Print

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

[Fleckenzwerg]

Hi,

I like all my tracks played at 100% volume. The important thing is, it should be always the maximum volume for a tack, no overall "the same" volume for different tracks. I do not get how to do that with volume leveling. Is there some hidden function that simply normalizes every track to 100%?

Thank you!

[6233638]

What you describe is Peak Normalization, which will not actually "normalize" the volume levels of your music.

The more dynamic range a track has, the quieter it will be, so it biases towards highly compressed audio, which sounds louder when you use peak normalization. This is the reason we have such compressed music in the first place, because that's how Radio and TV used to work.
You will also be susceptible to clipping due to intersample overs if you do this - which can approach 12dB in rare cases. In reality, you need far less headroom though - the 3.5dB used in the DAC2-HGC is probably enough for "real" audio, and if you are using ReplayGain you probably don't have to worry about it.
Clipping protection may already handle it though - I'm not sure. Matt is probably the person to ask about that.

ReplayGain - which is what the Volume Levelling DSP utilizes - is based on analysis of your music, and normalizes around what it believes to be the "average" volume level for a track, so that tracks with a lot of dynamic range, and tracks with very little dynamic range, should all play at roughly the same volume. It's largely successful, but I would like to see a move towards R128, which should do a better job - maybe that is something for MC19 though.

For video, I believe Adaptive Volume set to Light does this, but there's no option for music playback.

[Fleckenzwerg]

What I actually want is, if the maximum peak of a track is for example 98%, the volume of the track should be increased so that this peak is now 100%. And vice versa if a maximum peak is 110%, the volume should be reduced so that this peak is 100%. Nothing should be compressed or cut off. I also do not want that a highly compressed track sounds equally loud as a low compressed.

How can I do this?

[6233638]

As I said, what you want is Peak Normalization, and I don't think Media Center offers that.

I can't understand why you would want to use peak normalization though - that is going to cause the volume of music to vary wildly between tracks.

[Fleckenzwerg]

Peak normalisazion will help to use the dac with best possible THD+N. Because THD+N increases the lower the output of the DAC is, records with low volume will sound better if they are amplified to 0 dB in the digital domain. If for example a whole cd has a low volume, you still can listen to it without big volume differences. Or one could use peak normalization for a complete album. This way the volume between the tracks is the same.

[6233638]

Peak normalisazion will help to use the dac with best possible THD+N. Because THD+N increases the lower the output of the DAC is, records with low volume will sound better if they are amplified to 0 dB in the digital domain.

Well I can't argue with that - it's true. But clipping from intersample overs is probably much worse distortion than you're going to see from lowering the volume 3dB or so. And most DACs shouldn't have a problem with using ReplayGain.

If for example a whole cd has a low volume, you still can listen to it without big volume differences. Or one could use peak normalization for a complete album. This way the volume between the tracks is the same.

That won't give you the same volume for each track - unless they had peak normalization applied at the mastering stage.

However, I think using album-based levelling, and setting the Volume Levelling Adjustment to Automatic based on current playlist is close to the results you're trying to achieve.

[InflatableMouse]

Maybe this is overly simplistic, but what if you push MC's volume so that it always clips, but use clipping protection to keep it at 100% all the time?

Won't that effectively be what you want?

[6233638]

Maybe this is overly simplistic, but what if you push MC's volume so that it always clips, but use clipping protection to keep it at 100% all the time?

Won't that effectively be what you want?

I tried that out and it seemed like it was compressing the volume, because clipping protection seems to act in realtime, rather than analyzing the track in advance.

[whatsup]

What I actually want is, if the maximum peak of a track is for example 98%, the volume of the track should be increased so that this peak is now 100%. And vice versa if a maximum peak is 110%, the volume should be reduced so that this peak is 100%. Nothing should be compressed or cut off. I also do not want that a highly compressed track sounds equally loud as a low compressed.

How can I do this?

http://mp3gain.sourceforge.net/index.php

[Fleckenzwerg]

Maybe this is overly simplistic, but what if you push MC's volume so that it always clips, but use clipping protection to keep it at 100% all the time?

Won't that effectively be what you want?

No, because it will truncate the peaks.

However, I think using album-based levelling, and setting the Volume Levelling Adjustment to Automatic based on current playlist is close to the results you're trying to achieve.

I tried that. There is a plus 3,4 db increase for the current playlist (one CD from one artist) and a corrective gain for each track. You are right, this is almost what I want. The loudest track on the cd has a peak of 91%. With volume leveling this track will only gain 0.8 dB but a lot of other tracks gain 3 to 10 dB (that is better than nothing). Now, knowing how to analyse a playlist, I can tweak it to get what I exactly want:
The highest peak in the playlist is 91%. This is the same as -4.5 dB. Now I use volume adjustment in the parametric eqaulizer to add 4,5 dB of gain. That is it. All I want. Thanks guys for your help!

Dear developers: It would be very nice if this quite simple adjustment would be available by automation. I think it is very usefull especially if one is after the best possible sound quality from digital music files.

[Fleckenzwerg]

PS.: Maybe an adjustemnt to -0.1 dB would be better if the DAC would clip at 0 dB. But I do not know where the limit for clipping normally lies.

[6233638]

I'm sure that Peak Normalization would be an easy thing to implement, but I still don't think you actually want it - unless you are planning on adjusting the volume control on a per-track basis. (and I assume that's an analog volume control in whatever hardware you're using?)

Most audio is still well within the confines of 16-bit - including audio stored in 24-bit files. 96dB is a lot of dynamic range.
I don't think I've seen an adjustment beyond -18dB or so in ReplayGain - which is 3-bits. If you're using a 24-bit DAC, you probably have the headroom for that adjustment without any impact on sound quality at all.

PS.: Maybe an adjustemnt to -0.1 dB would be better if the DAC would clip at 0 dB. But I do not know where the limit for clipping normally lies.

Benchmark Media seem to suggest that it can be in the range of 2-3dB with lossless files - and could be higher if you are using lossy files.

[Matt]

I'm trying to understand this request, so help me along.

Are you playing in a power starved system where you're looking for ways to squeeze extra volume out?

As an example, on my home system, a comfortable listening level is about -30dB of internal volume.  This is a DAC hooked directly to a power amplifier.

If I peak normalized my songs, I might need -33dB for the same volume, so I'd get +3dB "for free."  However, I already have 30dB more than I can really use so the 3dB isn't particularly helpful.

Instead, I like the Replay Gain system that makes the volumes between songs sound about the same.  Since I have the volume headroom, it doesn't change the sound quality at all.

As 6233638 said, your DAC probably uses 24-bit, so turning a 16-bit down isn't actually removing information.  Even if your source is 24-bit, the last many bits are almost surely noise so you're not really removing "real" information with volume on those files either.

I'm not opposed to on-the-fly peak normalization (it's certainly better than baking it into the files!), but I'd like to understand the technical argument for it better.