TT Dynamic Range Meter

Discussion in 'Audio Hardware' started by Andrew Smith, Mar 9, 2011.

Thread Status:
Not open for further replies.
  1. Dukes Travels

    Dukes Travels Forum Resident

    He said some were based on erroneous info. Which in all law of averages is quite a sensible approach...considering this is the internet we are talking about.
     
  2. mwheelerk

    mwheelerk Sorry, I can't talk now, I'm listening to music...

    Location:
    Gilbert Arizona
    I don't care if he said one. I don't care if it is the Internet. Anyone making judgement on anyone else's opinion and view indicating they somehow have knowledge others do not is a problem for me. Be specific about the things you know and how you have any knowledge whatsoever what others do not know. Gross generalization serve no purpose.
     
  3. Dukes Travels

    Dukes Travels Forum Resident

    Well you can not care but the fact is that you were responding to my point and I made it pretty clear that i use the internet for info on the subject. And i have definitely read some erroneous stuff.

    I understand, you got your feelings hurt.
     
  4. Ham Sandwich

    Ham Sandwich Senior Member

    Location:
    Sherwood, OR, USA
    What is acceptable does depend on the type of music. Some music will suffer more from low DR numbers than others. It also depends on the production. The production for modern pop at low DR numbers is not the same as other styles of music at those same DR numbers. Don't judge all low DR numbers by pop music standards.

    Lady Gaga - Applause has a DR 5 and suffers (and makes me suffer).
    Adele - Skyfall single has s DR 6 and I can't make it all the way through the song.

    Some music doesn't suffer so bad. For example doom metal or stoner metal can get away with a DR 5 and still be sonically interesting and listenable. I wouldn't go so far as to say a DR 5 for stoner metal is acceptable, but it's not necessarily unlistenable and can still be enjoyable to listen to.

    Here's an example of stoner metal with a DR 5. It's on bandcamp so you can stream the songs and hear what it is like. That style of music is a bit of an acquired taste. I like it. I'd rather it was mastered with more dynamics and less compression, but I can deal with how it is and fully enjoy listening to it. This is probably not the best example to use because some people won't like the music. And it does have intersample clipping. It's not an example to put forth of good mastering. But there's lots of modern recordings that are done worse. Still, it's a DR5 and I can listen all the way through with no feelings of "gahhhh! this is awful sound, make it stop".

    Here's the bandcamp link: Stoned Jesus - Seven Thunders Roar (the band's name hints that it might be stoner music ;))
    And the DR analysis:
    Code:
    --------------------------------------------------------------------------------
    Analyzed: Stoned Jesus / Seven Thunders Roar [24/44.1]
    --------------------------------------------------------------------------------
    
    DR        Peak        RMS    Duration Track
    --------------------------------------------------------------------------------
    DR9        0.00 dB  -10.65 dB      7:47 01-Bright Like The Morning
    DR5        0.00 dB    -6.52 dB      9:21 02-Electric Mistress
    DR5        0.00 dB    -7.38 dB      5:00 03-Indian
    DR8        0.00 dB  -11.12 dB    16:02 04-I'm The Mountain
    DR8        0.00 dB  -10.07 dB      8:43 05-Stormy Monday
    --------------------------------------------------------------------------------
    
    Number of tracks:  5
    Official DR value: DR7
    Two of the tracks are DR5. The tracks with DR8 and DR9 do sound better than the DR5 tracks. Still, it's not "unacceptable".
     
  5. L5730

    L5730 Forum Resident

    ^ Whilst some material which happens to have a relatively low DR value (compared to what we would expected if it were an older release) may sound acceptable, the question was aimed more at determining the most preferred mastering. DR values won't really help there.
    Unfortunately we don't often get to compare two mastering with just a jack up in compression, without radically different EQ or shoddy sources.

    Now, people could be forgiven for wanting to hit me with a stick for what I am about to admit. We play a lot of tunes on a little netbook, and we actually run a compressor plugin for the bulk playlists. If the output was captured, and a comparison was made between the DR values of before and after compression, it wouldn't look pretty. But, for our purpose to playback like a radio station, we want to even out the level across all tracks beyond what simple ReplayGain can do.

    What I can tell is that clipping, crap tape sources and bad EQ choices sound a lot worse than bunging stuff through a compressor. So there is an example where a track with a higher DR value that has clipping and poor tonality would be worse than a modern compressed track with a low DR value, but perhaps more sensibly EQ'd mastering.

    Dynamic Range Day's target of DR8 for people making music is too conservative in my opinion for the most part. Most 1970's+ main genres would all sound good if they were left with the recorded and mixed dynamics intact. No engineer in the right mind should be putting compressors on the master bus at the mixing stage (*), it's simply a method to get more overall loudness - pointless. If the artist and the production team want to smooth out and squish individual instruments for effect, then yeah use track compressors, but there's no need for mindless bulk squishing - that's for the radio station to do for the purpose of a homogeneous feel when a new track comes on.

    * Dynamic Range Day webcast 2012:
    Charles Dye:
    "Well, that’s Ricky Martin’s Livin’ la Vida Loca, which, er, I’ll neither take credit for being the beginning of the loudness wars, but I would take credit for being how loud it is, that’s for sure.
    Others have blamed as being obnoxiously loud, and I will not disagree with that. But yeah, that was a song that I mixed, and made as loud as that when it was presented to the mastering engineer, Ted Jensen, and he essentially didn’t touch it as far as loudness goes. And he takes essentially no pride at all in it from a mastering standpoint simply because I didn’t give him, I say unproudly, I didn’t give him any headroom to do his job. And there’s no secret to how I did it, I just had L1, which was a fairly new and starting to become popular plug-in at the time, at that time that was mixed in 1999, and I had that across the mix bus and had it set at –6dB, so my mix was peaking at what it would have been peaking at normally, and then I made it about 6dB hotter, and then all Ted (Jensen) did was slightly alter the EQ curve."


    L1 - Waves plugins Ultramaximiser limiter. It makes stuff louder!
     
    thrivingonariff likes this.
  6. Ham Sandwich

    Ham Sandwich Senior Member

    Location:
    Sherwood, OR, USA
    The question was also about what is acceptable and whether the "bad" grouping of DR should be avoided. And the answer depends on the type of music. With some music you can dip into the "bad" region and still be OK. Some electronic music can also survive being in the "bad" region. For example, Shpongle and Infected Mushroom are mostly in the range of DR5 to DR8 and the albums sound fine for that type of music. I listen to both and never thought about complaining about the DR while listening. I don't have any Kraftwerk (need to get some) so I don't know what sort of DR would be acceptable for their music. To discount music in the "bad" range of DR7 and below would mean missing out on some interesting music that still sounds fine. To discount Shpongle and Infected Mushroom just because of their "bad" DR numbers would be missing out on some sonic treats. Both can be trippy when listening on good headphones.
     
    L5730 likes this.
  7. contium

    contium Forum Resident

    JRiver MC19 now includes dynamic range in it's audio analysis. For everything I have checked, it returned the same values as the foobar2000 plugin. Kind of interesting to be able to easily see the DR of the 40,000+ tracks I have. And it works on high res as well. MC19 also added R128 standard (kind of a broadcast industry standard of replay gain).
     
  8. Robert C

    Robert C Forum Resident

    Location:
    London, UK
    Novice question: When the meter spits out, e.g. "Official DR value: DR7", what is it 7 of?
     
  9. L5730

    L5730 Forum Resident

    It's not a measure " - out of", like a ranking like 5 out of 10. It's a calculated and measured value from the data, which is based on the crest factor of the audio. Crest factor is the difference between the bulk sound and the transients that jump out. A heavily compressed/limited track will have a small crest factor value, whilst others which haven't been so processed will have a larger value.

    Oh, and calling it a DR value (for Dynamic Range) is misleading. It should be called a CF value (for Crest Factor). The original plan and the heart behind it was in the right place - tying to encourage less crushed music.
     
  10. Robert C

    Robert C Forum Resident

    Location:
    London, UK
    Thanks. What is the value a measurement of then? Decibels?
     
  11. testikoff

    testikoff Seasoned n00b

    Here is the document explaining the TT DR measurement logic... ;)
     
  12. Stefan

    Stefan Senior Member

    Location:
    Montreal, Canada
    One big problem I have with the whole concept of DR is that a simple bass roll-off can increase DR values by 2-3 without there being any change at all in the real dynamics of a recording!

    Plus there's the whole concept of it discarding 80% of the audio information. A good example is the Adele Skyfall single mentioned above with a DR of 6. Yes, that may be valid for the main chorus sections, but if you take the song as a whole, there is a huge difference in level between the quiet verse parts and the chorus. It's hardly equivalent to the kind of squashing you find on a Metallica/Foo Fighters recording.
     
    L5730 likes this.
  13. loc4me

    loc4me New Member

    @soerin
    Earlier in this thread you posted that the DR component would write the tags to the files being analyzed? Is this correct. It doesn't appear to do that anymore it just writes a log file called foo_dr.txt

    Also is this component being actively developed? If the tags are not being written it seems like it should be simple to do so? Is there a place we can request features?
     
  14. L5730

    L5730 Forum Resident

    Soerin's site seems to have vanished some time ago.
    v1.1.1 does write two tags to the metadata, one for Album DR the other for Track DR
    http://www.hydrogenaud.io/forums/index.php?showtopic=101980

    He is a nice guy and made a couple of beta releases for me back when the project was quite fresh. v1.1.1 is the last known version and should do everything anyone needs with regard to the TT DR algorithm.

    What would be better to use now is something like ffmpeg and ebur128 scan. This should provide Integrated Loudness (average loudness in absolute LUFS) as well as Loudness Range.

    Download the latest ffmpeg, or compile it yourself with ebur128 support.
    If on Windows, extract ffmpeg binary to a folder with audio files to process (or c:\windows\system32 for system wide access).
    Try this command line syntax:
    Code:
    FOR %a IN ("*.mp3") DO (echo %a & ffmpeg -i %a -af ebur128 -f null - 2>&1 | FINDSTR /BIC:" " | FINDSTR /C:"LU")
    You'll get something like this:
    Code:
    cowboys.mp3
        I:         -13.0 LUFS
        Threshold: -23.3 LUFS
        LRA:         7.0 LU
        Threshold: -33.3 LUFS
        LRA low:   -16.7 LUFS
        LRA high:   -9.7 LUFS
    
    ffmpeg can read other audio formats, so you can just change the ("*.mp3") for ("*.*"), if you have a mixture, or just use the extension of the audio type's you are using.
    You can get more selective and use ffmpeg to actual write the info into the metadata of the file, but it involves more CLI syntax than I care to type right now.

    There has been an adjustment to the EBU-R128 algorythm that has not yet been implemented into ffmpeg or some other programs. Foobr2000 has been updated, so it seems, as have some of the loudness meter plugins such as Toneboosters Loudness.
    Although the vast majority of audio files will show the same results when scanned on any of these ebu-r128 scanners, occasionally a few have a discrepancy and are reported as being quieter than they really are - which leads me to believe that another gate was applied somewhere. I can't find the publication on the adjustment that happened sometime in 2014.
     
  15. L5730

    L5730 Forum Resident

    For example, Adele's track Skyfall from The Greatest Hits album:

    From TT DR 1.1.1
    Code:
    Statistics for: 01-Skyfall
    Number of samples: 12551448
    Duration: 4:45
    --------------------------------------------------------------------------------
    
                     Left              Right
    
    Peak Value:      0.00 dB   ---      0.00 dB 
    Avg RMS:        -9.41 dB   ---     -8.97 dB 
    DR channel:      5.99 dB   ---      5.58 dB 
    --------------------------------------------------------------------------------
    
    Official DR Value: DR6
    See the DR (Crest Factor) of only 6dB - incorrectly suggesting it's musically dynamically limited, and with very little peak headroom.

    ...and from ffmpeg ebur128 filter
    Code:
      Integrated loudness:
        I:          -8.3 LUFS
        Threshold: -19.4 LUFS
    
      Loudness range:
        LRA:        18.2 LU
        Threshold: -29.4 LUFS
        LRA low:   -23.6 LUFS
        LRA high:   -5.4 LUFS
    See the LRA of 18 LU - that's a lot of dynamic range in the music. It's still relatively loud, because of the louder parts being compressed and limited, but it's more of the truth about the actual music itself.

    ...and from Foobar2000 ReplayGain (ebur128 based)
    Code:
    Track Gain = -9.71dB
    Track Peak = 0.999969 %
    
    Bear in mind that ReplayGain is the amount of adjustment needed to make the file reach a target of -18 LUFS. So: -18 + 9.71 = -8.29 LUFS - pretty much spot on with ffmpeg.

    The peak is just shy of 100%, it's loud and needs turning down by 9dB to fit the comfortable target of what many commercial CDs were like in the early 80s.
     
  16. BrewDrinkRepeat

    BrewDrinkRepeat Forum Resident

    Location:
    Merchantville NJ
    Does anyone know if a 64-bit version of the AU plugin is in the works? I love to use it when mixing, but the latest version of Logic only supports 64-bit plugins (the DR Meter is 32).
     
    Robert C likes this.
  17. L5730

    L5730 Forum Resident

    Hi, Ian Shepherd (Production Advice) has a page about the TT DR meter and posted this video for using the 32bit plugin in 64bit Logic on Mac.


    Oh, just realised it's for a VST plugins, not AU. Uses JBridge to 'bridge' 32 to 64 Host. He mentioned AU plugins at the end 4m:12s. "...Vienna Ensemble Pro to host 32 plugins within Logic for example..."
     
  18. L5730

    L5730 Forum Resident

    If people have foobar2000 then I have created some syntax to provide a column for absolute LUFS for files scanned with ReplayGain (ebur128).

    Be sure to use fb2k replaygain scanner (ebur128) because other software may offset mono files by 3dB quieter than they are on a stereo system.

    Code:
    Installation:
    1.Right click on Fb2k columns header
    2.Choose "more..."
    3.Add new custom column
    4.Name it "LUFS"
    5.Copy and paste the entire chosen line below into the "Pattern" box
    6.OK
    7.Right click fb2k columns header
    8.Choose new LUFS column
    
    Track Gain Only Syntax:
    Only looks at Track replaygain values.
    
    ################################################
    $if([%ReplayGain_track_gain%],$puts(LUFS,$sub(-1800,$add($cut(%replaygain_track_gain%,1)$right([%ReplayGain_track_gain%],5),$mul([%ReplayGain_track_gain%],100))))$div($get(LUFS),100).$right($get(LUFS),2), )
    ################################################
    
    
    Track Gain Priority:
    Uses Track RG values if they exist, else uses Album RG values.
    
    ################################################
    $if([%ReplayGain_track_gain%],$puts(RG,%ReplayGain_Track_gain%),$if([%ReplayGain_album_gain%],$puts(RG,%ReplayGain_album_gain%)))$puts(LUFS,$sub(-1800,$add($cut($get(RG),1)$right($get(RG),5),$mul($get(RG),100))))$if([%ReplayGain_track_gain%],$div($get(LUFS),100).$right($get(LUFS),2)t,$if([%ReplayGain_album_gain%],$div($get(LUFS),100).$right($get(LUFS),2)a))
    ################################################
    
    Album Gain Priority:
    Uses Album RG values if they exist, else uses Album RG values.
    
    ################################################
    $if([%ReplayGain_album_gain%],$puts(RG,%ReplayGain_album_gain%),$if([%ReplayGain_track_gain%],$puts(RG,%ReplayGain_track_gain%)))$puts(LUFS,$sub(-1800,$add($cut($get(RG),1)$right($get(RG),5),$mul($get(RG),100))))$if([%ReplayGain_album_gain%],$div($get(LUFS),100).$right($get(LUFS),2)a,$if([%ReplayGain_track_gain%],$div($get(LUFS),100).$right($get(LUFS),2)t))
    ################################################
    
     
  19. JamesLord

    JamesLord Forum Resident

    Location:
    UK
    This is a very valuable & illuminating discussion - thanks to all who have contributed.

    I am a non mathematician, but if I am understanding correctly the meters will compare the loudest 20% of the music to the absolute peaks to determine the overall DR value. Is this correct?

    This would explain some perplexing readings I have got from my own music. For example a long track that was extremely dynamic but did have an intentionally very loud/driven section came out with a low DR reading, despite have extensive sections that were very quiet. The Reading baffled me, but if the logic above is correct then all is explained.
     
  20. JamesLord

    JamesLord Forum Resident

    Location:
    UK
    There is an irony here too. In some respects, the more varied/dynamic the music is the less accurate the meter readings will be (due to discarding 80% of the available data).

    That, and the fact that it doesn't work on needledrops, really reinforces the fact that, as a brave few on here have noted, these numbers are interesting and useful, but are really neither a definitive statement of a recording's dynamics and certainly can't be used as a synonym for 'sound quality'.

    Obvious to many I am sure - but seems to derail a lot of otherwise interesting discussions on here. Imho of course :)
     
    Stefan likes this.
  21. testikoff

    testikoff Seasoned n00b

    IMV, TT DR algorithm was specifically designed to look at the loudest portion of analyzed material & calculate the ratio between the second highest peak & the RMS. If the ratio is 4 (or higher), then material is regarded good (or better) dynamically, simple as that. TT DR value does not represent true dynamic range of the recording (which in case of RBCD audio may potentially be close to 96dB), it just gives an idea whether the loudest parts of analyzed audio are indeed alive (or not so much).
     
    thrivingonariff likes this.
  22. JamesLord

    JamesLord Forum Resident

    Location:
    UK
    Thanks. That's the distinction that I think gets missed. Yes, music with a lower DR rating has less dynamics between the loudest 20% and the absolute peak. But that doesn't mean there are no dynamics in the song.

    It's when I see discussions on here that go:
    A - how does x sound?
    B - its dr is n so it's great/crappy

    Or the even more spurious 'anything with a DR below 10 is unlistenable.'

    That it seems like these numbers are being given a level of significance way beyond what was intended.

    I have plenty of music with DR 6/7 that sounds fine to me, and a great many recordings with DR 12+ that are sonically anaemic and crappy.

    But, as noted above, it is very useful to have some clarification on how the meters work.
     
    thrivingonariff and Stefan like this.
Thread Status:
Not open for further replies.

Share This Page

molar-endocrine