this is unbelievable...

hello. i ripped my entire cd collection to hd, and using audacity (which is thought at default settings did absolutely nothing to a wav unless i applied effects to it myself) i amplified all of my wavs. most old recordings can be amplified on average by a couple of decibels and i like to have everything amplified… i guess it’s an ocd thing.

anyway, i just found out today that audacity by default has dithering (shape) enabled. WTF? what is this logic? so everything i’ve ran through audacity which i’ve only amplified are all degraded wavs with added noise now. JUST GREAT.

i really don’t see the logic in why audacity has dithering enabled by default, can someone please explain?

anyway, i want to know how i can set audacity so it does absolutely NOTHING to the wav i import. all i want to do is amplify each channel to it’s max without clipping and be done with it. i want NO NOISE ADDED TO MY WAV - just want to amplify it!

retail audio cd’s are 44000 16 bit, right? why is audacity set at 32 bit float by default? is this what’s messing me up?

unbelievable - i have to go through with 500 hour process ALL OVER AGAIN. please help me out here… god what a headache.


Audacity isn’t a WAV editor. It’s an audio or production editor. I believe you can go into Audacity Preferences and set things to match 44100, 16-bit, Stereo audio CD format and also stop dithering.

We’re betting you will not be able to hear the difference. The dithering signal is very low and it’s put there, as you said, to ease the transition between 32-floating internal format and 16-bit. Not the other way. You can make a 32 from 16 with little or no damage.

Most people insist on trying to edit and process truly terrible audio into a good quality show. If Audacity did filters and effects at the original sound format, the show would quickly turn to technical garbage. So we immediately convert to the very high internal format and apply filters and effects to that. Then convert to the target format later. It’s true. The input and output formats have nothing to do with each other.

You also need to know that 32-floating does not clip and overload like 16-bit does, so if you do run Audacity at 16-bit, you’re going to need to be very careful with your volume levels and other effects. Once 16-bit clips or overloads, it’s permanent.

And yes, only the OCD people are bothered by this. “How come my WAV export isn’t bit for bit perfect with the original show.” It also kills people trying to edit MP3 files. too. MP3 gets worse sounding each time you make a new one and you make a new one every time you edit in Audacity.

Nobody can hear a couple of dB and you are warned against using either Normalize or Amplify at the 0dB default settings. If you use your new music to make a compressed version for your portable music player, you may, and probably will cause clipping.

Amplify to -1 or -2, which may put you back where you were.


Unfortunately, unless you are amplifying by an exact multiple of 2 the laws of physics are against you.
For any other amplification amount the sample values need to be rounded. Rounding to the nearest 16 bit value (standard rounding, or “quantizing”) creates harmonic distortion due to the “step” patterns (similar to the “swirly pattern” effect when regular stripes or checks are shown on a TV). This harmonic distortion can be avoided by not rounding in the same direction for each conversion, but by adding in a randomisation factor to the rounding, This is called “dither” and it is designed to replace the harmonic noise that is generated by quantize errors with a different kind of noise that breaks up the quantize error patterns and increases the dynamic range of 16 bit audio to the equivalent of about 18 bit resolution.

In short, the act of amplifying, unless by a factor that produces exact numbers of bits (such as amplifying by 2 or 4…) creates noise. You have a choice of what sort of noise. The default in Audacity is for “shaped dither”, which places most of the noise where it is least audible. If you prefer harmonic noise, go into “Edit menu > Preferences > Quality” and set “High quality conversion: dither” to “None”.

“Audacity isn’t a WAV editor. It’s an audio or production editor.”

understood. i was always under the impression that audacity and cool edit are wav editors and cater to people who use it for general purposes (whatever that may be).

actually, i could hear the difference between a dithered wav and the original. it’s very slight, but the dithered wav obviously has noise during silence, and when there is music, the dithering makes it sound just SLIGHTLY warmer and not as dynamic. vocals also seem to sound more bare and ‘not all there’ at first i thought i was crazy and i’m known be to ocd and neurotic but it’s definitely there.

anyway, thanks so much for the informative response and advice. i swallowed as much as i could, but i’m really not as tech savvy as i may come across to sound (if i do at all).

just to be clear (aside from the amplifying in 16 bit mode) - if i disable dithering while importing and exporting my wav, i WILL end up with a bit by bit unaltered copy, correct? i know you touched on this when you said:

“And yes, only the OCD people are bothered by this. “How come my WAV export isn’t bit for bit perfect with the original show.””

…but i’m not sure if you’re referring to your paragraph before that about 16 bit mode clipping during amplification, or you’re speaking just in general.

wow, steve. as much as i am trying to understand your post i just can’t - i’m really not savvy in that way lol.

i was under the impression that if i loaded a wav, and audacity tells me that i can amplify by say, 1.2db without clipping, then i’m safe. my logic is that i can squeeze that extra 1.2 in so i won’t have to turn up my volume past say, 95%. i’m not sure if i’m on the same plane with you about the amplification by intervals of two - although when i play my music in audacity i always keep my volume number at either 75, 95, 125, 145, 165 etc. it’s always with a five at the end. and each of those numbers i listed increase the volume by two decibels. this is all considered that the said wav is amplified to it’s max… lol.

i’m probably just crazy and in my own little world here… man, if only things could be so simple.

That’s OK :slight_smile:
I’ll try and explain in “non techy” terms…

To make it easier to visualise, let’s imagine than instead of “16 bit” audio for CDs, it was just “4 bit” (in reality that would sound terrible).
In “4 bit” format there are only 16 possible values. As audio usually uses positive and negative values, our 16 values would be:
-8, -7, -6, -5, -4, -3, -2, -1, 0, 1, 2, 3, 4, 5, 6, 7

So let’s say that you have a recording and the maximum level of the recording does not use the full range of values . For sake of argument, let’s say that the maximum positive peak is only at +6 and the greatest negative peak is at -7. Clearly we are able to amplify the signal a bit so that the maximum positive peak is +7 and the greatest negative peak is -8. To do this, we need to amplify (which is the same as multiplying the number values) by about 1.167. Then our peak positive value becomes 6 x 1.16666, which is about +7, and our peak negative value becomes -7 x 1.16666, which is about -8. Because there are only 16 possible values, when we amplify by 1.16666, the values 4, 5 and 6 will be rounded up to 5, 6 and 7. The values -4, -5, -6 and -7 will be rounded to -5, -6, -7 and -8, but the values -3, -2, -1, 0, 1, 2,and 3, even though we are amplifying (multiplying) by 1.16666, when we round them to a 4 bit value they are still at -3, -2, -1, 0, 1, 2,and 3. Do you see the problem? There is “jump” in the amplified values that wasn’t there before we amplified.
Our output values after amplifying are:
-8, -7, -6, -5, -3, -2, -1, 0, 1, 2, 3, 5, 6, 7

The purpose of “dither” is to “smooth” out these jumps by adding a tiny bit of controlled randomisation.
3 x 1.16666 = 3.49998, which rounds down to 3.
4 x 1.16666 = 4.66664 which rounds up to 5.

What dither does is that it will sometimes round 3.49998 down to 3, and sometimes round it up to 4, so that on average it is rounded to 3.49998.
Similarly, dither will sometimes round 4.66664 up to 5, and sometimes down to 4, so that on average it is 4.66664.
The downside of this randomisation is that it can be just audible in what should be absolute silence. One way round this is to carefully trim the tracks to the right length and then insert un-dithered absolute silence between them - this is what they do on good quality commercial CDs.

i think i get it steve, at least as far as i can stretch my brain. i’ll have to re-read both your’s and kozikowski’s posts several times to fully digest them lol… but thank you both so much for the replies and your time. when i re-rip my wav’s, i think i’ll just leave their volume levels as-is in fear of causing any unwanted distortion or clipping. ocd sucks… why can’t i just be a kid again and be satisfied with listening to music off of analog mixtapes?

We’re pleased you’re not a basket case by now, so we’ll keep at it. Please know that we’re all volunteers and this isn’t a help desk. It’s a bi-directional forum. Anything we do to help you may be valuable when somebody else comes along with similar problems.

You shouldn’t think of digital audio as a smooth, graceful, integrated whole. It’s a pointillist construction. See illustration.
Screen shot 2013-03-18 at 11.03.30 PM.png
That’s a greatly magnified wave. You can do that, too by continuously magnifying a portion of your show.

Each of those points is a sample and the computer is only seeing the dots, not the lines between them. There are 44100 of them every second and they gather together in an orderly manner to reconstruct your show. The problem comes when you mess with the volume and there are no convenient dots to represent where you put the sound. There are dots too high and there are ones too low. You can’t have half a dot. So now it’s chaos and this mismatch can cause audible distortion. Dither is added to make sure the errors you created don’t line up to become serious damage. It’s impossible to have two successive dots off the same direction.

With dither you can sometimes hear a gentle hiss during dead quiet passages (particularly if you’re turning the volume all the way up like we know that you’re doing). But without dither you can get the harsh/warmth shifting in the show. Harmonic Distortion – distortion related to the musical tones in the show.

You can avoid all of these problems by leaving the CD audio right where it is.


I’ve fished out a couple of samples for you. These are from a high quality 32 bit master. One has dither and one hasn’t.

thanks guys. that does make the picture more clear for me, koz. as for the files - i can definitely tell that ‘a’ is dithered. it doesn’t have as much fidelity to it as ‘b’ does.

…or am i crazy?

Yes A is dithered, but it depends what you mean by “fidelity”. B has more distortion than A (if you listen carefully to the tail end of the fade-out the sound in B gains a “metallic” tone). Personally I find it pretty hard to hear any significant difference unless I crank up the volume to an uncomfortable level.

This is the 32 bit master:

Warm, lush fidelity when you didn’t start with any isn’t good, either.
It comes down to producing in a way that’s pleasing to you. People who mix on vintage Neve desks and use multi-track analog tape machines are right there with you. Surgical accuracy isn’t always the best thing.


Where do you change the dithering setting?
I have Audacity 2.0 from March 2012, and I went into EDIT - PREFERENCES and REAL TIME CONVERSION and HIGH QUALITY CONVERSION both have DITHER set to NONE.
Is this something different to what is being discussed here?

That’s it, Dither is disabled.

That’s right I remember I turned dither off myself, when I first download Audacity 2.0 I noticed when changing speed of the same file 2 times and comparing the 2 files in EAC WAVE Compare the samples would be different (but with a previous Audacity version changing the speed of a file twice would give the same samples), I remember back then Gale pointed me to this in the bug/fixes sections… Switching dither off fixed the problem and from there on changing the speed of a file would result in the same samples each time…

Dither noise is incorrectly applied by default if exporting to most formats where the bit depth is the same as (or higher than) the project. For example, this occurs if exporting to 16-bit WAV, 16-bit FLAC or MP3 from a 16-bit project. OGG is unaffected. Workarounds: Set “High Quality” dither to “None” in the Quality Preferences. To fix any files that have already been affected, see this Forum topic.