Help - Search - Members - Calendar
Full Version: Questions on Loudness and Playback
Hydrogenaudio Forums > Hydrogenaudio Forum > General Audio
Bourne
CODE

Scale (volume range):

silence                                              very loud

Looking at the dynamic range CD is capable of:

      -96dB FS                                        0dB FS
       |----------Total possible CD Range--------------|
    |-----------------Replay Gained-----------------|
       |--------------16-bit playback------------------|

    |--| < lost!                           empty! > |--|



Looking at the dynamic range of typical recordings:

                                  |--dynamic music--|
                                         |--top 40--|
    |--| < lost = irrelevant


1) How would that graph be like if you apply the 24-bit decoder and 16-bit source?

2) If a 24-bit source comes in the way, then will definitely lose half data if played with 16-bit decoder and if played with 24-bit decoder will happen just like your first graph...

3) This 16 or 24 bit decoding applies only to soundcards? Not to DVD-Players CD/MP3 features?

4) If a very compressed loud music is made quieter, then the peaks get also quiet and then the music may be "wimpy" and "poor" because where there's quiet, there can't be no loud too. Is is better to listen to compressed music the way it is (loud=more punches) or is it better to make it quiet (wimpy=weak punchs), or should we just make a "little" quieter but not to the point where it gets wimpy...?

Ref: http://br.youtube.com/watch?v=3Gmex_4hreQ

Added:

5) Is there any way to determine compatibiliy loudness between songs from different albums, considering the fact that they would not have just the same peak, but also the bass and treble intensities...

6) Would someone explain why 89.0dB is equivalent to 0.0dB, what is the pratical answer to this?
Lyx
Just plain and simply use replaygain and use the windows-mixer/analog volume knob to set your prefered loudness and stop being paranoid - or if you cannot stop it, just ignore replaygain and psychoacoustic lossy compression - also better dont use any EQ at all, since that changes the signal.

This is the answer to all your questions.

So in short, here is the "manual" on how to use replaygain:

1. Scan files with an RG-scanner.
2. Choose albumgain or trackgain for playback.
3. Adjust volume knob on your speakers/amp.... or via windows mixer.
4. Enjoy the music.

I know you'd prefer if it were more complicated and needed more "tweaking" - however, "unfortunatelly", Replaygain works indeed that simple in practice.
gasmann
I can't fully agree with the graph... why should replaygained audio not be able to use the full scale of volume range? a replaygain value of 0 would speak against that. Decently mastered recordings do often have a replaygain value of 0 or higher (see Star Wars Episode 2 OST, Titanic OST, Bach essentials, ...). Thus this 'Replay Gained'-bar really isn't exact for all audio. Some audio may not shift the bar at all (rg 0), while some might even shift it in the opposite direction (rg > 0). I think that's a big mistake you start off with.

I agree that replaygained clippressed tracks sound worse than replaygained dynamic tracks. But I would still not want them to be louder, because that defeats the sense of replaygain. I want volume to be constant and so am happy how rg works.

But anyway... I second what Lyx says that you only make your life more complicated than it really is.
NogginJ
"1) How would that graph be like if you apply the 24-bit decoder and 16-bit source?"

don't do that without dithering (you'll totally screw up the quiet parts otherwise).

"6) Why 89.0dB is 0.0dB? (According to foobar2000 ReplayGain setup)"

dB is relative, not absolute. so 89 could equal 00 in different circumstances.
S-12
I prefer WaveGain over the use of ReplayGain tags, because, although it uses a lossy process, it provides perfect portability. All of my audio players portray the effect of WaveGain, while only those which are "ReplayGain compatible" respond to RG settings in an encoded audio file.

And from some individual listening tests here several years ago, no one was able to perceptually differentiate any WaveGain loss, so I've always felt safe using it even on my most important archived music.
benski
You seem to be confusing a few related but not identical concepts.

The dynamics of a song are usually calculated as the difference between the RMS level and the peak level. An absolute volume change will not change this.
"punch" = max-avg

The dynamic range of a playback system is measured as the difference between the softest possible signal and the loudest possible signal. Since Replay Gain is an absolute volume change, you won't see any difference here.
dynamic range = max-min

Truncation due to loss of precision adds noise and effects dynamic range only as a side-effect. Dithering can retain most of the dynamic range due to precision loss, at the expense of additional noise. Applying Replay Gain to 16-bit source material will require more than 16 bits to store fully-precise results.
Martel
QUOTE(Bourne @ Feb 18 2008, 10:22) *

1) How would that graph be like if you apply the 24-bit decoder and 16-bit source?

2) If a 24-bit source comes in the way, then will definitely lose half data if played with 16-bit decoder and if played with 24-bit decoder will happen just like your first graph...

3) This 16 or 24 bit decoding applies only to soundcards? Not to DVD-Players CD/MP3 features?

4) If a very compressed loud music is made quieter, then the peaks get also quiet and then the music may be "wimpy" and "poor" because where there's quiet, there can't be no loud too. Is is better to listen to compressed music the way it is (loud=more punches) or is it better to make it quiet (wimpy=weak punchs), or should we just make a "little" quieter but not to the point where it gets wimpy...?

I saw this http://br.youtube.com/watch?v=3Gmex_4hreQ

I may not be correct, but as I understand it:
1) The decoded stuff (MPEG and equivalents) is usually floating point (no absolute range boundaries, there ARE some boundaries but they hardly apply here), then replaygain is applied (as floating point multiply, no rounding, minimum precission loss), then the floating point values get converted into 16-bit or 24-bit range (and rounded to integers). I think that decoding to 24-bit might help as you don't lose the resolution in quiet parts. If you had only 16-bit scaled-down file and turned up the volume, the quiet parts will have few effective bits (like 6 or so) while with 24-bit decoded file there will be 14.
Either way, DACs are not usually able to precisely decode the last 4 bits of 24-bit samples (there is some electric noise) so you have effectively 20-22 bits upon conversion to analog.
3) some MP3 players do it even worse - they don't even have a regular (resistor) volume control and the digital signal is scaled down before DAC in order to reduce output volume, which makes it utter crap, cutting the signal to only few effective bits of resolution...
4) The problem is that all the lossy codecs change the waveform shape (mostly due to frequency filtering). The peaks that were simply cut-off (clipped) in the original wave usually do not appear as clipped after decoding of lossy file and they do not fit in the original (16-bit) range. Clipping them again may produce different and even more distorted sound than the original so the replaygain is applied to prevent additional distortion.
Bourne
thanks Martel, I think you figured how to write a decent reply
cabbagerat
QUOTE(Bourne @ Feb 18 2008, 10:22) *

5) Is there any way to determine compatibiliy loudness between songs from different albums, considering the fact that they would not have just the same peak, but also the bass and treble intensities...
This is exactly what replaygain does. It calculates how loud the music is going to sound (irrespective of the peak level). The details can be found at http://replaygain.hydrogenaudio.org/calculating_rg.html.

Even with replaygain, some songs (between CDs) are just meant to sound softer than others. Bob Katz uses the example of a string quartet and a whole orchestra - the quartet is just much quieter, and shouldn't play right after the orchestra at the same perceptual level. This is where good mastering level practices come in - something that replaygain (for all it's advantages) is not an alternative for.
Bourne
RG is just a volume regulator right?

I am talking about BASS intensity in on song, and weak bass on another. That is... they come from different mastering, in that case, it's impossible to make the weak-bass stronger unless you edit the song with cooledit and make adjustments to gain more bass... if I were to put several tracks from different albums, how would one find how "bassy" is a song compared to the other...
Northpack
QUOTE(Bourne @ Feb 20 2008, 06:49) *

RG is just a volume regulator right?

I am talking about BASS intensity in on song, and weak bass on another. That is... they come from different mastering, in that case, it's impossible to make the weak-bass stronger unless you edit the song with cooledit and make adjustments to gain more bass... if I were to put several tracks from different albums, how would one find how "bassy" is a song compared to the other...

That's not what Replaygain is all about. As you said, it's a "volume regulator", not an equalizer! You may want to have some kind of equalizing profiles attached to the files, applied on replay just like Replaygain. I think Ipod's track-based equalizing works this way, but it just uses some silly presets like "rock", "pop" etc.

However, there is no way to automatically compute such a profile, because it's a subjective impression, if youl feel the bass of a track/album is to meek. If you would spread different files to equivalent loudness on each frequency, the results would sound horrible (imagine you'd try to make a bass out of a flute...)
Bourne
hmm it all boils down to the mastering...
retro83
QUOTE(Bourne @ Feb 20 2008, 13:43) *

hmm it all boils down to the mastering...


So is what you envisage something like replaygain, except with seperate processing for bass/mids/highs ?

I imagine that it would make the music sound a bit ropey. tongue.gif
Lyx
QUOTE(cabbagerat @ Feb 20 2008, 06:39) *

Even with replaygain, some songs (between CDs) are just meant to sound softer than others. Bob Katz uses the example of a string quartet and a whole orchestra - the quartet is just much quieter, and shouldn't play right after the orchestra at the same perceptual level. This is where good mastering level practices come in - something that replaygain (for all it's advantages) is not an alternative for.

Which is the main weakness of Trackgain. It has no clue about what is "meant" to sound softer and amplifies dynamic balads too much. I consider trackgain to be rather useless, but apparently, some people like it. Perhaps they listen to different music than i do.
2Bdecided
QUOTE(Bourne @ Feb 18 2008, 18:22) *

1) How would that graph be like if you apply the 24-bit decoder and 16-bit source?

CODE
Scale (volume range):

silence                                                                       very loud

Looking at the dynamic range CD is capable of:

     -144 dB FS                -96dB FS                                        0dB FS
                                |----------Total possible CD Range--------------|
                             |-----------------Replay Gained-----------------|
                                |--------------16-bit playback------------------|
                        
                             |--| < lost!                           empty! > |--|

                             |-----------------Replay Gained-----------------|
       |-----------------------24-bit playback----------------------------------|

= "nothing" lost (in the digital domain at least)



The limit in the 24-bit playback is the noise floor of the converter, amplifier etc, not the digital data. It could lie anywhere, from about -120dB to about -50dB: so depending on the quality of your equipment, you might keep all the original signal above your system's noise floor, or you might not: that's true whether you use ReplayGain or not.

QUOTE
2) If a 24-bit source comes in the way, then will definitely lose half data if played with 16-bit decoder and if played with 24-bit decoder will happen just like your first graph...
Yes, but this is all over simplistic. You are assuming the 16 or 24 bit sources have a noise floor at or below the last bit. This is never the case for 24-bit audio, and often untrue for 16-bit audio. You are also assuming the converters and amplifiers etc have a noise floor at or below the last bit. This is never the case for 24-bit audio, and sometimes untrue for 16-bit audio.

QUOTE
3) This 16 or 24 bit decoding applies only to soundcards? Not to DVD-Players CD/MP3 features?
They all decode to something! I would expect 16-bit decoding to be common, and 24-bit decoding to be uncommon (even on devices with 24-bit DACs).

QUOTE
4) If a very compressed loud music is made quieter, then the peaks get also quiet and then the music may be "wimpy" and "poor" because where there's quiet, there can't be no loud too. Is is better to listen to compressed music the way it is (loud=more punches) or is it better to make it quiet (wimpy=weak punchs), or should we just make a "little" quieter but not to the point where it gets wimpy...?
If ReplayGain (or anything else) reduce the volume of something in the digital domain, and if you then increase the volume (back to where it started) with your analogue volume control, there will be no difference in the sound if:

(a) the noise floor of the original recording stays well above the noise floor of the converters, amplifiers etc
(b) the converters and amplifiers are linear over the range in use

The only audible difference can come from bad converters, amplifiers etc. Most commonly, amplifiers in portable devices are used at the limit of their range, and have a different characteristic depending on input and output level. In this case, the sound can be dramatically different.

QUOTE
5) Is there any way to determine compatibiliy loudness between songs from different albums, considering the fact that they would not have just the same peak, but also the bass and treble intensities...
Like an EQ matcher? Someone asked if this was possible. Anything is possible. It's fraught with even more problems than ReplayGain. You can solve both EQ and loudness problems by sending everything through aggressive multi-band compression (as radio stations do), but that will change the sound of everything to your chosen "signature" sound.

QUOTE
6) Would someone explain why 89.0dB is equivalent to 0.0dB, what is the practical answer to this?
It's not equivalent. dBs are ratios - any dB figure is how much louder or quieter something is compared with something else. There is always an assumed reference, e.g. the reference for dB SPL = a particular variation in air pressure (which roughly corresponds to the threshold of hearing at 2kHz); the reference for dB FS = a full scale sine wave. The ReplayGain use is explained here:

http://replaygain.hydrogenaudio.org/calibration.html

though the 83dB has been replaced by 83+6=89dB.

I think Lyx is right though - you should stop worrying!

Cheers,
David.
Bourne
Yes, will stop worrying because you explained it well. Hell, you know way much more than other people who assume they know enough.
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.
Invision Power Board © 2001-2008 Invision Power Services, Inc.