IPB

Welcome Guest ( Log In | Register )

2 Pages V   1 2 >  
Reply to this topicStart new topic
Does Normalization decrease the sound quality of an audio file?
meurglys0
post Jun 15 2010, 22:03
Post #1





Group: Members
Posts: 16
Joined: 14-January 09
Member No.: 65534



I have searched the forum to see if what I'm about to ask now has already been discussed, but even though I found many threads on this topic, none of them discussed the matter at the level of my ignorance of these technical issues, so I'm going to
ask very simply:

I downloaded a great sounding vinyl rip (in flac format) of an album I love and it sounds much better than other versions of the album I have. But the volume level of this rip is lower than the necessary level for me to be able to listen to it via my digital audio player (zune 120) in public transportation. (I use lossless formats on the audio player, btw.) Anyway, I know that I could normalize the files to -1 db and it makes the album louder. However the dilemma is that I prefer this vinyl rip over the many versions of this album for it sounds better, but if because of normalization the sound is worsened, the cause will be lost.

So my questions are:

1. Does normalization of an audio file in wav format to -1 db result in decrease of sound quality?

2. Should I normalize tracks seperately or all of them at once?

3. There are options for limiting etc. if I choose to select all to normalize at once... I shouldn't go into that, right? Or is there limiting involved in normalization of individual tracks anyway?

Thanks in advance.
Go to the top of the page
+Quote Post
DVDdoug
post Jun 15 2010, 23:15
Post #2





Group: Members
Posts: 2534
Joined: 24-August 07
From: Silicon Valley
Member No.: 46454



QUOTE
1. Does normalization of an audio file in wav format to -1 db result in decrease of sound quality?

Short answer - No! Normalization does not degrade quality. Digital volume adjustment is considered lossless... Mixing engineers, mastering engineers and other people involved in audio production do it all the time, and they don't give it a 2nd thought. (I normalize my vinyl transfers.)

I assume that you are setting the peaks to -1dB? That's OK.

But if you want to "get picky", there are tiny rounding errors with almost any digital alteration, including normalization.

QUOTE
2. Should I normalize tracks seperately or all of them at once?
Some songs are supposed to be louder than others, so you should normalize the album as a whole. (All of the songs should be adjusted by the same amount.)

QUOTE
3. There are options for limiting etc. if I choose to select all to normalize at once... I shouldn't go into that, right? Or is there limiting involved in normalization of individual tracks anyway?
What software are you using? Normalizing should be a linear volume adjustment. Limiting (and compression) is not-linear and CAN degrade sound quality. The only time limiting would be involved is if you boost the volume to the point where the peaks would be above 0dB.


Go to the top of the page
+Quote Post
Light-Fire
post Jun 16 2010, 01:16
Post #3





Group: Members
Posts: 420
Joined: 5-August 06
From: Canada
Member No.: 33645



QUOTE (DVDdoug @ Jun 15 2010, 18:15) *
...But if you want to "get picky", there are tiny rounding errors with almost any digital alteration, including normalization.


So. Normalization decreases sound quality.
Go to the top of the page
+Quote Post
Axon
post Jun 16 2010, 01:42
Post #4





Group: Members (Donating)
Posts: 1984
Joined: 4-January 04
From: Austin, TX
Member No.: 10933



Yes, but to a degree that is quite provably negligible for all listening scenarios.

If you normalize and output to the sound card with 24 bits of precision, the normalization error will be 20-40db under the noise floor of the DAC. If you output to 16-bit, but perform internal processing at higher precision and apply dither/noise shaping, the only sound quality loss is an addition of background noise, not nonlinear distortion. When dithering is not performed one would be extremely hard-pressed to find a quantization error spectrum which could rise significantly above that DAC noise floor... &c.

If you're concerned about such things enough to avoid normalization, then you probably have no business listening to recorded music whatsoever, as virtually all digital and analog recordings suffer from distortions far worse in magnitude.
Go to the top of the page
+Quote Post
AndyH-ha
post Jun 16 2010, 03:05
Post #5





Group: Members
Posts: 2204
Joined: 31-August 05
Member No.: 24222



For some strange reason people seem to think it is all or nothing. Normalize. Save the normalized result to a new name. Listen to it. If it doesn’t appeal to you, go back to the original.

It is very unlikely you could notice any result of the quantization errors in normalizing a 16 bit file, but ... .
One must have the software with the required functionally, but converting the file (new name again!) to a floating point format should be available in quite a few application, I think. Normalizing in that format will certainly make unhearable quantization errors. Do a proper dithered conversion back to your target bit depth. You will never hear the tiny amount of added noise.
Go to the top of the page
+Quote Post
benski
post Jun 16 2010, 03:30
Post #6


Winamp Developer


Group: Developer
Posts: 670
Joined: 17-July 05
From: Brooklyn, NY
Member No.: 23375



If the audio is sufficiently quiet, it's most likely that the small amount of quantization noise introduced from the normalization process will be outweighed by the reduced amplification required in the analog domain (which equates to less noise) to actually be able to listen to the music.
Go to the top of the page
+Quote Post
meurglys0
post Jun 16 2010, 09:50
Post #7





Group: Members
Posts: 16
Joined: 14-January 09
Member No.: 65534



I thank everyone for the replies...

QUOTE (DVDdoug @ Jun 15 2010, 17:15) *
QUOTE
2. Should I normalize tracks seperately or all of them at once?
Some songs are supposed to be louder than others, so you should normalize the album as a whole. (All of the songs should be adjusted by the same amount.)

QUOTE
3. There are options for limiting etc. if I choose to select all to normalize at once... I shouldn't go into that, right? Or is there limiting involved in normalization of individual tracks anyway?
What software are you using? Normalizing should be a linear volume adjustment. Limiting (and compression) is not-linear and CAN degrade sound quality. The only time limiting would be involved is if you boost the volume to the point where the peaks would be above 0dB.


The problem is I downloaded the vinyl rip and the files are already seperated. I'm using Adobe Audition 3.0 and when I select an individual track, under the "favorites" tab there's "Nomralize to -.1 db" option, but that's not applicable when I select all the files; if I select all and right-click and select "group waveform normalization" a more complex normalization menu appears and the limiting etc. options are there.

I gave normalizing tracks individually a shot and the result was fine for most of the tracks. But two quiet tracks resulted sounding a bit louder than they should, I guess.

So can you recommend me a software that will take separate wav files and normalize them all at once without losing the quietness of the quiet songs compared to the loud ones... ?

Regards.

This post has been edited by meurglys0: Jun 16 2010, 09:51
Go to the top of the page
+Quote Post
db1989
post Jun 16 2010, 09:55
Post #8





Group: Super Moderator
Posts: 5275
Joined: 23-June 06
Member No.: 32180



You could use foobar2000's Converter to create new files with 'hard' (i.e. applied directly to the audio data, not just stored in tags) album-mode ReplayGain, although you'll probably need additional amplication to approach -1 dB, and I'm not certain if this option exists (there doesn't appear to be a page about the Converter's Processing options on the wiki).

This post has been edited by dv1989: Jun 16 2010, 10:00
Go to the top of the page
+Quote Post
Arnold B. Kruege...
post Jun 16 2010, 11:06
Post #9





Group: Members
Posts: 3647
Joined: 29-October 08
From: USA, 48236
Member No.: 61311



QUOTE (meurglys0 @ Jun 15 2010, 17:03) *
1. Does normalization of an audio file in wav format to -1 db result in decrease of sound quality?


Generally, no. In rare borderline cases, yes.

There is a tiny theoretical loss quality that is inherent in just about every kind of processing. However, it takes a lot more than a tiny theoretical loss to cause an audible loss.

QUOTE
2. Should I normalize tracks seperately or all of them at once?


I presume that you are normalizing to obtain a collection of recordings that have consistent loudness. While normalizing may actually accomplish that in some cases, it is far from the best way to do it. Normalizing relates to amplitude, but due to the complex properties of human hearing, amplitude does not always mean the same thing as loudness. So, normalizing does not always improve the consistancy of perceived loudness.

QUOTE
3. There are options for limiting etc. if I choose to select all to normalize at once... I shouldn't go into that, right? Or is there limiting involved in normalization of individual tracks anyway?


AFAIK the best way to cause a collection of tracks to have the same perceived loudness is to be skilled at doing such things, and doing it manually. The second best way is to use a program that uses a psychoacoustic model to evaluate loudness, and follow its recommendations. IME, the second best way is actually very good, and it is the best choice for most people who don't want to make a career out of audio production.
Go to the top of the page
+Quote Post
meurglys0
post Jun 16 2010, 11:32
Post #10





Group: Members
Posts: 16
Joined: 14-January 09
Member No.: 65534



QUOTE (Arnold B. Krueger @ Jun 16 2010, 05:06) *
AFAIK the best way to cause a collection of tracks to have the same perceived loudness is to be skilled at doing such things, and doing it manually. The second best way is to use a program that uses a psychoacoustic model to evaluate loudness, and follow its recommendations. IME, the second best way is actually very good, and it is the best choice for most people who don't want to make a career out of audio production.


Thanks.... But what software -which will do the job and is simple also- should I use?

This post has been edited by meurglys0: Jun 16 2010, 11:33
Go to the top of the page
+Quote Post
Arnold B. Kruege...
post Jun 16 2010, 11:41
Post #11





Group: Members
Posts: 3647
Joined: 29-October 08
From: USA, 48236
Member No.: 61311



QUOTE (meurglys0 @ Jun 16 2010, 06:32) *
QUOTE (Arnold B. Krueger @ Jun 16 2010, 05:06) *
AFAIK the best way to cause a collection of tracks to have the same perceived loudness is to be skilled at doing such things, and doing it manually. The second best way is to use a program that uses a psychoacoustic model to evaluate loudness, and follow its recommendations. IME, the second best way is actually very good, and it is the best choice for most people who don't want to make a career out of audio production.


Thanks.... But what software -which will do the job and is simple also- should I use?


Last time I had this situation, I used FOOBAR2000.

I double checked its results by ear and was satisfied that the FOOBAR2000 recommended level adjustments were about the same as I would develop using lengthy listening sessions and a lot of experimentation. BTW, I applied the adjustments with Audition.
Go to the top of the page
+Quote Post
2E7AH
post Jun 16 2010, 11:46
Post #12





Group: Validating
Posts: 2424
Joined: 21-May 08
Member No.: 53675



As dv1989 mentioned, foobar2000 in Album Replay Gain mode is what you should try. It's easy and you have many options: after you have RG scanned your tracks, you can use preamp in Playback preferences if you want it louder (with processing: apply gain and prevent clipping) and then even convert the files when you are satisfied with the results (Converter also has preamp option in processing tab). Thou you don't have to alter your tracks if you play the tracks with RG aware player

...too late

This post has been edited by 2E7AH: Jun 16 2010, 11:48
Go to the top of the page
+Quote Post
Notat
post Jun 16 2010, 14:33
Post #13





Group: Members
Posts: 581
Joined: 17-August 09
Member No.: 72373



QUOTE (meurglys0 @ Jun 15 2010, 15:03) *
But the volume level of this rip is lower than the necessary level for me to be able to listen to it via my digital audio player (zune 120) in public transportation.

I assume this is because the volume control won't go up high enough to overcome noise. What kind of headphones are you using? I see people wearing standard iPod earbuds on the subway and figure if they're able to make out the music, they must be doing damage to their ears. Best to use in-ear style headphones that seal and isolate.

Have you measured the peak levels of the tracks? Normalization won't do anything if the tracks already hit full scale.
Go to the top of the page
+Quote Post
2Bdecided
post Jun 16 2010, 15:59
Post #14


ReplayGain developer


Group: Developer
Posts: 5059
Joined: 5-November 01
From: Yorkshire, UK
Member No.: 409



You just need to amplify all the tracks by the same amount, that amount being the amount needed to bring the highest amplitude peak to -1dB or -0.1dB or 0dB or whatever you prefer.

You can do it manually in Audition. I only have Cool Edit Pro (much older version of Audition), but it's easy: load all the tracks into a single track (i.e. append them all), check the gain change required to peak normalise that long full-album track to the level you want (there's an option for that under "amplify", and it's also listed in the track statistics, but these can take a while to gather), close the full-album track, and then apply that same gain change to each and every track separately. Job done.

If you don't want to load all the tracks into a single track, load them all individually, check the required gain change for each track, and then use the smallest value you find.

As others have said, foobar2k will do this for you automatically, but peak normalisation-by-album requires the correct settings - which, from memory, mean ReplayGaining the tracks, and then applying the album gain with clipping prevention with a "with ReplayGain" pre-amp at maximum. It'll only work if the required gain change is less than the combination of the ReplayGain change and the pre-amp setting. It's simpler than it sounds - though it's a bit of a waste of time because it calculates the ReplayGain information and then doesn't use it (apart from the peak value).

Cheers,
David.

This post has been edited by 2Bdecided: Jun 16 2010, 16:00
Go to the top of the page
+Quote Post
[JAZ]
post Jun 16 2010, 18:11
Post #15





Group: Members
Posts: 1751
Joined: 24-June 02
From: Catalunya(Spain)
Member No.: 2383



Normalization:
- Usually useless (Cd's, for example, tend to be normalized already)
- Generally inconsistent (If it even gets to do something, as stated above, amplitude is not the same as loudness, and two files normalized to the same value can have a much different loudness).
- So 1990's.
- For the most part the effect of the change is bigger than the side effects. (so quality should not be affected perceptibly).

Compression/limiting:
- May be mostly useless (CD's are for the most part compressed, FM Radio/TV compresses on emitting)
- Difficult to get good results (There isn't one setting for all, and most general settings either are soft, or too constant).
- They are good tools, if used properly. Here at hydrongeaudio, when this topic shows up, it is mostly for the bad use of it.

Replaygain/loudness adjustment:
- The proper method to do what most people think that normalization does.
- Like normalization, it's just a change of amplitude. Unlike normalization, it applies different amplitudes depending on how it sounds, not how it looks like.
- If you have a player that supports replaygain (there are some portable players that start to support it, but there aren't many), then applying just a tag is enough.
- If the player does not support replaygain, the change has to be applied to the file, (like normalization does).


Like suggested above, the proper way to replaygain a flac album for use on a portable that does not support replaygain is:

-Add all flacs album files in foobar2000.
-select them, context menu(right mouse click), replaygain, scan all files as an album.
-select them all again, context menu, convert.
-select the output format (you can create it wav, mp3, flac...).
-select processing tab, enable replaygain, and select album gain mode. Inside the replaygain configuration, you can increase the final gain, if 89dB is not enough (probably your case with the portable. 92 or 94 is a good suggestion. if you neeed higher than 96, there is probably a problem with your player).
- and let it convert.
Go to the top of the page
+Quote Post
Ethan Winer
post Jun 16 2010, 18:40
Post #16





Group: Members
Posts: 248
Joined: 12-May 09
From: New Milford, CT
Member No.: 69730



QUOTE (Arnold B. Krueger @ Jun 16 2010, 06:06) *
There is a tiny theoretical loss quality that is inherent in just about every kind of processing. However, it takes a lot more than a tiny theoretical loss to cause an audible loss.

This is the right answer. I recently applied 120 gain changes to a 32-bit Wave file containing a 400 Hz sine wave, and after all of 120 processes the distortion products were all below the -150 dB limit of my FFT program. If you do the same 120 volume changes with a 16-bit Wave file, the distortion is still below 0.1 percent. So for one gain change the distortion added is surely insignificant.

--Ethan


--------------------
I believe in Truth, Justice, and the Scientific Method
Go to the top of the page
+Quote Post
AndyH-ha
post Jun 16 2010, 21:55
Post #17





Group: Members
Posts: 2204
Joined: 31-August 05
Member No.: 24222



If you want the experience intended by the performance, you want the same amplification on each track, preserving the differences between tracks. In Audition you should find Open Append on the File menu. If the files are all alone in one folder, select all to get them all into one file, or select the individual tracks while holding down the Ctrl key (select in reverse sequence: 10, 9, 8, ... to get the original sequence: 1, 2, 3 ...). All will end up in one file with que ranges delimitating the individual tracks. Normalize to the value you want under Transform/Amplify/Normalize.

If you want the result back to individual files, open the Que List, select all, Batch. You will have to rename the output files if you want the original names. That is quick and easy with copy-paste.
Go to the top of the page
+Quote Post
Arnold B. Kruege...
post Jun 17 2010, 13:17
Post #18





Group: Members
Posts: 3647
Joined: 29-October 08
From: USA, 48236
Member No.: 61311



QUOTE (AndyH-ha @ Jun 16 2010, 16:55) *
If you want the experience intended by the performance, you want the same amplification on each track, preserving the differences between tracks.


One could hope that the origional mastering engineer actually thought about setting artistically relevant levels for the various tracks. However, in these days of hypercompression, that is not always a safe assumption.

Reading the OP again, it appears to me that there was problem with a digital player/headphone combination that could not play every track loud enough in noisy listening environments, which is rally a different set of priorities.

I agree that the best solution would be to correct the problem where it started, and obtain a digital player/headphone amp/headphone setup that had enough reserve gain to be satisfactory as it is being used. Fiio has some under $20 headphone amps that may be highly useful even in noisy mobile environments.
Go to the top of the page
+Quote Post
chrizoo
post Jun 21 2010, 23:37
Post #19





Group: Members
Posts: 345
Joined: 25-March 08
Member No.: 52274



QUOTE (Ethan Winer @ Jun 16 2010, 17:40) *
QUOTE (Arnold B. Krueger @ Jun 16 2010, 06:06) *
There is a tiny theoretical loss quality that is inherent in just about every kind of processing. However, it takes a lot more than a tiny theoretical loss to cause an audible loss.

This is the right answer. I recently applied 120 gain changes to a 32-bit Wave file containing a 400 Hz sine wave, and after all of 120 processes the distortion products were all below the -150 dB limit of my FFT program. If you do the same 120 volume changes with a 16-bit Wave file, the distortion is still below 0.1 percent. So for one gain change the distortion added is surely insignificant.--Ethan


In this regard, there is an interesting experiment, see postings #98 to #102 here:
http://www.gearslutz.com/board/remote-poss...malizing-4.html

Go to the top of the page
+Quote Post
Notat
post Jun 22 2010, 03:41
Post #20





Group: Members
Posts: 581
Joined: 17-August 09
Member No.: 72373



Floating point is a different animal than the integer math Ethan used. You should consider Ethan's results worst case and be happy that many modern DAWs use floating point math.
Go to the top of the page
+Quote Post
meurglys0
post Jul 2 2010, 11:29
Post #21





Group: Members
Posts: 16
Joined: 14-January 09
Member No.: 65534



I thank all for the answers.

On foobar2000, when applying the replay gain to the tracks during conversion,

Source: Album
Processing: apply gain and prevent cllipping
with rg info: left at 0

without rg info: THIS IS WHERE I DON'T UNDERSTAND.

I read that it should be slided to -8db but I don't understand what this is... could you please help me with this...

Thanks in advance.
Go to the top of the page
+Quote Post
shakey_snake
post Jul 2 2010, 12:49
Post #22





Group: FB2K Moderator
Posts: 4322
Joined: 1-November 06
From: Cincinnati
Member No.: 37036



With that slider, if your files are not tagged with replayGain data, you can tell fb2k to make a volume adjustment.

With modern popular music recordings, the replayGain adjustment usually lowers the volume to match the reference level of 89dB. So if most of your music is averaging out to having replayGain values of -5, then it would be reasonable to assume that, on average, any music you might wish to play that does not yet have replayGain tags should be adjusted -5dB--it's basically a guess.

If you are busy listening to your RPG'd files at your desired volume and you try new CD or something that is rather loud and doesn't have RPG tags, it can cause you to reach for your volume knob rather quickly! tongue.gif

This post has been edited by shakey_snake: Jul 2 2010, 12:54


--------------------
elevatorladylevitateme
Go to the top of the page
+Quote Post
meurglys0
post Jul 2 2010, 23:13
Post #23





Group: Members
Posts: 16
Joined: 14-January 09
Member No.: 65534



QUOTE (shakey_snake @ Jul 2 2010, 06:49) *
With that slider, if your files are not tagged with replayGain data, you can tell fb2k to make a volume adjustment.

With modern popular music recordings, the replayGain adjustment usually lowers the volume to match the reference level of 89dB. So if most of your music is averaging out to having replayGain values of -5, then it would be reasonable to assume that, on average, any music you might wish to play that does not yet have replayGain tags should be adjusted -5dB--it's basically a guess.

If you are busy listening to your RPG'd files at your desired volume and you try new CD or something that is rather loud and doesn't have RPG tags, it can cause you to reach for your volume knob rather quickly! tongue.gif


90 % of the stuff that I listen to were recorded during the 70s. That's the very reason I'm looking for a way to boost the volume of some of my albums, for in the 70s most albums weren't recorded at high levels compared to the modern recordings.

With what settings would I get the loudest result without sacrificing quality (ie. clipping)?
What happens if I set the "with replay gain" setting higher than 89db?
What happens if I set the "without replay gain" setting at -1db or even slide it to +db levels?

Here's one example:




EDIT:

I converted the above album with settings at: 89db, with gain 0, without gain -3db.
The resulting files turned out even queiter than the orginal files!


So with what settings should I convert this album?

Thanks in advance for your replies.

This post has been edited by meurglys0: Jul 2 2010, 23:36
Go to the top of the page
+Quote Post
DVDdoug
post Jul 3 2010, 00:06
Post #24





Group: Members
Posts: 2534
Joined: 24-August 07
From: Silicon Valley
Member No.: 46454



Your older recordings are probably less (dynamically*) compressed...

Modern recordings are compressed to make them sound louder. Compression is used to increase the overall average volume without boosting the peaks or or clipping the peaks. (Sometimes it's over-done, and some modern CDs are clipped.)

Does the make sense? Your modern recordings sound louder even though they are normalized to the same peak level as your older recordings.


QUOTE
90 % of the stuff that I listen to were recorded during the 70s. That's the very reason I'm looking for a way to boost the volume of some of my albums, for in the 70s most albums weren't recorded at high levels compared to the modern recordings.
If your 70's recordings are normalized (maximized), you can't increase the volume (linearly**) any more without clipping.

So... if you want to match the volume and you can't increase the volume on the older dynamic recordings, your only choice is to reduce the volume on the louder tracks. This is why Replay Gain reduces the gain on most tracks.

QUOTE
What happens if I set the "with replay gain" setting higher than 89db?
You are not leaving RG as much "room to work". If you've set-up RG to prevent clipping, it will end-up having no effect on many tracks. If you allow it to clip, the tracks will be louder and many will be clipped.

QUOTE
What happens if I set the "without replay gain" setting at -1db or even slide it to +db levels?
The tracks without RG will play louder. Since RG is reducing the volume on most tracks, these non-RG tracks will generally be louder than the RG tracks. If you go above 0dB, many of these tracks will clip.

Normalization works on the peaks. The peak volume does not correlate well with perceived volume.

Replay Gain tries to adjust the perceived volume. (BTW - The 89dB value is an acoustic loudness reference level, whereas the 0dB reference is the maximum peak digital value in a computer file or fed into a DAC, etc.)






* Dynamic compression is not related to file compression like FLAC or MP3.

** Compression is non-linear. You can use an audio editor to compress your older recordings and make them louder... But, reducing dynamic contrast will make them sound boring, just like modern recordings!

This post has been edited by DVDdoug: Jul 3 2010, 01:04
Go to the top of the page
+Quote Post
Notat
post Jul 3 2010, 05:00
Post #25





Group: Members
Posts: 581
Joined: 17-August 09
Member No.: 72373



QUOTE (meurglys0 @ Jul 2 2010, 16:13) *
With what settings would I get the loudest result without sacrificing quality (ie. clipping)?
What happens if I set the "with replay gain" setting higher than 89db?
What happens if I set the "without replay gain" setting at -1db or even slide it to +db levels?

You actually want to set the reference level lower than 89 to accommodate a library with a wider range of loudness. Anything that gets a positive Replay gain reading may potentially clip when played back with volume leveling enabled. The Replay gain specification suggests using 83. I personally use 85. 89 seems to work well for material only back to the 1990s.

MediaMonkey has a "clipping prevention" feature associated with its Replay gain implementation. It's not real clear how this works but I believe it lowers playback gain below the Replay gain suggestion if the Replay gain value would results in clipping.
Go to the top of the page
+Quote Post

2 Pages V   1 2 >
Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



RSS Lo-Fi Version Time is now: 24th July 2014 - 14:24