Keni
Thanks Craig...
 
I understood that anytime you change sample rate (not bit depth) you were supposed to dither so that the unmatched spaces get slurred with the dithered noise...?
 

 
Wherever you heard this, please remove them from your list of trustworthy technical information sources. 
 
Unfortunately there is lots of stuff written about digital audio that's "easy to understand" but just plain wrong. Also unfortunate is that the correct technical explanation is often just not easy to understand intuitively.

Below is a paragraph from Dithering With Ozone by Izotope Page 8. 
 
"Section III: "Over on the corner there's a happy noise"
Down on the Corner by Creedence Clearwater Revival
So what to do? We dither. We'll get into the details of dithering later, but for now consider it as
adding very low-level noise to the audio before it is converted from 24 bits to 16 bits."
 
To me that is a simple way of putting it but its a useful way in understanding it. I use the word mask because that is what dither does. The artifacts are there but cannot be heard due to the masking effect of the noise that is added. 

John
Below is a paragraph from Dithering With Ozone by Izotope Page 8. 
 
"Section III: "Over on the corner there's a happy noise"
Down on the Corner by Creedence Clearwater Revival
So what to do? We dither. We'll get into the details of dithering later, but for now consider it as
adding very low-level noise to the audio before it is converted from 24 bits to 16 bits."
 
To me that is a simple way of putting it but its a useful way in understanding it. I use the word mask because that is what dither does. The artifacts are there but cannot be heard due to the masking effect of the noise that is added. 

In audio, the word masking is a technical term that has a specific meaning. It is not useful in understanding dither simply because dither does not rely on masking to work. I suspect it's "useful" to you only because you don't actually understand what's going on. Using incorrect, but perhaps easy to understand, explanations do not aid actual understanding and instead only add confusion. 
 
And the artifacts can be heard after dither, but they have been transformed from artifacts that are correlated with the signal (bad sounding distortion or modulated noise) to a less ugly and less obtrusive steady noise.

John
To me that is a simple way of putting it but its a useful way in understanding it. I use the word mask because that is what dither does. The artifacts are there but cannot be heard due to the masking effect of the noise that is added. 

Actually, the cool part of dither is that's not the case. Dithering doesn't cover up quantization noise, it replaces it with a different noise of our choosing instead. It's pretty cool. Monty Montgomery has a reasonably complete explanation that isn't too technical in this video (the whole video is a great primer on digital audio).

We'll just have to agree to disagree on this. I see nothing wrong with the way I described dithering. 
 
Masking may be a tech term. I was using it as a word with its normal meaning. 

John
We'll just have to agree to disagree on this. I see nothing wrong with the way I described dithering. 
 
Masking may be a tech term. I was using it as a word with its normal meaning. 

There's no agreeing to disagree. It's not a matter of opinion. This one is cut and dry and you're simply wrong.  
 
We can go through how dither works and masking in excruciating detail. Dither doesn't mask the errors - it transforms them into something else.

Six of one, half dozen of the other....
 
Personally, I lean toward replacing as the proper concept as it's done all in the digital domain - quantization noise isn't something that is treated or filtered separately. Once dither is blindly added, the quantization noise is gone - it can't be recovered. (Which isn't to say that new quantization effects can't be created by further downsampling.)

And if you don't wish to believe me:
 

drewfx1

John
We'll just have to agree to disagree on this. I see nothing wrong with the way I described dithering. 
 
Masking may be a tech term. I was using it as a word with its normal meaning. 

There's no agreeing to disagree. It's not a matter of opinion. This one is cut and dry and you're simply wrong.  
 
We can go through how dither works and masking in excruciating detail. Dither doesn't mask the errors - it transforms them into something else.

Sorry thats a distinction without a difference. BTW we should keep our focus on the thread and its context. My post was meant to simplify the concepts in such a way that the OP would have a better idea of what is really happening. You may think masking is a poor way to define dither. I think its a good way to understanding the idea behind dither. 

The issue lies in the definition of mask (meaning to conceal), vs replace, which is to substitute. 
 
The first one purports the object being concealed still exists but is hidden, where the second suggests the item is substituted, and therefore no longer exists. 
 
 
update:
 
Wikipedia:  The premise is that quantization and re-quantization of digital data yields error. If that error is repeating and correlated to the signal, the error that results is repeating, cyclical, and mathematically determinable.
 
I think the above statement indicates that the quantization noise should still be detectable to be considered to exist, and therefore merely masked.