Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[JustinM]

I believe you could improve sound quality, if replay gain raised the volume of quiet songs, instead of reducing the volume of louder songs.
eg:  I have 24bit recorded songs  and  a 24bit external dac ,, my 24bit songs recorded at 100% are being reduced in gain by 18db,  then are outputed to my 24bit dac..  Didn't I just lose
18db of dynamic range ?  .. (3 bits)
Some people will also be running line level outs to amps/recievers.... at lower line levels their system will also be more susceptible to added noise.


I know alittle knowledge can be dangerous Am i missing something? Is there an advantage to lowering the gains ?

 

[Matt]

You can use the 'Automatic' adjustment mode in DSP Studio > Volume leveling to apply the safest gain for the current playlist that maintains similar volume and will not clip (based on Peak Level from Audio Analysis).

[JustinM]

Hi,Thx Matt.

But again it reduces the volume for tracks that are already below 100% peak level.

[Matt]

Automatic mode uses the highest output level that:
1) Maintains equivalent perceived volume (Replay Gain)
AND
2) Does not clip

The only way to provide a higher level than the existing Automatic mode is to sacrifice #1 or #2, which would either defeat the purpose (#1) or sound bad (#2).

Some more technical details here:
http://wiki.jriver.com/index.php/Technical_Information_on_Replay_Gain_Adjustment

[Alex B]

In addition to what Matt said, the typical loud compressed tracks have a very narrow dynamic range. You can attenuate them by several tens of dBs without sacrificing any quality. Your 24-bit output has plenty of headroom. For example if the total attenuation is 20 dB, you still have over 120 dB of available dynamic range in the digital domain.

[JustinM]

... the typical loud compressed tracks have a very narrow dynamic range.

Yeah your right .... you still get the resolution...

Automatic mode uses the highest output level that:
1) Maintains equivalent perceived volume (Replay Gain)
AND
2) Does not clip

The only way to provide a higher level than the existing Automatic mode is to sacrifice #1....

ok Matt, I'll read up, It just seems to me that if peak level of a song is 23%,   Replay Gain shouldn't be reducing the gain another 2.89 db...  I assumed you would reduce any songs above 100% and raise any songs below...  

cheers  

[JustinM]

Hi Matt :  If your interested:



I don't think the Replay Gain feature assumes there will be ANY improperly ripped songs....  
and peak levels over 100% are not accomadated/ corrected .

[seiferflo]

In addition to what Matt said, the typical loud compressed tracks have a very narrow dynamic range. You can attenuate them by several tens of dBs without sacrificing any quality. Your 24-bit output has plenty of headroom. For example if the total attenuation is 20 dB, you still have over 120 dB of available dynamic range in the digital domain.

HI guys,
Sorry to just create an account for that but I was deeply into some research regarding Replaygain and I saw this great answer from Alex B which is still raising questions in my head. I hope I still could get some info.

Actually I'm sharing the point of JustinM regarding the loss of dynamic.
I did some tests on my music library. I have from very quiet classical music to some 0db peak electro, and I do understand the point of ReplayGain to offer a more "volume conformity" listening experience.
However, when using ReplayGain (83dB) I did decrease around 0 to 20db on probably 75% of my library just to have the same level of the other 25%. From this, only around 5% of it has 24bit resolution.

So when I look, sorry hear at my 95% of 16bit music which got their gain decreased by up to 20dB, it hurts. There is not plenty of headroom left when we take it from 90dB of dynamic.
When I push my audio card to 100% and listen to my music with RP activated, (even some high dynamic classical), it is quieter. So I am loosing some dynamic.
Is there a way to arrange that, I read there are some settings we can play with, but didn't get it right?

This is the thing I'm not really sure about, please do correct me if I'm wrong.
We do have a loss of dynamic using ReplayGain, right?
Or is it something that either we can't hear because we are talking about digital sound... ?

I'd love more if you could develop your answer more please.
Thanks in advance for your help.

[6233638]

You can't do digital volume adjustments without losing dynamic range, because it is impossible to go above 0dB without damaging audio quality (clipping) so you have to make negative adjustments.

I would suggest not using ReplayGain at all if you are only outputting 16-bit from your PC. Losing 3-bits of dynamic range is probably going to be audible.

However, if you are outputting 24-bit or greater from your PC, you instantly have 8-bits of adjustment before you lose anything - that's 48dB.

Looking through my library, the worst I have is -17dB for Album Gain, and -19dB for Track Gain. But I only use Album Gain to keep volume relative between albums and maintain track dynamics. So that still leaves me with 21-bits being output, which is still well above the 16-bit we have for CDs.

[mwillems]

You can't do digital volume adjustments without losing dynamic range, because it is impossible to go above 0dB without damaging audio quality (clipping) so you have to make negative adjustments.

I would suggest not using ReplayGain at all if you are only outputting 16-bit from your PC. Losing 3-bits of dynamic range is probably going to be audible.

However, if you are outputting 24-bit or greater from your PC, you instantly have 8-bits of adjustment before you lose anything - that's 48dB.

Looking through my library, the worst I have is -17dB for Album Gain, and -19dB for Track Gain. But I only use Album Gain to keep volume relative between albums and maintain track dynamics. So that still leaves me with 21-bits being output, which is still well above the 16-bit we have for CDs.

I'm curious about this; what you're saying makes mathematical sense, but I wonder how audible it is.  To be clear I'm genuinely curious and want to make sure I'm not missing something (recognizing that the -18 dB replaygain is an extreme example, and that most music won't receive the full 3 bits of attenuation)

A very quiet room has a 30 dB noise floor (a treated studio might get down to 20, a typical room in a house with ventilation and electronics running is more like 40 or even 50).  The recommended system calibration is 83 Db at -20 DBs pink noise.  In that arrangement it seems like the loudest sound coming out of your system would be around 103 dB.  So it seems like even 78 dB of dynamic range would cover what's possible to hear above the "quiet room" noise floor unless you're in a recording studio or have your system calibrated so that your peaks exceed 108 dB (which, to be fair, I'm sure is the case for many folks). 

I use 24 bit output myself because it's free and I can, but do you really think in most applications the difference between 78 dB of range and 96 dB of range will even be audible? I just want to make sure I haven't misunderstood the point somehow (I'm a little dense occasionally  )

[6233638]

You're probably right. Under normal listening conditions, even using ReplayGain when outputting 16-bit is probably fine, especially if you use the dithered 16-bit output. And I seriously doubt anything is using close to 60dB of dynamic range to begin with. (assuming 60dB of dynamic range played at -18dB)

But just about anything will output 24-bit these days, and there's no reason not to.

From doing some basic testing, I struggle to hear a -78dB 1kHz tone above the background noise level in the room at my normal listening level - though I do listen at much lower levels than most.

[mwillems]

But just about anything will output 24-bit these days, and there's no reason not to.

From doing some basic testing, I struggle to hear a -78dB 1kHz tone above the background noise level in the room at my normal listening level - though I do listen at much lower levels than most.

That's basically where I was on both points: I can't really hear much below -75 dB in my room, but 24 bit output is easy to do with most hardware, so why not, just in case?

[kstuart]

mwillems:
"I use 24 bit output myself because it's free and I can, but do you really think in most applications the difference between 78 dB of range and 96 dB of range will even be audible? I just want to make sure I haven't misunderstood the point somehow..."

So, here is a professional mastering engineer who has done hundreds of major label CDs about why it is more than just the difference between two large amounts of dynamic range, and why the noise floor is not necessarily relevant:

" The components of a signal that are below full scale are encoded using less than the full word length.
With real music, this means a great deal of what is musically as well as sonically significant information is encoded using less than the full word length.

This means the same signal in a 24-bit system is going to be encoded with 8 bits more than in a 16-bit system.

A signal that is 40 dB down in a 16-bit system gets encoded using 10-bits. In a 24-bit system, it gets encoded using 18-bits. To my ears, 10-bits is pretty cheesy sounding. Lower level signals, encoded with even fewer bits, will sound cheesier still."

He is talking about using 24-bit rather than 16-bit, but the same argument applies here.

[mwillems]

mwillems:
"I use 24 bit output myself because it's free and I can, but do you really think in most applications the difference between 78 dB of range and 96 dB of range will even be audible? I just want to make sure I haven't misunderstood the point somehow..."

So, here is a professional mastering engineer who has done hundreds of major label CDs about why it is more than just the difference between two large amounts of dynamic range, and why the noise floor is not necessarily relevant:

" The components of a signal that are below full scale are encoded using less than the full word length.
With real music, this means a great deal of what is musically as well as sonically significant information is encoded using less than the full word length.

This means the same signal in a 24-bit system is going to be encoded with 8 bits more than in a 16-bit system.

A signal that is 40 dB down in a 16-bit system gets encoded using 10-bits. In a 24-bit system, it gets encoded using 18-bits. To my ears, 10-bits is pretty cheesy sounding. Lower level signals, encoded with even fewer bits, will sound cheesier still."

He is talking about using 24-bit rather than 16-bit, but the same argument applies here.

I'm not sure I understand. The quote you provide seems to be describing encoding a given signal as part of recording, not playback.  There's a clear and compelling argument for using 24 bits when making a recording (and for playing back 24 bit recordings in 24 bit), but I don't understand how that applies to this.  

My understanding is that when JRiver outputs a 16bit signal as 24 bit it just pads the signal with zeros, it doesn't attempt to "re-dynamicize" the recording.  Those zeros give you some extra pad when adjusting the volume down, but they don't provide extra "resolution" in any meaningful sense.    So you still have the identical 16 bit signal.

Likewise when JRiver adjusts the volume of a 16 bit recording down I don't understand it to be performing dynamic compression in real time on the signal; if it were performing dynamic compression as part of volume control your point would be well taken. My understanding of how the internal volume works is that it essentially just shifts the signal down, so the quietest bits just "fall off" the bottom (to the extent there are any bits encoded 80 dB down to begin with). In that case it's not clear how any given sound can sound more or less "cheezy" than it otherwise would because it's being played back with the same number of bits as before, unless some of those bits are very, very quiet (and probably below the noise floor).  The signal is the same, but truncated in the quietest parts.  The only missing parts are the parts that are below the noise floor.

I'm left scratching my head a bit, because I'm not sure how decreasing the volume could do what you're describing unless it also performed dynamic compression or otherwise monkeyed with the dynamic range of the recording (which, if you have adaptive volume set below medium, it should not be doing).  

[seiferflo]

Thank you all for your answers, it helps understanding more what's happening.
However I just want to go back to my simple question regarding ReplayGain.

Whatever happens mathematically when using 16bit ou 24 output/file it does not change the fact that applying Replaygain is decreasing the dynamic of the listening volume (I don't know if I can say of the file).

If I want to listen a song at value 50 on my Amplifier, well it's now quieter because of Replaygain. So if I want to have the same listening volume, I have to increase to 60. Of course the next song will be played at the same level thanks to Replaygain so I don't really need to change the volume again, but at then end, I decreased the listening quality of my song. Because Replaygain decreased the volume of song to align to a norm and to other files, my song requires more hardware power to get played at the same level.
Seeing this on a smaller configuration: on a laptop headphone output, when I am at the maximum output level of 100, I can't go beyond that. At then end, in my hears, it sounds quieter, so my dynamic of listening has been reduced.
As "6233638" mentioned earlier: "You can't do digital volume adjustments without losing dynamic range".

Changing the output between 16bit to 24bit does not really matter to me, the result on a basic scale is the same for me.
Please do give me your feedback on an listener experience level (sorry for my lack of technical knowledge).

[6233638]

If I want to listen a song at value 50 on my Amplifier, well it's now quieter because of Replaygain. So if I want to have the same listening volume, I have to increase to 60. Of course the next song will be played at the same level thanks to Replaygain so I don't really need to change the volume again, but at then end, I decreased the listening quality of my song.

Do you actually hear an increased level of hiss/background noise from the speakers at your listening position when you turn them up to 60? If not, it means the noise level is still below the noise floor in your room. (which is likely)

Seeing this on a smaller configuration: on a laptop headphone output, when I am at the maximum output level of 100, I can't go beyond that. At then end, in my hears, it sounds quieter, so my dynamic of listening has been reduced.

I would be more concerned about your hearing if you are listening to headphones at maximum volume.

As "6233638" mentioned earlier: "You can't do digital volume adjustments without losing dynamic range".

To be clear, what I meant by that was that if you digitally reduce the volume of a 16-bit file, in the 16-bit domain, you are losing dynamic range. Whether that affects your music playback depends on the adjustment.
Most music will have less than 30dB of dynamic range, and I seriously doubt anything will approach or even exceed 60dB. What this means is that you probably have plenty of room for volume adjustment even in 16-bit, because the lower bits are just silence/noise anyway, rather than music.

If you have a 16-bit file and output at 24-bit, then you have 8-bit (48dB) of volume adjustment before there is any change to the original audio. And even then, you probably still have room to reduce the volume even more.

24-bit audio tracks are basically all noise/silence beyond 16-bit anyway, so reducing the volume of a 24-bit file, while outputting 24-bit isn't going to harm audio quality.
24-bit makes sense when recording/mastering content, but there's no need for it with the final file mastered for playback. It probably only makes sense to have 24-bit files if you are using a player that outputs 16-bit when you send it a 16-bit file, and you are reducing volume in the digital domain.

Changing the output between 16bit to 24bit does not really matter to me, the result on a basic scale is the same for me.

Leave it on 24-bit (dithered) or higher if your hardware supports it.

[kstuart]

24-bit audio tracks are basically all noise/silence beyond 16-bit anyway, so reducing the volume of a 24-bit file, while outputting 24-bit isn't going to harm audio quality.
24-bit makes sense when recording/mastering content, but there's no need for it with the final file mastered for playback. It probably only makes sense to have 24-bit files if you are using a player that outputs 16-bit when you send it a 16-bit file, and you are reducing volume in the digital domain.
Leave it on 24-bit (dithered) or higher if your hardware supports it.

You just seem to be ignoring the counter-arguments.   Here again is the central part:

" A signal that is 40 dB down in a 16-bit system gets encoded using 10-bits "

The point is that for a low-level signal, you are not getting 16-bits.   You only get the full 16-bits for the loudest signals.

So, 16-bits is sufficient, but most of the signals aren't getting 16 bits of resolution.

[6233638]

Bit-depth is just volume level (dynamic range) with PCM signals. You aren't losing anything. My mistake, you were talking about quantization when you say "resolution" - which is, in all likelihood, inaudible.

[kstuart]

Bit-depth is just volume level (dynamic range) with PCM signals. You aren't losing anything.

Both of those statements are false.  Read the explanation above.