I have to say that in 15 or so years of being involved in audio production, I never have read or heard anyone speak on this subject as clearly and concisely as you have here. So, thank you a million times. I finally GET it now.
At 14:23 "ideally that dither should happen as the very last thing in the chain." For anyone curious as to why dither must be the last process, I just put up a quick demonstration video on my channel. The tldr is because your daw processes audio at a higher bit depth than the audio files themselves, any change you make after a dither immediately reverts the signal back to your daws internal processing bit depth. And to Justin, I think this mic sounds good for the podcast!
Hi Justin, thanks so much for all those valuable insights. This is one of the best youtube channels relating audio for me. Quality content and marvelous interviews. I would kindly appreciate if you could make an UNBIASED clipping plugins comparison. Some claims they can compete with clipping from 10k converters, but I don't like the hype on this one regarded to the already available such as SIR Standard Clip, or the one from Boz Digital, or the one from IK Multimedia to name a few... Please feel free to pick your own selection into that comparison. Thanks so much and keep up the good work man !
I would be interested in going deep into the weeds about dither but I'm a nerd. I realize that a lot of people don't really care about "how" it works. I have read a lot about it but I'm always interested about hearing how different people explain things. It will sometimes give me insights to things I may not have though about or realized. Love your podcasts. Thank you.
Dither? Dropdown list. Triangle? I guess, cool, I just want to print this thing. Never really played with different versions of the same song within the same DAW. I probably couldn't hear it anyway. I'll guess it wouldn't matter with today's bitrates, etc, and today's DAWs. Certainly not to the average listener. I'll watch the video now. Edit: Demystify the dither, geek out :) I'll watch for sure. Thanks for the vid.
Thanks Justin. I did a remix of your podcast. A 20-second version. It goes like this: "Hi this is Justin Colletti and this is the Sonic Scoop podcast. Thanks to Soundtoys. Ok. Does dither matter. No. Thanks. Bye". I can upload it if you want haha. Cheers for the podcast dude. Stay safe.
I look at it like this. If you are going to be playing your music live and through loud sound systems, specifically if the dynamic range is high, it doesn't hurt to be on the safe side. Realistically, most modern music that isn't a live performance is medium-low dynamic range and the chance of hearing the dither noise floor is practically 0.
The purpose of dither is not to mask "bad sounding noise". Its purpose is to correct inherent losses in digital conversion to better preserve the original signal by introducing random noise, which statistacally results in more data being preserved, along with a higher noise floor. it's a tradeoff.
You are correct that it is not "masking" so much as "replacing" the noise floor. I could have said that part more precisely, and in hindsight, wish that I did. Thanks for the comment! -Justin
If you do a "dither demystified" follow-up, I must admit that I have one burning question I've always been curious about... Since I work in Ableton Live which has a perfectly good triangular dither option when printing to 16 bit, I have always wondered why on Earth you would want/need a limiter plug-in to have a dither option included?? I guess the quick answer to this would be that the option is there for printing files in stand-alone mode OR that the quality of the dithering algorithm might be better than your DAW's default option, but using Waves L316 limiter as an example, that particular plug-in has several shaping options and also 2 types of dithering. One type doesn't automatically mute itself when there is no audio information playing, the other one does. So even without having the most perfect hearing in the world, I did actually notice that I could hear the dithering noise when mastering one of my tracks after limiting the pre-master - a reason I will never use that option on that limiter ever again in my workflow. Plus, if DAWs process audio at a higher bit-depth (32-bit floating point) in a session regardless of the source bit-depth (as I have been led to believe by reading information online etc), then does anyone know if the dithering noise is being applied before the quantization stage actually begins, and therefore if that makes any difference whatsoever (?). I have multiple plug-ins that offer dither in the limiting stage but all I know for sure, is that I have NEVER ever noticed the dither noise when using Ableton's default options, so using the dithering offered by a third-party plug-in just seems like an opportunity to make a simple situation needlessly complicated without being able to say for sure whether the third-party dither is actually a preferable solution or not? From what I have learned, I should just ignore this subject and carry on as I was, but it does also beg further questions such as - if you are recording a file at 24-bit (say a vocal recording from a microphone) - which isn't actually what my settings do btw, I record at 32-bit, but nonetheless, the 24-bit option is there if you want it - what is now happening when the DAW is handling the audio file at a higher bit-depth than the recording is created at in regards to further editing - say, if you now "freeze" the track (which would then be 32-bit floating point again), but then export the recording as either 24 or 16-bit in order to reduce the size of the files you have to manage on your hard drive? So lastly, what I mean there is that if you are converting 32-bit files down to 24 bit as a matter of course during the session creation, but then finally ending up at 16-bit for the final master, have you not applied dither at multiple stages, each applied to solve quantization artefacts at differing bit-depth scales? Should I care at all about this? Should anyone care at all about this lol!? Well, if someone could definitively tell me the answer than it's one less thing to worry about late at night haha! ...can't believe I've just written such a long comment about dithering of all things?!!
I didn’t comprehend dithering or analog to digital (or vice-versa) conversion until I understood the Differential Calculus I failed in high school. Rounds out the stepped edges of the interpolated waveform. I always use it (eg UV22) before making an mp3 for upload.
Hey! You're using the BB29 for your voice now. Sounds really good, Justin. Thanks for all you do. Can you do a video on Sample Rate Conversion? Maybe a part of a tune with a few of your different SRC's? I use my DAW's stock SRC and I'd like to know your opinion on if I should spend $500-$1000 on a "better" SRC. Thanks again!
To be accurate, dither replaces the quantization errors rather than masking them yes. I believe I say this more precisely at 3:15. Hope that helps! -Justin
We work in 32 bit float. From that to 24 is no problem but going all the way to 16 bit makes a big difference. So, yes, dither is important. If you’re a mastering engineer.
The idea that the noise that needed to be dithered is inaudible is simply an untrue statement. It's going to depend on the recording and the processing of of the person mixing and recording. The noise is caused by the over tones bouncing back into the audible range from lack of not high enough sample rate. If the processing and recording has caused a lot of it then it will definitely be audible. The problem is you might not know how audible it is until you take it out. Which dithering will not totally do. The best thing to do is to understand what causes it and what you're doing in the process that puts it in in the 1st place and avoid it or minimize it.
Dither provides resolution below the LSB (Least Significant Bit). That being said, the analog noise which some analog simulation plug-ins provide, in themselves, could be a good source of dither.
Yes, it can give you effectively greater dynamic range at certain frequencies! That is neat, and I mention it in the episode somewhere. But, would we really notice or benefit from that additional dynamic range below -96dB? Seems unlikely. But it is cool that it has that effect. Yes, in many cases, the noise floor of your analog audio would be higher than the noise floor of the dither, so your analog noise would likely mask even undithered quantization errors in many cases But the cool thing about dither is that it doesn't just mask the quantization errors with noise. It actually REPLACES the quantization errors with noise. That's neat. So you could have relatively lower levels of noise and still eliminate the quantization errors that way instead of just masking them. It's definitely a best practice to use dither when going down to 16 bit. Most DAWs now do this by default. But if you somehow didn't, it probably wouldn't be a big deal on any sufficiently loud modern production styles. Hope that makes sense! -Justin
I usually love all your videos, and i reaaaaally do not want to come off as an idiot here, but i thik this is not exactly accurate. Dither is something a bit more complex than just exporting to 16 bits. Dither also prevents audio distortion anytime there is bits truncation, and since every daw, work in 64 or 32 FP, and the DA converters work at 24 PCM, (and taking into account that many plugins work internally at higher bit numbers) Theres is always truncation. I agree that with or without dither you can´t really fuck it up, but still, it is important to use dither both while mixing, while mastering and when exporting audios in different bit depths.
Thanks for the feedback. Not sure there are any inaccuracies pointed out here! We actually seem to be in agreement on pretty much everything you've written here. Yes, truncation and quantization errors can be described as a type of distortion, but they occur at or around the level of the noise floor, which is about 96dB down for 16 bit, and 144 dB down for 24 bit. That's very very very unlikely to be heard in most practical cases. Dither can be used to replace these errors with white noise (or shaped noise). But actually hearing this effect generally requires cranking up the level during exceptionally quiet moments to unrealistic levels. The likelihood of anyone ever knowing you've dithered at 24 bit is probably close to zero, but it's a best practice to do it anyway, for sure! May as well. More or less the same can be said for 16 bit, though perhaps less emphatically. "Important" is subjective. You've said you'll probably be fine either way.... so is it THAT important? :-) That said, I basically agree with your recommendation! You should do it anyway for sure. It's a best practice, and it doesn't cost you anything to do-so absolutely, why not? Most DAWs will do it for you automatically these days, especially when going down to 16 bit. As far as needing to do it at 24 bit if someone else is going to bring it down to 16 bit after you..... Not necessarily, because the errors at 24 bit (or the dither) are going to be truncated away when you go down to 16 bit anyway. So you're basically fine either way in that case. But if you're *releasing* at 24 bit, you may as well dither it..... Even if no one will probably ever know.... But again, why not! Can't hurt and could maybe help in some extreme and unlikely cases. So yes, "always dither" is fine advice, and I basically agree. Just trying to put it into perspective :-) Hope that makes sense, Justin
@@SonicScoop Makes a whole lot of sense! Incredibly well put! I hope people get to read this comment of yours beacuse it certanly brings more light to the video! Thanks for taking the time to reply to my comment. I really apreciate it. Looking forward to the next video, as allways!
All dither does is push audio that's so quiet it's at the limit of your bit range and add noise to it so it gets a little louder. The upside is better fades, the downside is added noise. Honestly, for 99.9% of listeners it's totally irrelevant which dither you choose, just turn on powr and worry about something else.
Pardon my bad memory, but there's something I'd like to ask. I started trying my mixes/masters with and without dither. I can still hear the difference (only when there's sound) . . . and I liked without dither more. So here's the question: Is it necessary to dither a mastered audio file? As some sort of standard to prevent problems with online platforms, because tbh even with 16 bits I'd preffer the sound of no dither at all.
I have a friend thsat only Dithers his tracks, and sends them to Spotify, etc. He never masters them, because he says the streaming platforms with adjust the volume for the website. Sounds crazy to me. What do you think?
You are correct. It is crazy. First, if your track is too quiet it’s not going to get boosted to -14LUFS. Second, tracks that are mastered hotter than -14 often sound louder even after they are turned down. Third, mastering isn’t just about average level. It’s also about frequency balance, microdynamics and macrodynamics of the track. Hope that makes sense, Justin
I maybe crazy, but I think going from 24bit to 16 or even Mp3 sounds better no dither! My friend and I both agree as we compared on very good speakers. So maybe we are both crazy?..
But but but, there are different TYPES of noise that'll you'll almost certainly never ever hear at 16 bit or above!! 😂 Doesn't that matter, like a LOT???
SonicScoop - Thank you for this useful video. I have one technical question for the mastering process and dithering. So, I mix in 24bit, export the file for mastering in 24 bit, then during mastering the last plugin in the chain is the limiter that has a Dither, I use either AOM Invisible limiter or Ozone Maximizer. Once I'm happy with the processing of the file in mastering I print the audio. My question is what bit depth should I select during the printing of audio the file, 24 bits or 16 bits in the Dither menu? In Pro Tools, I export the newly mastered printed in both 24 bits for Streaming and 16 bits for CD and I always wondered if I should print two versions, one with 24-bit dithering, and then export in PT for the 24 bit audio; and another print with 16-bit dithering and export that file in 16 bit for CD releases. It sounds to me that there are some inefficiencies and duplication of effort to print twice the same song just for dither. The bit depth is really determined during the export of an already processed file and I'm not sure about the proper process. I currently print just one track with 16-bit dither and then export in 24 bits and 16 bits, realizing that the 24-bit audit export actually had a 16-bit dither process. Thank you!
Is the production done in 32 bit float? Then you don’t need dither for the 24 bit mixing print. The 16 bit file for streaming and cd needs dither. It has less dynamic range compared to the original production and 24 bit mixing file. So to be short: use dither on the 16 bit print and print a separate 24 bit without dithering. Hope this helps.
Yes, if you are sending them the final version for release. (But if you forget, it’s unlikely anyone will notice.) No if they are doing mastering for you. (Though if you forget, it’s unlikely anyone will hear it.) -Justin
I always thought dithering was similar to converting a RAW photographic image to a jpeg, so some info is thrown away and replaced with information that is borrowed from elsewhere in the source.
I exported 32 bit float to 24 bit without dither. No matter how loud I crank up the quietest parts, I can't hear any quantization distortion. Should I dither nevertheless and if yes, could I bounce the already bounced file with dither as I am lazy af?
so essentially dither does identify unwanted white noise-frequencies and turns down their volume like an eq? a bit like noise cancellation, at least when it comes to the detection part. just without the cancellation by countering the frequencies as to not leave any holes in the recording. did i get that right?
Not exactly. The noise inherent in digital audio does not sound perfectly random like white noise. It sounds a little glitchy and artifact-y and distorted. Dither replaces this awkward-sounding low level noise with white noise (or some other type of "shaped" noise). Which sounds better. That is, if you can actually hear it 🙃 It's not so much detection as it is randomization of quantization errors. Basically, the very last value in a digital "word" is totally randomized making the noise more perfectly random (in the case of a very basic dither.) I could go into much more detail on this, but I'm not sure how useful it would be! It would require talking a little bit about word lengths and truncation and how digital files are written to the point where you'd probably be like "wait, why do I care, can I get back to making music please?"
Lowering bit depth to 16 bit doesn’t really lower quality. It does reduces the dynamic range. But it’s still overkill for consumer playback. All that happens is that you have a louder noise floor. 96dB instead of 144dB. But no analog gear you can record with can actually use the full 144dB of 24bit to begin with, and the 96dB of 16-bit is already overkill for consumer playback. Remember that even pretty quiet masters aren’t really going to use more than 14 or 16 dB of dynamic range. A lot of commercial masters barely use 10dB! (Or less.) Going down to 320kbps mp3 doesn’t perceptibly reduce quality. But like going down to 16 bit, it reduces size. This is an important feature. What if you can reduce file size by 2x or 5x or 10x without hearing a difference? Wouldn’t that be good? Or do you want to run out of hard drive space, internet bandwidth and batteries 2x, 5x or 10x faster? At a certain point, going for bigger file size will not improve quality perceptibly, but will have other costs. A 1 trillion bit file isn’t going to sound perceptibly better than a 16 bit one. But it will sure cost a whole lot more to start filling your libraries with files that size! Try streaming one and see how you like it :-) Hope that helps! -Justin
Dither is used to replace the noise for in digital audio to make it sound more pleasant. Oversampling is an approach used by some plug-ins (and converters) to work at a higher sampling rate than the actual sampling rate of your session and audio files. Oversampling can be useful for things like saturators and synths to help them perform better, and can be used by AD converters to help make their anti-aliasing filter even more transparent. Hope that helps!
Dither doesn’t just simply mask distortions, it is involved in retaining information from your music in digital form when losing bits. This is why your DAW is constantly dithering when you touch anything. I guess just not doing it once isn’t going to hurt, the problem would only come cumulatively.
You might be dithering without knowing it. A lot of DAWs will automatically apply dither when going down to 16 bit by default, but may not apply it in other cases. That's a pretty good solution if you ask me! Hope that helps, Justin