Discussion in 'Mastering' started by ray1018, Apr 8, 2008.
Dithering Set while Bounce or Burn into CD?
I'm not sure exactly what your question is, but that's when I do it, yes...it's one of the last steps before putting it on a CD for me.
When working on the track & mix, I keep everything at 24/44.1 until it's time to make the CD. Then the dithering goes on, during the conversion/burn. I use PowR-3, FWIW. I like the way it doesn't sound - the 16 bit version (on CD) sounds identical (to my ears) as the original 24 bit.
You answered my questiong..glad and Great!!!
will try and taste it now!!!
I just love this subject of dither. Does anybody here actually know what dither is? You might be surprised to find out that dither is nothing more than broadband noise of the white, pink or, brown variety and comes down to nothing more than HISS, in different flavors.
This nasty noise is important when converting from different bit depth's & sample rates. They put such exotic names on to such common crap. It's noise. Digital recording without dither sounds really awful! Any reverb turns into scratchy nasty sounding stuff as it fades into the abyss without some element of noise behind it, to make the converters have something to operate properly, upon. Otherwise, the analog to digital & digital to analog converters would get confused if there is no signal to convert. So you add dither, to taste.
Tasty noise cooking
Ms. Remy Ann David
Obviously you don't.
You might be suprised to hear there is rather more to it than that. Dither noise is not random, it is carefully designed to keep the LSB of the target bit-depth toggling constantly, and to average out to precisely zero.
Nope, nothing to do with samplerate conversion. Dither is just for when you truncate your word-length. The idea is: once you get the LSB of the target file toggling on and off constantly you can represent the original fractional value (that would otherwise be lost completely) as a probability that the LSB will be high or low.
The same principle applies to images (look at the cute cat pictures )
Imagine a photograph of a diagonal line. If you reduce the image resolution, what happens to that line? If you don't use dither, the process is completely deterministic: the values will be rounded up or down to the nearest integer value, and the diagonal line will become a series of steps. However, if you add some precisely calculated dither noise the result is a statistical system in which the probability of each sample being rounded up or down corresponds precisely to its original fractional value.... in other words, instead of a stepped line we get a cloud of dots which imply a perfectly diagnonal line, albeit with slightly blurred definitiion.
So, dither doesn't just 'mask the nastiness of digital audio', it actually increases the resolution of the resulting file by including extra statistical information about its low-level detail that would otherwise have been lost. The penalty you pay for that is a slightly increased noise floor (the blurriness of the 'diagonal' cloud of dots) which is why it is important to only apply dither once if possible. ie: work at 24 bit resolution or greater up to the very last stage.
While it has nothing to do with SRC, it usually does need to be applied because most SRC's will change the bit depth.
I know this is an oversimplification and somewhat silly, but I like to think of the dithering process as "keeping the CPU Busy", forced to write all that extra low-level information it would otherwise ignore or toss out.
Since it's "hearing" all that very low-level noise, it MUST keep the calcutions going, resulting in a more detailed render.
I know there's a lot more science behind the process, but it works for me.
We tend to think about dithering during the post-processing (e.g. mix, mastering) environment. Dithering is also highly valuable in the analog-digital conversion; however, it needs to be applied during conversion to 'keep the A/D busy' and linearize the bottom bits of the converter.
Basically everybody's response is like I said, the end result appears as a smooth, perceived as, low level broadband noise, in the end, at the ground-floor of the conversion process. For those not as technically oriented, who don't know balanced from unbalanced, proper gain staging and any other general understanding of audio, my explanation was aimed generally for the novice. Which you and I and others here are not. I certainly appreciate the mathematics behind it all since it is technically a Band-Aid like fix for otherwise a bad sounding crappy thing to record our audio with, known as digital. Kind of like explaining where babies come from to a five-year-old. You really don't want to get into the actual specifics of how it all works, precisely.... Which we all know is from a kiss. Precisely why I recommend everybody use a condom on their microphones. Otherwise, SM57's end up looking like U47's which visually would appear contradictory based upon numerical value since 57 is already more than 47.
I think we got there from dither and yawn?
Ms. Remy Ann David
Separate names with a comma.