Okay. In the context of subtractive dither, I understand the confusion. Yes, the purpose of subtractive dither is to reduce the total dither noise remaining in the output, typically after adding dither during bit reduction. Clearly, that will increase the SNR and allow you to hear more of the signal.
> On Jan 10, 2022, at 12:58 PM, Zhiguang Zhang <[email protected]> wrote: > > sampo started this thread as OP so not sure what this discussion is about, i > think he was talking about 'subtractive' dither. i actually think the > problem is more nuanced, perhaps a music business problem dealing with music > publishing and licensing law > > On Mon, Jan 10, 2022 at 12:55 PM vicki melchior <[email protected] > <mailto:[email protected]>> wrote: > Eric, > > You’ve lost me. I don’t follow what you are worrying about. This > discussion isn’t about the audibility of dither. In fact white noise > generally is fairly benign from an audibility standpoint. > > The reasons dither is added have to do with avoidance of distortion due to > undithered quantization error, It is this distortion that can be quite > audible and can produce coloration, because the distortion is coherent in > nature and not noiselike. > > Vicki > >> On Jan 10, 2022, at 12:33 PM, Zhiguang Zhang <[email protected] >> <mailto:[email protected]>> wrote: >> >> well i'm not here to talk about whether or not i can discriminate dither >> from music, it is pointless for me as someone who listens to music >> >> On Mon, Jan 10, 2022 at 12:31 PM vicki melchior <[email protected] >> <mailto:[email protected]>> wrote: >> No, I never said that dither becomes a coherent signal. Dither is noise. >> >> >>> On Jan 10, 2022, at 12:14 PM, Zhiguang Zhang <[email protected] >>> <mailto:[email protected]>> wrote: >>> >>> Vicki, >>> >>> that is rather incredible to me if true, that dither is detectable as a >>> coherent signal but i suppose that the dither that i was referring to is >>> necessarily a part of the program material signal because it is the dither >>> that has already been added during the recording chain and thus not a >>> separate coherent signal >>> >>> https://ask.audio/articles/the-how-and-why-of-dithering-in-pro-tools >>> <https://ask.audio/articles/the-how-and-why-of-dithering-in-pro-tools> >>> On Mon, Jan 10, 2022 at 9:15 AM vicki melchior <[email protected] >>> <mailto:[email protected]>> wrote: >>> Eric, I’m not sure I get the gist of your question, but “hearing into the >>> noise” refers to the fact that coherent signals can be detected at some >>> level (around 10-15 dB) below the RMS level of the noise (whether the noise >>> is dither or part of the signal). The mathematical analogy for this is >>> coherent/noncoherent gain; the hearing system integrates both noise and >>> signal over the bandwidth of the particular cochlear filter. Noise >>> integrates non-coherently while signal integrates coherently, leaving a net >>> gain in SNR. This is relevant for a number of reasons. First, you can >>> (maybe) detect actual signal at those depths below noise. But second, you >>> can also hear distortion lying well below the noise floor if it is >>> relatively coherent, especially the peaks associated with truncation >>> distortion when dither has been omitted. These arguments are highly >>> relevant to determining the bit depth needed to convey program material, >>> and that in turn, is a function of the dynamic range audible to humans >>> along with an understanding of the noise sources present in the given >>> system. So it is not about hearing the noise, but rather hearing signal >>> below the noise floor. >>> >>> >>>> On Jan 9, 2022, at 8:10 PM, Zhiguang Zhang <[email protected] >>>> <mailto:[email protected]>> wrote: >>>> >>>> not sure if this point is important, but the dither that is added before >>>> you hear the program material being reproduced isn't actually supposed to >>>> be 'heard' - so this argument doesn't appear to make much sense in my >>>> mind. engineers might hear the dither when they're familiar with the >>>> studio that they work in, but past that, i'm not sure i get the point of >>>> discussing the practical limits of hearing something added which, for all >>>> intents and purposes, is hidden. it's almost like you're trying to >>>> reverse engineer what recording interface an audio engineer was using >>>> >>>> On Sun, Jan 9, 2022 at 5:48 PM Brian Willoughby <[email protected] >>>> <mailto:[email protected]>> wrote: >>>> Thank you for these titles. I've already found them in the AES library. >>>> >>>> Brian Willoughby >>>> >>>> >>>> On Jan 9, 2022, at 13:43, vicki melchior <[email protected] >>>> <mailto:[email protected]>> wrote: >>>> > As far as measurements of how far “into the noise” we can hear, there >>>> > aren’t a lot of good published numbers that I know of (having reviewed >>>> > the subject a couple of years ago), but Bob Stuart and Peter Craven >>>> > argue dynamic range and, to a certain extent, audibility below the noise >>>> > floor in a couple of papers published in JAES in 2019. They are based >>>> > on psychoacoustic arguments as well as listening test results, the >>>> > latter as part of their studio and lab work on MQA. If interested, >>>> > their (open access) papers are in the AES e-lib, “The Gentle Art of >>>> > Dithering” and “A Hierarchical Approach for Audio Capture, Archive and >>>> > Distribution”. >>> >> >
