IPB

Welcome Guest ( Log In | Register )

3 Pages V  < 1 2 3 >  
Reply to this topicStart new topic
do you lose quality when lowering a wav file volume then raising it up
dhromed
post Apr 17 2013, 09:22
Post #26





Group: Members
Posts: 1244
Joined: 16-February 08
From: NL
Member No.: 51347



QUOTE (coolhotfun @ Apr 17 2013, 00:17) *
Thats what i'm wondering.. wheres the point where that data gets lost when i do -inf compared to -59db.. its getting lost somewhere. So wheres the point where stuff gets thrown out and if i go to raise the volume later are there things that'll be lost?


A regular WAV is sound represented as a huge series of samples with a certain value. If it's a 16-bit integer WAV, those values can only be whole numbers, just like these forum post can only contain whole characters. There's no such thing as half a character.

If you reduce the volume a lot, then at some point, the values will snap to 0. And there's no escape from 0, obviously.

QUOTE
But apparently somebody up there mentioned that file size doesnt have to do with that


A book with empty pages takes up just as much space as one with a story printed in it. Unless you compress it, and then you can write "this book is completely empty" on a scrap of paper. That's some space savings!

QUOTE
You guys got all that? and dont use any big words like attenuation or floating point format cause i dont know what that means. I didnt go to college for all this, like you guys ya know.


It's really advisable to learn about a few of those terms if you're going to do audio editing. Not kidding.
A few quick notes: Attenuation is making things quieter. Floating point is a notation system of representing a decimal number like 45.7958 in bits so that your computer can work with it. Integer, conversely is a whole number. A 16-bits WAV means the samples can have 2^16 different values.

Wikipedia is pretty good if you want more details.
Go to the top of the page
+Quote Post
db1989
post Apr 17 2013, 10:34
Post #27





Group: Super Moderator
Posts: 5159
Joined: 23-June 06
Member No.: 32180



QUOTE (dhromed @ Apr 17 2013, 09:22) *
QUOTE
You guys got all that? and dont use any big words like attenuation or floating point format cause i dont know what that means. I didnt go to college for all this, like you guys ya know.
It's really advisable to learn about a few of those terms if you're going to do audio editing. Not kidding.
This. None of these concepts are anything that require college, i.e require more than a few minutes and Google. You will learn far more if you make a basic effort to teach yourself instead of expecting everyone to spoon-feed you. Alleging that such elementary concepts are obscure and academic comes across as rather patronising.

Just IMO, etc., blah-blah-blah. Other members are welcome to walk you through this if they want to. I just personally donít think they should be expected to, and I donít think youíd be getting the most out of the experience if they did.
Go to the top of the page
+Quote Post
coolhotfun
post Apr 17 2013, 14:45
Post #28





Group: Members
Posts: 18
Joined: 4-July 11
Member No.: 92047



i guess nobody knows what goes on with wav files.... i guess i'll just have to hope that all the information is still in the wav files if i want to raise the volume later in the future.
Go to the top of the page
+Quote Post
Propheticus
post Apr 17 2013, 14:55
Post #29





Group: Members
Posts: 218
Joined: 10-September 11
Member No.: 93615



Some seem to do know what goes on. Heck someone even explained that integers when they round to 0 will have lost all data. If a sample has a low amplitude to begin with, let's say 1 and you reduce volume 60% this becomes 0,4 which will round to 0. increasing 0 by 60% again is...you guessed it 0.
The higher the bit depth the less you'll notice this. With 32bit floating point (please just google this) there is a lot of headroom, so 2 volume changes likely won't affect too much. If you're using 16 bit and to a lesser degree 24 bit projects, you be losing dynamic range every time you do the 'volume up-down dance'.
Go to the top of the page
+Quote Post
db1989
post Apr 17 2013, 15:43
Post #30





Group: Super Moderator
Posts: 5159
Joined: 23-June 06
Member No.: 32180



QUOTE (coolhotfun @ Apr 17 2013, 14:45) *
i guess nobody knows what goes on with wav files....
Plenty of people know. You have been given a wealth of information and pointers to other sources from which you can obtain more. Please do not continue to patronise people by falsely equating an impatience with reading comprehension on your part to a lack of knowledge on theirs.
Go to the top of the page
+Quote Post
dhromed
post Apr 17 2013, 15:46
Post #31





Group: Members
Posts: 1244
Joined: 16-February 08
From: NL
Member No.: 51347



QUOTE
if i want to raise the volume later in the future


They will be more noisy. Maybe not audibly noisy, but technically, yes.

So basically, don't do it.
Go to the top of the page
+Quote Post
bennetng
post Apr 17 2013, 16:04
Post #32





Group: Members
Posts: 215
Joined: 22-December 05
Member No.: 26587



QUOTE (coolhotfun @ Apr 17 2013, 21:45) *
i guess nobody knows what goes on with wav files.... i guess i'll just have to hope that all the information is still in the wav files if i want to raise the volume later in the future.


I posted a video in the previous page, I guess you didn't see it. If you don't want to learn any technical stuff then I can tell you, If you use 32-bit float format, you will not have AUDIBLE quality loss even if you change volume with a large range (e.g. over 80dB)
http://youtu.be/E_D6d83nVUE
Go to the top of the page
+Quote Post
coolhotfun
post Apr 17 2013, 16:13
Post #33





Group: Members
Posts: 18
Joined: 4-July 11
Member No.: 92047



naw i'm only using 44khz 16 bit like the good old days. Thats cd quality and i dont notice better sound quality difference with higher ratios. Anyways.. hope if i gotta raise the volume in the future i dont lose some data from having lowered the volume. Its not by much but sometimes if its really high pitched i have to lower it really small. Generally the file peak tops are about -8db so below the -6db line.. At least now its not hurting my ears when i'm editing. The original files are higher up like peaks of maybe -3db so i can hear any hiss on those then take that out and lower it. The -40 stuff i was talking about earlier was just tests to see if i'd lose data by lowering it then raising it. Stuff isnt that low.. D'ya guys get whats going on now? Its just a bunch of samples for the next project. Its crazy.. gonna take a year and a half and about 800-1000 hours for a 4min track and video.
Go to the top of the page
+Quote Post
Propheticus
post Apr 17 2013, 16:58
Post #34





Group: Members
Posts: 218
Joined: 10-September 11
Member No.: 93615



I'm done here, you're obviously not getting it. Do the project and edits in 32bit floating point and export the end result in properly dithered 16bit.
Go to the top of the page
+Quote Post
drewfx
post Apr 17 2013, 17:03
Post #35





Group: Members
Posts: 65
Joined: 17-October 09
Member No.: 74078



I think you need to answer some questions, because though many people here understand how .wav files work at the lowest level of excruciating technical detail, what you're trying to accomplish is either really unclear to us or you appear to be trying to do something that doesn't make any sense.

So:

1. Where is the noise that you are trying to hear (and remove) located? If it's in the .wav file then lowering the level of the signal also lowers the level of the noise.

2. What are you doing to the file to edit it (other than change the volume)?

3. What program are you using to do this stuff?

4. What is your end goal/what are you trying to accomplish?


What you have to understand is that the answer to the questions you asked is often going to be "it depends", which may be why you are not getting as direct answers as perhaps you are looking for. In floating point, lowering/raising the volume will produce no audible difference. But under fixed point, it will introduce errors/noise/distortion, but the noise may or may not be audible. To further complicate things, your audio editing program may (very likely) process 16bit or 24bit fixed point files using 32bit (or 64bit) floating point math internally.
Go to the top of the page
+Quote Post
Arnold B. Kruege...
post Apr 17 2013, 17:51
Post #36





Group: Members
Posts: 3537
Joined: 29-October 08
From: USA, 48236
Member No.: 61311



QUOTE (saratoga @ Apr 16 2013, 14:38) *
QUOTE (greynol @ Apr 16 2013, 13:06) *
QUOTE (Arnold B. Krueger @ Apr 16 2013, 10:56) *
Any good volume control will attenuate all in-band frequencies the same.

I'd much prefer one that is fitted to my equal loudness profile, personally.


Would this be worth implementing? It could be implemented.



I suspect that it has been implemented, more than once.

I suspect that this product includes such a thing:

http://www.audyssey.com/audio-technology/dynamic-eq
Go to the top of the page
+Quote Post
bennetng
post Apr 17 2013, 18:00
Post #37





Group: Members
Posts: 215
Joined: 22-December 05
Member No.: 26587



Anyone worries about losing quality after editing should backup the original files before editing. This should be the simplest answer. I was wrong to talk about floating point to confuse the OP.
Go to the top of the page
+Quote Post
bernhold
post Apr 17 2013, 19:14
Post #38





Group: Members
Posts: 41
Joined: 22-March 13
Member No.: 107330



A side note for audio software users playing around with WAV volume changes: For a quality loss to occur, you actually have to save the file as WAV and then re-import that WAV file between every volume change. It's not enough to lower the volume and raise it again without saving the file. If you do that, the volume change will always be lossless because the changes happen in-memory and the limitations of the WAV format won't come into play.

In short:

- Normalize to -90db, normalize back to previous volume = lossless
- Normalize to -90db, save as WAV, re-open WAV, normalize back to previous volume = quality loss (added noise)

At least this is what I observed using Audacity.

This post has been edited by bernhold: Apr 17 2013, 19:16
Go to the top of the page
+Quote Post
bennetng
post Apr 17 2013, 19:31
Post #39





Group: Members
Posts: 215
Joined: 22-December 05
Member No.: 26587



QUOTE (bernhold @ Apr 18 2013, 02:14) *
At least this is what I observed using Audacity.


It depends on the software you are using. That's why I actually exported the wave file in 32-bit float in my video illustration (at 2:40). In Adobe Audition 1.5, if you don't explicitly convert the file to float before editing, then you will lose a lot of quality after reducing 90dB in a 16-bit file, even if you don't save and reopen the file.
Go to the top of the page
+Quote Post
Brazil2
post Apr 17 2013, 19:35
Post #40





Group: Members
Posts: 119
Joined: 9-May 10
Member No.: 80499



QUOTE (bernhold @ Apr 17 2013, 20:14) *
- Normalize to -90db, save as WAV, re-open WAV, normalize back to previous volume = quality loss (added noise)
At least this is what I observed using Audacity.

Because Audacity is applying dither by default, check the Quality tab of the preferences.
Go to the top of the page
+Quote Post
bernhold
post Apr 17 2013, 19:51
Post #41





Group: Members
Posts: 41
Joined: 22-March 13
Member No.: 107330



QUOTE (Brazil2 @ Apr 17 2013, 20:35) *
QUOTE (bernhold @ Apr 17 2013, 20:14) *
- Normalize to -90db, save as WAV, re-open WAV, normalize back to previous volume = quality loss (added noise)
At least this is what I observed using Audacity.

Because Audacity is applying dither by default, check the Quality tab of the preferences.


Yes but why does it matter if dither is applied or not? I tested with and without dither and there was severe quality loss.

To clarify, I used WAV default settings (Microsoft 16-bit PCM). By the way, in Audacity advanced uncompressed WAV options, what's the difference between "signed 32 bit PCM" and "32 bit float"? Which is better? Both options produce the same file size.

This post has been edited by bernhold: Apr 17 2013, 19:58
Go to the top of the page
+Quote Post
bennetng
post Apr 17 2013, 20:02
Post #42





Group: Members
Posts: 215
Joined: 22-December 05
Member No.: 26587



QUOTE (bernhold @ Apr 18 2013, 02:51) *
what's the difference between "signed 32 bit PCM" and "32 bit float"? Which is better? Both options produce the same file size.

Try to normalize a file to -190dB and save as "signed 32 bit PCM" and "32 bit float" then normalize to 0dB again.
Go to the top of the page
+Quote Post
saratoga
post Apr 17 2013, 20:22
Post #43





Group: Members
Posts: 4718
Joined: 2-September 02
Member No.: 3264



QUOTE (Brazil2 @ Apr 17 2013, 13:35) *
QUOTE (bernhold @ Apr 17 2013, 20:14) *
- Normalize to -90db, save as WAV, re-open WAV, normalize back to previous volume = quality loss (added noise)
At least this is what I observed using Audacity.

Because Audacity is applying dither by default, check the Quality tab of the preferences.


That has nothing to do with dither. If you throw away most of the signal, of course there will be quality loss.
Go to the top of the page
+Quote Post
bernhold
post Apr 17 2013, 20:22
Post #44





Group: Members
Posts: 41
Joined: 22-March 13
Member No.: 107330



To bennetng: I just did it and can't hear any difference. But I'm on laptop speakers so I may not notice if the difference is subtle. What is the difference?

This post has been edited by bernhold: Apr 17 2013, 20:23
Go to the top of the page
+Quote Post
pdq
post Apr 17 2013, 20:27
Post #45





Group: Members
Posts: 3305
Joined: 1-September 05
From: SE Pennsylvania
Member No.: 24233



Signed 32 bit PCM encodes values between plus or minus one with a resolution of approximately one two-billionth.

32 bit float encodes values many orders of magnitude larger or smaller than that, but with a resolution of only approximately one sixteen millionth OF THE VALUE.

Both are complete overkill for audio data.
Go to the top of the page
+Quote Post
DVDdoug
post Apr 17 2013, 20:32
Post #46





Group: Members
Posts: 2441
Joined: 24-August 07
From: Silicon Valley
Member No.: 46454



QUOTE
...like those tests i ran where i took a normal sounding wav and lowered it by -40db then raised it again.. sounded the same. You'd think doing that would add some kind of hiss or something.

But it didnt..
Try the experiment again at 8-bits. By simply saving at 8-bits, you should be able to hear the quantization noise. If you go through the process of reducing the volume, re-saving at 8-bits, and then re-boosting the volume, you should hear the quantization noise get worse. (Maybe start with -12 or -20dB because -40dB is going to be too much at 8-bits.)

With 24-bit integers, you can probably get away with a 40dB change. At 16-bits, you might get away with it too, since at 16-bits your digital noise floor is more than 90dB down, and if you "throw away" 40dB of dynamic range, you still have more than 50dB of dynamic range remaining. In many cases, your acoustic/analog dynamic range is worse than 50dB. So a little quantization noise below -50dB might not be noticeable.

QUOTE
Thats what i'm wondering.. wheres the point where that data gets lost when i do -inf compared to -59db.. its getting lost somewhere.
Volume reduction is done by division. Volume boost is done by multiplication.

Let's say you have a sample value of 15123 from a 16-bit WAV file. If you reduce it by 40dB (a factor of 100) and represent the result as an integer, you get 151. If you re-amplify it, you now get 15100. (You wouldn't actually hear an error of 23 out of 15000, but with smaller sample values the errors become proportionally larger.)

QUOTE
By the way, isn't floating point discarding accuracy as well? What about values like 5,3333333~?
Yes, but the errors/rounding are very small. You can take any 24-bit (or 16-bit) integer, convert it to 32-bit floating-point, and convert it back with no change in the integer value. So, converting to floating-point never makes things worse.

It gets a little weird, because like everything in the computer, the floating point values are in binary (base 2). So, values like 1/10th are not perfectly represented either! But that's OK. These errors are so small that we don't have to worry about them.

And, the smaller the number (with a large negative exponent) the smaller the error. It's kind-of nice when the size of the error (or rounding) is proportional to the number value... If you are buying a Ferrari (or a Honda) and the price goes up $100, it's no big deal. But if the price of a hamburger goes up by $100, it's a very big deal!

But really, we are talking about errors/rounding in the Ferrari price that are a tiny-tiny fraction of a penny. When you convert the price back to dollars and cents, and display it on the computer screen, the errors way-way off to the right of the decimal point are truncated and what you see is perfect to the penny!

This post has been edited by DVDdoug: Apr 17 2013, 20:41
Go to the top of the page
+Quote Post
bennetng
post Apr 17 2013, 20:33
Post #47





Group: Members
Posts: 215
Joined: 22-December 05
Member No.: 26587



QUOTE (bernhold @ Apr 18 2013, 03:22) *
To bennetng: I just did it and can't hear any difference. But I'm on laptop speakers so I may not notice if the difference is subtle. What is the difference?

Maybe Audacity applied a limit to the normalize value so even if you enter -190 Audacity did not actually reduced 190. I can hear difference but the difference of reducing -190 and -300 are the same.

Try to use some quiet piano music and use headphone and reduce the volume again.
Go to the top of the page
+Quote Post
bernhold
post Apr 17 2013, 20:37
Post #48





Group: Members
Posts: 41
Joined: 22-March 13
Member No.: 107330



QUOTE (pdq @ Apr 17 2013, 21:27) *
Signed 32 bit PCM encodes values between plus or minus one with a resolution of approximately one two-billionth.

32 bit float encodes values many orders of magnitude larger or smaller than that, but with a resolution of only approximately one sixteen millionth OF THE VALUE.

Both are complete overkill for audio data.


Thanks
Go to the top of the page
+Quote Post
Brazil2
post Apr 17 2013, 20:54
Post #49





Group: Members
Posts: 119
Joined: 9-May 10
Member No.: 80499



QUOTE (saratoga @ Apr 17 2013, 21:22) *
That has nothing to do with dither. If you throw away most of the signal, of course there will be quality loss.

Yes, but I was thinking about the added noise he mentioned.
Go to the top of the page
+Quote Post
bennetng
post Apr 17 2013, 21:01
Post #50





Group: Members
Posts: 215
Joined: 22-December 05
Member No.: 26587



QUOTE (bennetng @ Apr 18 2013, 03:33) *
QUOTE (bernhold @ Apr 18 2013, 03:22) *
To bennetng: I just did it and can't hear any difference. But I'm on laptop speakers so I may not notice if the difference is subtle. What is the difference?

Maybe Audacity applied a limit to the normalize value so even if you enter -190 Audacity did not actually reduced 190. I can hear difference but the difference of reducing -190 and -300 are the same.

Try to use some quiet piano music and use headphone and reduce the volume again.


http://www.hydrogenaudio.org/forums/index....howtopic=100466
I posted a piano clip and the processed clips here
Go to the top of the page
+Quote Post

3 Pages V  < 1 2 3 >
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: 20th April 2014 - 17:10