Thanks Ethan !

​@@ProductionAdvice Hello. I think I might shine a different light for you onto the reality of preserving sound resolution. First of all, soundwaves are not infinite and there was never a question of infinite values. Yes, there are some engineers, even brilliant ones, that talk about infinity, because they are making scientifically incorrect projection of the mathematical model onto the physical reality. In physical reality soundwaves are not infinite and one doesn't need infinite values in sampling (neither bit depth, nor sample rate, nor time, etc.). Let's say one hears a 5000Hz soundwave. All one needs to record it correctly is to sample 1 bit at 10kHz. In mathematical model it doesn't matter if you sample 1bit/10kHz or 8bit/10kHz, or 16bit/10kHz, or 24bit/10kHz, it's all the same, there is no difference, because the reconstructed signal is in each case a mathematically perfect 5000Hz. In physical reality, however, there is noise, and all kinds of physical and technological limitations. The aforementioned different sampling techniques may involve different ADC technologies (esp. 1 bit vs 24 bit) that may give slightly different results, having different filtering requirements, etc. Larger bit depths were actually intended as headroom in editing, so that an editor could do away with truncating bits but still maintain signal properties. In reality human ears are sensitive to modulating noise. If one could achieve stable 'perfect' reconstruction filtering (huge processing power needed, ie. for ultra-precise time measurements), such reconstructed signal would certainly be easier on the ears, at least to some. Someone could call it 'higher resolution', if one insisted. Anyway, it's the sampling rate that in reality is the major factor in regards to preserving the sound resolution. If one could sample at least at 3GHz (yes, gigahertz) and more, that would definitely bring the recorded sound to the level of super-realistic illusion (beyond current analogue masters) that would become difficult to distinguish from the reality.

​@@ProductionAdvice Hi again. I've read more of what you wrote about resolution and I'm afraid there's a cognitive error that you are making. You're describing resolution as if it had to be necessarily absolute in order to be even spoken of. No, it does not have to be, in fact it never is absolute. The Sampling Theorem is also not absolute. It's correct, but not absolute. Meaning that the restored soundwave is a 'mathematically perfect' approximation of captured reality, indeed a more or less realistic illusion, but in reality it does not ever become the soundwave that was heard by a microphone. It does in the mathematical model (which is essentialy an approximation), but not in the reality. It's a very important distinction that is usually overlooked by digital engineers, even the brilliant ones. The term resolution itself is relative. It can be used to desrcibe sound or to describe one's reflection in the mirror. Soundwaves are material, so they can be quantified, described in numbers. Sampling process is essentially taking measurments of numerical characteristics of soundwaves traveling in pressured space and time (as heard by a microphone and represented by voltage), according to a proper scale that is provided by the sampling theory. It does not matter how you illustrate this process, whether by waving hand or water flow, or stairs falling and rising, or pixelation. It doesn't matter. As said above, in physical reality soundwaves can be quantified. Say you are sitting in a concert hall, listening to an orchestral piece. There exists a number that you could apply to take a complete (but not absolute) record of soundwaves that sourround you (well, their representation as voltage), using the sampling theory as the basic principle. Later on you could reproduce soundwaves using the captured record in order to recreate a near-perfect (but never perfect) illusion of the concert. The PCM technology is a great attempt to make such complete capturing process possible. However in PCM the implementation of the sampling theory is based on 44100Hz as the universally applicable sampling rate. By no means it is a sufficient measure in most real world cases. DSD sampling that uses higher sample rates, is another improved variant of a complete attempt. But the true universal sampling rate needs to be much higher, in gigaherz regions, if it is to properly implement sampling theory into real world scenarios (i.e. a symphony orchestra playing in a large concert hall). The result would be sound reproduction that goes beyond what any master analogue recording can offer. Obviously present technological limitations deem using such high universal sampling rate uneconomical, if even possible.

most things "audiofool"" people talk about is sample rate, and higher sample rate, the higher up in frequency you can go. and even if its not audible (>20k hz) it can still add percieved clairty and sparkle to a mix.

​@@AboveEmAllProduction Higher sample rates are not simply to obtain higher frequencies, but to preserve soundwave spectrum. One cannot record a symphony orchestra at 44100Hz and expect that one can reproduce a perfect illusion of that orchestra in your bedroom. Impossible. You can, however, reproduce a superficially generated 22050Hz soundvawe perfectly in your bedroom at 44100Hz sampling rate. That is if you enjoy listening to a perfect 22050Hz (or lower) single soundwave 'music'.

@@getthecandies what's wrong with using words to describe things? all those three words have very specific meaning to describing subjective things such as audio

@@getthecandies so warm doesnt mean bassy/organic without sharp treble.. phasey doesnt mean problems with stereo image.. punchy doesnt mean a present 100-300hz range... damn i must be delirious hallucinating all of this, and you are completely of sound mind. cheers to you aswell

@@john_scott I bet you could, provided you retained a high dynamic range.

Nyquist frequency en.wikipedia.org/wiki/Nyquist_frequency AD converter aliasing en.wikipedia.org/wiki/Analog-to-digital_converter#Aliasing Real-world example and practical usage of sample rates higher than 48kHz: kzfaq.info/get/bejne/Y9BzqqykuZa8aIE.html&ab_channel=fabfilter

Easy to argue. Lost detail from lower resolution will statistically follow a chaotic distribution which will sound like noise. But is in fact lost detail between the 8- and 24-bit versions.

@@PartyMusic775 No, the "lost" detail would be correlated to the musical signal (if there was any) so not random. The only loss of detail is making by higher noise-levels. When comparing 16 and 8-bit this has significant impact, but not so much comparing 24 and 16-bit. You can actually have 8-bit shaped dither which sounds really quiet good - similar to tape hiss, for example.

More than double that now.

Wow! This”experiment” was the best explanation on bit depth that I have seen. This video will be watched in many audio engineering classes. This will become part of the “required” syllabus. Thank you!

Pop has near-zero DR, and is just about immune to noise...

5 to 6 bit equivalent is probably more correct with type 1 cassette tape.

yeah basically the people that came up with the cd standard are a bunch of morons, right?

His Show and Tell vid is truly superb. In fairness, he doesn't dive deep into recording, DSP processing, mixing, mastering, etc., when the lower noise floor of a higher bit depth is super beneficial. So it's not that 24 is totally excessive and useless, just that it makes little to no difference in playback applications.

LOVE Monty's video ! I send it to people several times a month :-)

The Monty Montgomery's video only managed to properly illustrate the sampling theory working in the mathematical model of reality, using superficially generated single frequency sound waves (1000Hz, etc.). You don't listen to single-soundwave-computer-generated 'music', do you? I don't think so. The Philips' PCM technology was an attempt to properly apply the sampling theory in the real world, but was by no means a perfect application. Sony tried a more sophisticated sampling method with DSD, but it has been a commercial failure, largely due to the fact that all computers in the world are using native PCM. Regardless, the 16bit/44.1kHz so called 'Red Book' format is just a commonly accepted format from the 80's, that by the way has been since brutally diluted by the lossy compression code. What Monty Montgomery seems to fail to understand is that the mathematical model of reality, although correct, is just approximation, meaning it's not the reality itself. In order for you to understand it, imagine that you have 2 apples and you add another 2 apples to those 2. In the mathematical model of reality a certain numerical system is being used to count your apples, say: 2+2=4. But is 2+2=4 in the real world? No, in physical reality it's always less (than 4). If you don't understand then you can yourself undertake a painstaking task of counting molecules of a chosen element in all of your mathematically correct 4 apples - you'll see that the number of molecules is constantly decreasing. And this is valid in all the material universe, for any molecules, in any part of the universe. Although there is no contradiction whatsoever in using mathematical models in the physical reality. I'm using it right now, sayin '2 apples', or 'number of molecules', etc. Regarding sound waves. One cannot record a symphony orchestra at 44100Hz sample rate and expect that one can reproduce a perfect illusion of that orchestra in one's bedroom. One can, however, reproduce a superficially generated 22050Hz soundwave perfectly in one's bedroom at 44100Hz sampling rate, just what Monty Montgomery did in his video. If Monty enjoys listening to a 'mathematically perfect' and indeed properly reproduced 22050Hz (or lower) single soundwave 'music', then godspeed to him and everyone that follows undereducated engineers alike.

​@@royschwaben9646 Indeed, Roy. The higher bit depths were intended to be used as headroom for digital editing purposes. This is why 20 bit master recordings were being made in the late 90's and later on 24 bit has become a universal format in pretty much all recoding studios. The recording companies started selling 24 bit masters (realistically in many cases probably usually less than 20 bit), marketing them as 'higher resolution' or 'master tracks', etc. Do they sound better? Many of them do, not having been crudely truncated to 16 bit or less, having more sophisticated dithering algorithms applied. Many were previously mastered using cheap algorithms or in the editing process gain was used that had no dither, etc. These may tremendously benefit sonically from being more carefully remastered.

Lol ! Thanks :-)

only chodes listen on mobile devices and laptops.

was wondering the same thing. I bet it does.

I believe if you record live at 8 bit it will be the same as the non dithered file. You can only get this quality post rendered with dither from 16/24 bit. Commodore amiga suffered with this in tracker software etc.

The difference would probably just be the noise floor. And considering how little difference it was at 16 and 24bit depth its extremely negligeble. Still, given how much space we have on harddrives these days you may aswell record at 24bits, just know that It wont really make much of a difference in the end product.

This is a common question, and one I had, too, actually. Checking back with people whose maths and understanding is far deeper than mine, the null signal is genuinely pure noise. In fact the theory shows that the difference can _only_ be noise, and this is just a practical demonstration of that. The only time you'd hear something that sounds like music is with very quiet signals if you *don't* dither (or there's a flaw in the dither implementation) in which case you'll hear something like the very tail end of the example at 6:20 In other words it's not that the dither noise in the null test is hiding or masking tiny musical details - the truncation error has been randomised by the dither and is therefore just noise. So it's literally true to say that all you have is the music signal plus noise, as the demo shows. It's confusing stuff, I know - does that help ?!

The real answer is detailed maths that I can't help you with. But here's an attempt at a "hand-waving" argument: productionadvice.co.uk/no-stair-steps-in-digital-audio/

Sure, and that's because there's a difference. To be clear, I'm not saying there isn't a difference, just that the difference isn't to do with the resolution of the audio. The noise floor is lower for 24-bit audio, and with some material that will be audible. Although I took a quick look at that paper and they don't mention dither once. If the 16-bit versions weren't correctly dithered, the difference will be much more audible, because people would have been hearing 16-bit truncation distortion.

They are basically "equalised" or "not equalised" (shaped, to put it correctly) to be more or less noticeable under extreme cases.

You’ll get a difference signal starting at the Nyquist frequency of the lower sample rate file, but this video is about bit-depth so I’m not sure what you’re getting at.

Looking at individual samples won't tell you anything helpful. Each version is a digital representation of the analogue audio waveform, which 'contains' all the musical details you're asking about. 24-bit is "better" than 16-bit because it has a lower noise-floor (assuming both are correctly dithered). That's it. Have you seen this video ? You might find it useful: kzfaq.info/get/bejne/ma-BbKyJuLvepYE.html

Yes, the only use for extended bit depths and in some cases sample rates is in studio work for further audio processing.

​@@miialamia1653 High sample rate is useful for mainly one thing: less steep anti aliasing filter. Since human hearing has a upper limit of at around 20khz, you gain nothing else from using higher sample rates than ability to use anti aliasing filter (LP-filter) that has smaller Q (slope) and that doesn't create that much issues at the corner frequency (steeper slopes have a hmp before the slope starts) and can get rid of ultra sonics and any possible aliasing they cause in bandwidth limited signals much more efficiently. That being said, there are some DSP-algorithms that can benefit from oversampling (EQ's), but for the same reasons as any filter would (rising the corner frequency) and thus some EQ's have built in oversampling to account for audio that's lower than let's say 48khz...

@@Mtaalas This video is about bit depth ... not sample rate.

Agreed - people think higher bit-depths mean great resolved dynamic detail, and that's what the video aims to dispel. Technically speaking higher SNR is "dynamic range" as you say

That's true, but it's not necessary for the purpose of this test. The files were full-res PCM when I created the test, you only need to hear the *results* of the null via KZfaq's codecs

​@@ProductionAdvice Sure, but I am not talking about your test, I am talking about what one can hear, and I am directly responding to your line in the video when you said you didn't know if people can hear the 24 bit.

@@blerblybliggots9801 Fair enough, I missed your point, apologies

Probably. This was one of the reasons classical music enthusiasts drove the development of the CD. They had long looked for a better playback option than the LP record.

So 8, 16 or 24-bit files or output. Floating-point formats like 32-float or even 64-float don’t need dither.

@@ProductionAdvice Oh, okay. yea, thank you for clarification. Have a good one, Ian!

They won't die as long as the people pushing them are making money hand over fist. There is way too much snake oil in the audio world. One of the leading purveyors is Stereophile. Here is a sample of the BS they publish from a review of a pair of $900 speaker cables. "For the rest of that day and a few more days to come, I left the SuperFlatlines in my system and repeated the pattern of listening to a short piece of music on my warmed-up system, letting the cables cook with a few sides of background music, then replaying my earlier selection. The Nordost cables continued to sound clearer, more open, and, especially, more colorful-up to a point. If they exhibited additional performance gains beyond their second day in my system, those distinctions were too subtle for me to report, hand on heart, as real. I dare say they did most of their running-in within the first few hours of near-steady use-and the degree of that change was laugh-out-loud surprising. " Riiiiiight.

Yes, you'll get exactly the same result for any music, regardless of frequency content

Navi Retlav if you’re working at 24 bit and planning on exporting to a lower bit depth, then yes you should dither. However not dithering probably won’t ruin your project

Yeah, you should. go to airwindows.com and download the Ditherbox plugin on your mixbus (use TPDF 24 option), you'll notice a difference right away: Transients are clearer and the sound isn't as crunchy. Btw, I recommend to mix with a dither plugin on the mixbus always activated and render the final mix through it, as it changes the sound from how you perceive it without it, you wouldn't want the mix to sound a certain way inside the DAW and in different way after exporting.

Ian answers this in the video. The answer is YES. You always dither when saving to fixed bit depths (after any calculations / processing, even a simple gain change).

You only dither the target file when ever you're going _down_ in bit-depth. If you have 24bit original recorded file, you don't gain anything by dithering it. But when you convert down to for example 16bit for a CD, you want to use dither.

@@Mtaalas Surprisingly many people seem to be unaware that their DAWs and plugins are signal processing/calculating the audio in 32 bit floating point (or higher) REGARDLESS of the bit depth of the source material. That means than when you save a file after ANY processing (even a simple gain change or a fade) to fixed bit depth (24 bit or 16 bit) then you use the appropriate dither (16 bit dither for 16 bit and 24 bit dither for 24 bit). You can use the plugin Bitter from Stillwell to see this clearly in your DAW of choice.

You can’t keep all the high frequency details, but the good news is most people can’t hear them ! To get the best results, use a high-quality converter (eg. iZoptope RX) and add 16-bit dither when exporting to the final file 👍

Well, you definitely wouldn't like vinyl or analog tape!

@@prep74 I like and listen to both. I still have a pretty good collection. The old school in me.

Well, strictly speaking hiss is also distortion ! But hiss is a very natural, benign "analogue" type of distortion, unlike enharmonic quantisation distortion. It's also random, which is less audible than truncation distortion which is correlated to the audio signal, because our ears confuse it *with* the audio signal. So yes, it's much more than masking. Truncation distortion is effectively prevented (because it's turned into hiss) and as a result the wanted audio signal isn't damaged. So from a technical perspective dither should be applied *every time* we export to fixed bit-depth (ie. not floating-point). Now, will there be an audible difference at 24-bit ? Very possibly not, but I say why take the risk when the solution is so simple ?

@@ProductionAdvice thank you so much for your answer! that completely cleared it up for me

Got the same question as you. I've been reading RME's forum for this: www.forum.rme-audio.de/viewtopic.php?id=25399 Hope someone could shed some light as well.

If using a poor DAC, yes. ESS DAC chips use the internal 32-Bit depth to do volume control, and they have a noise floor of around 24-Bit, so you can adjust 16-Bit down >40dB and it still is perfectly 16-Bit out of the chip.

It's mainly for mixing and processing in the box.

dreinertje no

Of course

He's talking bit depth not bit rates. In any event, the sound of cymbals will not change with a lower bit depth providing it is properly dithered. Do you think cymbals sound worse on SACDs, they are only 1 bit.