24bit vs 16bit, the myth exploded!
Jun 25, 2017 at 11:53 PM Post #4,156 of 7,175
It just comes down to what sounds more pleasing if you were theoretically going to hear it. Undithered errors would take the form of distortions that correspond to the signal, which is generally thought to be displeasing when compared to the homogeneity of white noise. We tend to associate white noise with analogue and randomized distortion with digital artifacts, and dithering is often thought of as a way of making the noise floor of digital sound more like traditional analogue. It's not audible, but in theory that's why distortion is traded for noise.
Except that, apart from the very most basic dither, it's not just simply white noise. There are several different ways of shaping the noise to get it to work well as a dither signal, but be less audible as a noise signal.

If you want to jump in with both feet...this is a pretty good read, although obviously product specific, the principles are there.
 
Jun 26, 2017 at 8:05 PM Post #4,157 of 7,175
They got it right in the first line of the intro, it's the audio equivalent of transmission fluid. Not even that, because you actually need transmission fluid, more like washer fluid. I'm sure anything can be made better with more intelligent plugins, it's a matter of being worth it or not or whether it will ever be heard, like whether or not the fancy blue washer fluid is worth it to clean bird poo off, or if I'm better off with distilled water and some blue food coloring (just kidding, I wouldn't put food coloring in my washer lines). Because they were so honest at the onset, however, and the content is interesting, I will read it through.
 
Jun 26, 2017 at 8:23 PM Post #4,158 of 7,175
A lot of audiophila is like working out the ratio of a circle's circumference to its diameter to way more decimal places than you really need to slice a nice piece of apple pi for dessert.
 
Jun 26, 2017 at 9:16 PM Post #4,159 of 7,175
I liked the life of Pi
 
Jun 27, 2017 at 5:48 AM Post #4,161 of 7,175
[1] I'm sensing you're all riled up about this for some reason and didn't actually read my post properly.
[2] Perhaps you should have a think about the mechanism, dither can only span those quantisation gaps over time, time a transient may not have.
[3] You appear to be disputing a mathematical fact.
[4] You see this is where I'm disagreeing with you, dither can only work over a number of samples because it's a statistical method.
[5] I suspect the problem is that you haven't reviewed many digital waveforms close in and seen how much each point counts.
[6] Take the old Genesis 'Lamb Lies Down' album, very dynamic ...
[7] Well no, again you simplify, mastering doesn't just use TPDF, shaped dither is a popular way to get a better sound from the CD format: your declaration that is 'is best' disagrees with all those mastering engineers who use a shaped dither.
[8] Who can't tell the difference?
[9] Less hysteria, more thought please.
[9a] What makes you think 96dB covers all possible cases?
[10] How can you re-create a live concert sound in your house or headphones with 96dB?
[10b] Or even the sound of a well hit snare drum?
[11] It seems to me you are arguing for Mid-Fi, not Hi-Fi.
[12] You also keep ignoring the fact that some people post process their digital music as which point 24bit would be rather more useful.
And you keep refusing to explain your reasons for not liking 24bit as a 'consumer' medium, what's the deal with that? 'Because you don't think other people need it' wasn't registering here as a reason BTW.

1. Unfortunately, your "sensing" is backwards, it's you who haven't read your own posts properly! It was you who quoted that dither prevents non-linear behaviour and since then you've tried to argue that it doesn't?

2. Again, you've got it backwards and YOU need to think about the mechanism! I notice that you have yet again failed to answer the question and presumably you do not realise that transients, like every other sound, only exist over time. If the frequency of the transient is too high/fast (beyond the Nyquist point) then it cannot be captured and cannot be dithered. There is no transient or any other audio waveform which can be captured with just a single sample, at least two samples are required. If it can be captured, then it can be dithered.

3. Again, completely backwards. You posted the quote, which is the end result of the "mathematical fact" and you are the one now disputing that quote/mathematical fact!!

4. And that's precisely why you're stupid to disagree! Yes, dither does take a number of samples BUT so does encoding a waveform as digital data in the first place and so does reconstructing it back into an analogue waveform. A to D and D to A conversions are themselves statistical processes. This really is digital theory lesson #1, which you clearly do not know and is presumably why you're coming out with all this nonsense ... and you accuse others that a little knowledge is dangerous, sheesh!

5. It's really impressive how much you are able to get so utterly wrong with just one simple sentence!! Firstly, it demonstrates that you don't understand even the absolute basics of digital audio theory and secondly, your "suspected problem" couldn't be more ludicrous if you deliberately tried! 1. You are looking at digital data points, not a waveform! That digital data represents coordinates which will only become a waveform once processed by a sinc function, until it is, then of course it will contain errors, digital data is NOT an analogue of the waveform!! Come on, this is the basics of digital audio developed 90 years ago! 2. I've "reviewed" digital audio data "close in" pretty much every working day of my life for the last 20 years!

6. "Very dynamic" compared to what? Compared to many modern pop music releases sure, but compared to the dynamic range offered by 16bit, NO, it has a tiny dynamic range, about 100 times less than 16bit!!

7. This just gets better! Now you're telling a mastering engineer what mastering engineers do.

8. Duh, I'm quoting YOU! "I'm still not sure why you keep pushing 16bit, 24bit is clearly a better format - even if you can't tell the difference...". Are you not reading your own posts?

9. Less delusion and ignorant nonsense please and more logic and facts!!
9a. How many commercial recordings can you name with more than 60dB dynamic range? If nothing else answer this question!

10. Been to many live concerts or houses with a 0dBSPL noise floor have you?
10b. You normally sit with your ear an inch or two away from the snare drum during a live gig do you? The only situation I can imagine where you might have your ears that close to a snare drum during a performance is if you were giving the drummer a blowjob! In which case, some ear defenders would be essential and a moist towel would probably come in handy as well!

11. It seems to me that you are arguing for exactly the same "fi" but just want it to be bigger and cost more!

12. Explain why a 24bit distribution format would be more useful if people want to post process their digital audio.

G
 
Last edited:
Jul 20, 2017 at 1:25 AM Post #4,163 of 7,175
I still didn't understand why exactly 24bit is useful in recording and mixing music. I record and mix all the time and I've tried both 16bit and 24bit and have compared them carefully. These doesn't seem to be any difference here either, just like in playback.
 
Jul 20, 2017 at 1:36 AM Post #4,164 of 7,175
If you record at a decent level and don't do a lot of filters, there probably isn't any real difference.
 
Jul 20, 2017 at 5:37 AM Post #4,165 of 7,175
I still didn't understand why exactly 24bit is useful in recording and mixing music.

In terms of recording, the advantage is purely in terms of headroom, not directly in terms of audio quality. With 24bit we don't have to be concerned with the noise floor of the recording medium, so if we set a level which results in a max peak of say -20dB it's not an issue, whereas that could be an issue with 16bit, after leveling and compression are applied during mixing and mastering. So with 16bit recording we tend to record hotter but then that risks unexpected clipping.

In terms of mixing, it makes no real difference at all, assuming an appropriate signal to noise recording mentioned above. However, the bit depth of the mixing environment (as opposed to the bit depth of the audio file itself) does make a difference and most mix environments today are 64bit float and even going back 20 years, it was typically 48bit fixed. This added bit depth in mixing is required to combat the issue of successive rounding errors generating noise, which would likely become audible in a 16bit mix environment, due to the sheer number of processing steps in a typical mix. With a 64bit float mix environment tens of thousands of processors would be required for the rounding errors to reach audible levels and it therefore becomes a non-issue.

G
 
Jul 20, 2017 at 7:48 AM Post #4,166 of 7,175
In terms of recording, the advantage is purely in terms of headroom, not directly in terms of audio quality. With 24bit we don't have to be concerned with the noise floor of the recording medium, so if we set a level which results in a max peak of say -20dB it's not an issue, whereas that could be an issue with 16bit, after leveling and compression are applied during mixing and mastering. So with 16bit recording we tend to record hotter but then that risks unexpected clipping.

In terms of mixing, it makes no real difference at all, assuming an appropriate signal to noise recording mentioned above. However, the bit depth of the mixing environment (as opposed to the bit depth of the audio file itself) does make a difference and most mix environments today are 64bit float and even going back 20 years, it was typically 48bit fixed. This added bit depth in mixing is required to combat the issue of successive rounding errors generating noise, which would likely become audible in a 16bit mix environment, due to the sheer number of processing steps in a typical mix. With a 64bit float mix environment tens of thousands of processors would be required for the rounding errors to reach audible levels and it therefore becomes a non-issue.

G
OK, I get it now. Thanks. I always record at 24bit/48hz, and recently considered going back to 16bit/44.1 because I couldn't tell a difference and didn't see a point in filling up hard drive space unnecessarily. Might as well stick to 24bit then.
 
Jul 20, 2017 at 8:12 AM Post #4,167 of 7,175
OK, I get it now. Thanks. I always record at 24bit/48hz, and recently considered going back to 16bit/44.1 because I couldn't tell a difference and didn't see a point in filling up hard drive space unnecessarily. Might as well stick to 24bit then.

If what you record has fairly predictable max peak levels, then you'd have nothing to gain from recording in 24bit, you'd just be filling up hard disk space unnecessarily. In commercial situations we're often facing wildly different max peak levels from client to client, so poor predictability and clipping a take is unacceptable, so 24bit is useful purely from the point of covering our a**es! For over a decade, commercial engineers only had 16bit recording and had to dedicate more of their time to covering their a**es by being more assiduous with their input levels to start with but then we generally have a great deal less time to record today than we did 2+ decades ago.

G
 
Jul 20, 2017 at 9:23 AM Post #4,168 of 7,175
If what you record has fairly predictable max peak levels, then you'd have nothing to gain from recording in 24bit, you'd just be filling up hard disk space unnecessarily. In commercial situations we're often facing wildly different max peak levels from client to client, so poor predictability and clipping a take is unacceptable, so 24bit is useful purely from the point of covering our a**es! For over a decade, commercial engineers only had 16bit recording and had to dedicate more of their time to covering their a**es by being more assiduous with their input levels to start with but then we generally have a great deal less time to record today than we did 2+ decades ago.

G
I may be misunderstanding you but do you mean to say that if my recordings are ever so slightly clipping, I might be able to fix that if they're in 24bit, as opposed to 16bit?
 
Jul 20, 2017 at 9:34 AM Post #4,169 of 7,175
nope that's not what he meant ^_^. just that if you have "only" 16bit for your tracks, you will spend more time trying to get the recording not too far from 0dB, from fear of potentially audible background noise. while with 24bit, you just don't have to care. you can leave a good deal of headroom to be absolutely sure that you won't clip anything and also don't have to worry too much about the background noise becoming audible when you add some gain later on those tracks to match them with others. it's just an easier/safer process.
 
Jul 20, 2017 at 9:42 AM Post #4,170 of 7,175
nope that's not what he meant ^_^. just that if you have "only" 16bit for your tracks, you will spend more time trying to get the recording not too far from 0dB, from fear of potentially audible background noise. while with 24bit, you just don't have to care. you can leave a good deal of headroom to be absolutely sure that you won't clip anything and also don't have to worry too much about the background noise becoming audible when you add some gain later on those tracks to match them with others. it's just an easier/safer process.
Thanks. I understand it better now. Basically, I can record "colder" than before and not worry about the possibility of clipping. This is definitely worthwhile. Can't tell you how often a perfectly good take was almost ruined because of a few unexpectedly loud hits here and there.
 

Users who are viewing this thread

Back
Top