure.)

    The same principles can work on anything, including certain types of mastering, if the industry decided to throw its weight behind it. So, real education is not necessary as long as trends and branding still work.

    I agree it's unlikely to happen, but it's far from impossible. It would just require the goals of better mastering and arbitrary marketing to align (more or less) by chance.

    Anyway, we're pretty far off topic now, I am talking about branding strategy in a thread that's supposed to be about spectrograms and such.

    Anyway, on that subject, I would like to vouch for the value of watching a live spectrograph in training one's ear. There is not much better than watching the output of a visualizer as you adjust different EQ bands, to train your ear to pick out frequencies.
    Last edited: Mar 8, 2018
1 2 3 4 5 6 7 8 9 10
12 13 14

Share This Page