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
11 12 13 14
  1. bigshot
    You know that the 2 CD complete box set is transferred at the wrong speed, don't you? They were recorded at 76 rpm and played back for the CD set at 78. They found that out because when the box set came out a lot of guitar players tried to figure out his fingering and discovered that it was impossible at the pitch of the CD. But when they slowed the disc down a bit, it fit right in with normal chording.
    Glmoneydawg likes this.
  2. Strangelove424
    That's what everyone is trying to do.

    Ironically enough, I think there's a pretty good concentration of old school outlooks here, but an amp or source don't make someone old school. In my experience, alcohol makes the forums so much better. I'll join you! :beerchug:

    Numbers, schnumbers! Who wants to calculate RPM speeds when they're listening to music?

    (and yes, that is tongue-in-cheek)
    Glmoneydawg likes this.
  3. gregorio
    The problem is, the analysis tools which exist are NOT meant to "supplement listening impressions", they are meant to measure and/or display information about the properties of sound waves, the frequency, amplitude and timing. This is an important distinction because "listening impressions" are typically only vaguely/indirectly related to the actual properties of sound waves and in some cases, not related at all. Loudness is a good example because while it may seem trivially easy to determine, loudness is NOT actually a property of sound waves, it's a "listening impression". On the other hand, loudness is a bad example because it's just about the only exception to the general rule that there are no tools capable of measuring/analysing "listening impressions", although it should be noted this analysis tool took many years and a significant international effort to develop.

    When it comes to the production of music (or audio in general), we cannot measure how much of anything (EQ, reverb or any other processing type) has been added during mixing and/or mastering. Some things, such as EQ, can be measured because it's clearly defined by sound wave properties but still, we can only measure the properties the sound wave has, not how much EQ has been added. Other things, such as reverb, we can't accurately measure at all, let alone how much has been added. Compression falls into this category of something which can't really be measured at all, let alone measure a "listening impression" of it.

    Some analysis tools, the DR Database being a good example, are essentially marketed to consumers as being a tool "meant to supplement listening impressions" but in practice, that's just marketing. While having tools that "could prove" listening impressions might seem like a good idea, as they could potentially aid consumers and solve numerous audiophile debates/arguments/assertions, such tools simply do not exist and aren't likely to in the foreseeable future. And, it's dangerous to assume that the analysis tools which do exist are capable of proving anything about "listening impressions".

  4. Glmoneydawg
    Heard a lot guitar players talk about his i know why....i guess it wasnt the deal with the devil devil then?
  5. Zapp_Fan
    Pretty much agree with all of this. There is no utility that can tell you "This song is mastered better than that one" any more than there is a camera that can tell you which painting or sculpture is "better". However, I would quibble with the idea that you can't tell if reverb or compression is added. You can't really tell if it's been done with the intention of not letting the listener hear it. But it can also be pretty obvious.

    I would say this - if you can't easily hear whether there is compression or reverb or EQ added... it's going to be nearly impossible to determine it using a software tool. That would go into the realm of forensic audio analysis, not just advanced music listening.

    That said, I would say various tools are good for checking your impressions, although more to calibrate your own ears, rather than judge a mix. If a mix sounds good, it's good. There's no need (or way) to prove this further with analytical tools. If you think a mix sounds bad, you can't be proven wrong with a spectrogram or vectorscope plot.

    However, if you think you notice something about a mix, looking at the spectrogram, waveform, or vectorscope can be a good way to train yourself to recognize things about the mix, or quirks of your own listening setup. If you think to yourself "this high end sounds harsh" you can then use the tools to determine if the extra harmonics are actually in the mix, or if they're coming from your amp, headphones, etc.

    I would not at all discourage people from using audio analysis tools to back up their subjective impressions, but not with the attitude that you can prove something about artistic decisions that have been made.
    Last edited: Mar 2, 2018
  6. bigshot
    The way to tell for sure is to listen to different masterings and use your ears to determine which one sounds better.
  7. Husky
    Thanks for your contribution. I can mostly agree, except intersample overs.
    While it is true that some DACs can handle intersample overs better than others, it's not a real problem by itself.
    Tracks with an audible amount of intersample overs typically have clipping and/or hard limiting as well, resulting in lots of flat areas on the waveform during more than two samples. That means there is distortion anyway no matter what DAC is used. You can see that quite well on the track from the article you've linked.
    So I would first blame the loudness war.
  8. pinnahertz
    As I said, "Clipping audibility covers the full range from completely inaudible to gross distortion, but to assume that just because something looks clipped that it is, and that clipping is audible at all would be naive."

    The same applies to intersample overs.

    "Distortion" does not automatically mean "audible".
  9. Husky
    I was quite busy these days so I couldn't reply earlier.
    1. Already explained. And of course it's not the only reason.

    2. "Inappropriate over-compression" is a long wording. And why shall "over" be more neutral than "hyper"?
      If there was a consensus about the appropriateness of over-compression for a certain piece of music, in my view it could be called hypercompression as well. I have the feeling that it is difficult to precisely and officially define the appropriateness because apparently tastes are not all the same, so we would end in the same fruitless debate again.

    3. Your more detailed description makes clearer what you meant and I wouldn't have argued that in the first place. Still and even more I can't see what you think that I wouldn't understand about the loudness war.

    4. Tell to Metallica fans who prefer the guitar hero 3 version of Death Magnetic (see their forum) that they don't listen to heavy/thrash metal...!
      And to make it clear: the CD version is 10dB louder, the guitar 3 version has RMS of -16 to -19 dB without any clipping and moderate compression (ClippingAnalyzer compression indicator 0%-4% vs. CD 47-65%):
      CD vs. GHIII
      Result_DM_CD.PNG Result_DM_GH3.PNG

    5. I have listened to it. Do you think there is "over-compression" used?
      How do we know to what extent the voice level was regulated by Cobain himself by varying his distance to the microphone, or later by (linear) amplification or by compression?
      The loudest chorus part towards the end still has RMS -13.3dB and ClippingAnalyzer shows a moderate compression indicator of 13%. As already stated I've never pretended that there was no compression - but at least no "inappropriate over-compression" or whatever wording you like.

    6. I have no statistics, do you? I think that the loudness war on the pop/rock genres gives pressure to all genres, first to jazz (already happened) and then to classical. A reason for this may be that there are people who think the loudness differences between the genres shouldn't be too big and irritating, or the attempt to attract younger listeners assuming that they are more used to high loudness.

    7. As I've already said, you have to differentiate.
      7a. See 4.

    8. I've already explained this and obviously you don't want to differentiate, it's the same as with 4. and 7.
      Maybe two other examples help you to understand.
      First let's take The Art of Noise. I've got the best of CD from 1988 and like it very much. There is plenty of distortion (and I guess heavy compression) on some sounds - where it musically makes sense! At the same time there is air and dynamics. Overall RMS and compression indicator look very good.
      For example the drums in "Close (To The Edit)" are very distorted, but it's suitable and the track breathes.

      The second example is Yello (electronic genre). I've got an unmastered version of "The Expert" (from the Stereoplay magazine 01/10). The mastered version sounds good, but the unmastered version sounds better.
      Boris Blank tells about the unmastered version that it is "how it leaves my studio. The original. Overall the volume is lower of course, but the dynamic range is even bigger, with quiet parts and extreme peaks. But you actually get the chance to hear how much happens in the quiet parts. My gift to all stereoplay readers." (I hope my translation from German is adequate.)
      Mastered (compression indicator 27%) vs. Unmastered (6%)
      YelloMasteredResult.PNG YelloUnmasteredResult.PNG
      You can see that the unmastered version has a histogram similar to pink noise (yellow line), especially in the range of 0 to -3dBFS (details right corner of the histogram), while the mastered version goes well above of that.

    9. If technically limiting is a form of compression and clipping is a form of limiting, then logically hypercompression includes clipping. As the term "hypercompression" isn't exactly defined, I think we shouldn't argue too much about it.
      I assume you agree that if the engineer uses a limiter with gain and hard clip and a limit of -0.1dBFS resulting in flat areas, this is equivalent to clipping.

      9a. I've never ever said that there was no compression before completing a mix!! Seems you have completely misunderstood my statements.
      If you are a mastering engineer, do you prefer a mix that is already extremely loud (over-compressed/limited/clipped and whatever)? I guess no? So possibly your artists listen to you and give you a less "hot" mix? If yes, then we have two versions and the consumer could be given the choice.
      I know that in many cases already the mix suffers from the loudness war, but in most cases the mastering engineer adds some loudness on top, doesn't he?

      9b. Would be great if some open minded engineers see the chances for better supplying the consumers demand by product differentiation. And I hope you can break out of your genre dogmas...
  10. bigshot
    A lot of people don't understand about the loudness war because they get all their info from other people on the internet and popular articles in the press and don't actually speak to engineers directly about it. You are lucky. You're actually getting to hear how an engineer thinks about it. Take advantage of that. Listen and you'll certainly learn something.

    We're all here in this forum to be helpful and learn from each other. If we were all supposed to be experts whose job it is to set everyone straight, we would be getting paid for this.
    Last edited: Mar 3, 2018
  11. RRod
    The 'unmastered' content appears to have no appreciable dynamism (can't read the scales but the amplitude axis appears linear?); have you done an actual volume-matched comparison?
  12. Husky
    • The amplitude axis (waveforms) is linear. (To be able to read the scales you might have to right click on the picture, let your browser show it separately and magnify it to the full resolution)
    • What do you mean by a volume-matched comparison?
      • If you mean a listening comparison with equal loudness (based on ReplayGain values): yes, of course.
      • ClippingAnalyzer shows the waveform as it is, whereas for the measurements (RMS, clippings, compression indicator) and distribution graphs it does a normalization first (peak 0dBFS).
    • Below is a screenshot where the two tracks are volume adjusted to have the same loudness (the mastered was 4.2 dB louder) and to have the unmastered normalized to 0dBFS as well. So you can better see the different amount of dynamics in the waveform.
      The first track is mastered, the second is unmastered.
    Last edited: Mar 3, 2018
  13. RRod
    Ok, thanks. Too hard to eyeball how big a difference trimming off those peaks makes. Still don't see this as 'dynamic' material. Having a snippet of both tracks would be helpful.
  14. bigshot
    The easiest way to tell what the proper mastering is would be to listen to the first release. That is the one that all of the artists and original engineers signed off on. After first release, all bets are off and the artists are just signing off on the royalties. Perhaps comparing waveforms of the original release to remasterings might tell you something. But you still can't just judge by eye to make sure they look the same. There are lots of remasterings that sound much better than the first release. When it comes right down to it, I don't know how you can judge music by a waveform. I've seen waveforms that didn't look very much alike that sounded pretty much the same. Our ears are processing the sound and doing their own internal balancing as well. Waveforms are best for finding your place in a track while you're editing. I've worked with them a lot, and I can't "read" them for much beyond that. Maybe egregious errors? I dunno.
    Last edited: Mar 3, 2018
  15. gregorio
    1. I'm not saying that one can't tell if reverb has been added, I'm saying we can't measure if and how much reverb has been added. Even when reverb is obvious, it can be difficult or impossible to know if it's reverb which has been added or reverb which occurred at the time of recording (and was recorded). With an experienced "ear" and significant experience and knowledge of recording practices it is more possible to determine what reverb has been added but only sometimes. Software cannot make this distinction though, it cannot even measure the total amount of reverb, let alone how much has been added.

    2. I would say - that it completely varies from case to case; sometimes huge amounts of EQ may have been added but you can't easily hear it, sometimes relatively small amounts of EQ may have been added which are quite easily heard. Likewise with software analysis or rather, not really "likewise". "Likewise" in the sense that it can sometimes give us a good measurement/indication of added EQ and other times it can't but not "likewise" in the sense that those "sometimes" are often not the same "sometimes" as with hearing. In other words, if you can't easily hear whether there is added EQ then sometimes software can make a determination where our hearing cannot. And vice versa, where we can easily hear whether EQ has been added, sometimes software analysis can also make that determination but other times it can't. BTW, I'm talking about forensic audio analysis software here and, I've picked EQ because compared to compression it's relatively easy to measure in the first place.

    In case it's not clear, I'm broadly agreeing with your post, except that there's no real correlation between what we can hear and what software is able to determine and of course, both of them in effect only work as accurate measurement tools sometimes.

1 2 3 4 5 6 7 8 9
11 12 13 14

Share This Page