https://www.gearslutz.com/board/6235423-post76.html

 

I set up logic in 24bit for my project and bounce my final work 24bit 48khz -6db ready for mastering (I never master my own tracks ) any sampling I do I bounce at 24bit for a clean sound . It's just the way my engineer likes me to do my work and I've always stuck to it .

 

I actually did some tests at 8bit, and the only situation wher dithered sounds better is if the signal is barely grasping above the bitdepth floor, in which case any noise smooths that transition. But that's when it's really quiet. I DID say at sensible levels, so you won't encounter this if you're sensible. And btw, 8 bit tests aren't very useufl, because we export at 16bit at minimum, which has a dynamic range 2^8 = 256 times better than 8bit, so again, if you're sensible, you won't noticed this. I don't.

PS: I don't mix or master at 8bit.

It's simple marketing. If you write advice saying you need it, you'll get more sales.

 

I usually work in 44.1 32bit then bounce dithered to 24bit, that unless certain situations where i need to track higher or say amp sims or some type of simulation i might work in 192 32 then en up 44.1 24 in end i noticed this day and age almost any delivery can handle the 24 bits i havent used mp3 in years just flac of course what happens when im done who knows

 

"
I actually did some tests at 8bit, and the only situation wher dithered sounds better is if the signal is barely grasping above the bitdepth floor, in which case any noise smooths that transition. But that's when it's really quiet. I DID say at sensible levels, so you won't encounter this if you're sensible. And btw, 8 bit tests aren't very useufl, because we export at 16bit at minimum, which has a dynamic range 2^8 = 256 times better than 8bit, so again, if you're sensible, you won't noticed this. I don't.

PS: I don't mix or master at 8bit."



i dont get your logic or what you are saying at all its like saying "well my eyes cant see the difference between 4k and 1080p so instead of making the project in 4k and then 1080p delivery of product , i just work in 1080p the whole time"

there are too many factors involved to base much directly on what you think you can hear or see in these situations. your ears on your system has nothing to do with much of squat. people cant hear the difference between their crappy mix and one of the top notch pro mixes so does that mean there IS no difference? or does that mean for whatever reason they DON'T hear it?

 

So if you see difference at 8 bit, why do you ignore this at 16 bit? You don't hear this difference because the noise of your speakers and soundcards are louder than 16 bit artefacts? Or undithered 16 bit is totally enough for delivering all you need? Less is more, more is worse? Or what?
For example if I don't hear so much higher than 18 kHz, so I should request final formats and recording at 36 kHz sample rate? It is total bullshit.
I hear difference between dithered and truncated at 16 bit, especially with headphones. Bob Ohlsson said that even in well treated and maximum silent room you can hear it more. So it means more you do to decrease shit in sound, better the sound. Yes, on shitty speakers 10bucks headphones mp3 is much more than enough, maybe you shouldn't mix at all for this shitty stuff. But remember about hi fi and achieving maximum results for such minimum quality formats as 44.1 16 bit

Nobody mix and master at 8bit, it is for tests, to hear the difference at once. So to say, difference in your face.

Airwindows is not marketing, it is "4-10 free but extreme quality plugins per month", they beat a lot of hundredbucks plugins that are real marketing and beautiful analog GUIs

 

Last edited: Mar 16, 2017

[OT] The word "dither" in this thread is used 44 times even if the question by itself has nothing to do with dithering (diretlcy). Dithering is one of the most overestimated technqiues, IMO.

[Update: 46] [/OT]

 

Errorjan,
If guy asked about samples, project with high sample rate and bit depth, it means he wants or tries to achieve maximum quality of final digital audio product. So it means dithering is what he should use for 16 bit to avoid some problems in audio. Digital audio problems are usually aliasing and quantization errors /noise and nothing more, only these two create harshness, death cold robotic sound, that is why we need minimize this. Other digital problems are lossy codecs.

 

I totally understand, MPM - I just wanted to mention that dithering is one issue of many. And it has no influence on the sample rate, but on the bit depth (both related to the question - sure). When reading this topic, I just noticed that everybody was talking about dithering - so I just want to note that this is not the most important issue here.

More important, and again, IMHO, is how to downmix 88.2/96 to 44.1. A poor man's method would be to just interpolate 2 (96: 3) samples to one. Here the "real" quantasation effects, which are noteable even to someone without a perfect pitch, comes into play. As long as freq above 22khz were used and not filtered before... I don't and can't want get deeper into freq-conversion, but I'm pretty sure that in a mastered song, in case the sound mixer forget to dither the final take before it gets mastered on CD will have a smaller impact that when he/she forgets to apply a hicut filter when downsampling.

 

Errorjan,
Don't know why one needs to use high cut filter before downsampling, because extremely steep filter starting at 22000 and ending at 22050 with linear phase causes no aliasing/intermodulation distortions (or much quieter than -180 db for 32 bit int files), only inaudible and extremely fast pre and post ringing at 22.05 kHz area. Such src are finalcd sharp mode, dbpoweramp src (I did this test on src.infinitewave.ca), sox, izotope rx4 advanced (for precise adjust of filter hold ctrl and drag filter cut fader), the brick (Mac only), shibatch Ssrc ...
No matter do you use 48000 or 96000 or another sample rate for 44100 conversion. Today is 2017, not 1989. Maths problems for standalone top quality src system have been solved.
Lesha Vaneev aka Voxengo promised many years ago to release R8brain version 2 with much improved filters for src and other options, but now we have only 1.5 pro version which is very cool and ideal for 2000s quality, but today it is not so very bad, but not excellent as posted above. But some people still do like it. Some prefer Weiss saracon, but it is gentler and has some aliasing at ultrasonic area, so it is good for stuff with not so much highs, also it means less pre and post ringing, but the frequency range of this ringing will be wider.

 

go in your daw look in the resample tab then when you bounce dither , but why ever leave 24bit in 2017?

 

>Today is 2017, not 1989
Yes, unfortunately (personal issue with my age and the kind of EDM these days;)

I think that you are experienced in this area and know the basics of DSP theory. Then you also understand why a waveform _looking_like_ a square wave @22khz is resolved to a sine wave after the signal passes (a good) 44.1khz DAC since the 22khz wave matches the nyquist frequency and nothing above that freq is representable (like a square wave would behave at lower frequencies). I'm just talking about theory and not what the cool Voxengo plugins are doing with the signal when, for example, oversampling is used (I assume some kind of hicut - would be awesome if Lesha Vaneev is active in this forum and would be so kind and clarify that).

So what I want to say, is, that w/o any conversion algo (which *is* probably applied automatically, like you said) when downsampling, the signal will be distorted in case the track contains signals in the freq area >nyquist. Proof: Comparsion screenshots attached. 30 khz sine wave at max amplitude (ok, extreme example:p) in one screen measured with simple downsampling (note the muddy peaks) and the other in 96 khz (also some peaks there but far less – don’t ask me why – of course the peak at 30khz is desired).

 

Let's put it this way: if you had a PROteus MPS+ PCM workstation from the early Nineties that was 16-bit sound, and you wanted to record it to a DAW at 88.2 kHz/24-bit, it would be the same thing as bouncing a DAW track from a sample library that was recorded at 44.1 kHz/16-bit (well, maybe with better fidelity than the PROteus). You're certainly not losing anything by recording the VI track to your DAW, and you're maintaining the ideal sample rate and bit-depth you want for your project. The 44.1 kHz/16-bit samples in your sampler VI will sound the same regardless of what fidelity you're recording at.

 

Absolutely correct that nothing is lost. And please, I don't want make a science out of this tiny issue now, it's really not so important since every tool/converter should care about hicut when downsampling (as said, I just talking about theory).

So let me quickly try to explain: When there is no signal above 22.5 khz in the 96khz stream, there is also nothing to be taken into account or to filter out when downsampling (proof: combining signals is always an additive process, and 0 + 0 = 0 ^^).

In most cases there aren't signals in that region, but if so, mostly only very very weak signals (noise) which aren't hearable even when it's (wrongly) not filtered out by whatever software-converter or analog system.

In case there is a signal, say a very loud sine wave @30khz like in my example (which is of course not common - just an extreme example), then there is something != 0 (DC) in the 96 khz signal. And when just closing the eyes and taking every 2nd sample as downsample algorithm (BTW which is not the best way - interpolating 2 samples is better, but that's not the issue right now and has no influence on this explanation) and not carry about possible signals which aren't representable after the conversion process, the downsampled data might become muddy.

It's unfortunately a fact that 44.1 khz signal data can only "hold" singals from >0Hz to <= the so called Nyquist frequency, which is half of the samplerate, so 22.05 @44.1 khz. Please have a look at the sample picture (the 44.1 khz one) of my last post: There are peaks (mud) at different locations (see the peak at 500, 800, 1000, 1100 and so on) as the 30 khz sine wave was not filterted out when I downsampled it in Ableton using a sampler and the "resample" function (NOTE in case you are familar with Ableton: when using the "HQ" option [small checkbox], then Ableton is doing exactly what I'm talking about - it takes care that "Nyquist issues" [there are more than only this example of downsampling] are solved by the program so the user don't need to care; since this cost some little extra performance, it's an option).

Oh, just got the genius idea to check whether there is already an explanation - and there is - all writing for nothing x-p
http://blog.dubspot.com/optimizing-sound-quality-in-ableton-live/

I think this is all just a misunderstanding since you see the issue from the perspective as user of a wav-converter/other software which already uses the math and I in turn saw it from the perspective of a theory guy used to do DSP programming some time ago.

Hope & keeping my fingers pressed that this is solved now

Cheers.

 

Oh, just noticed I totally missunderstood your answer - sorry. It really helps reading *every* word of a post before answering

So please forget my last answer.. And I agree with your answer - in case the source material is not in 96 (or more) khz, it makes nearly* no difference whether to work in 44 or 96, except you need a faster CPU when using 96 khz (personlly, my none-top but modern CPU wouldn't handle the projects I make in realtime in case I *would* work in 96 khz mode - I always work in 44 - better using 40 tracks, 10-15 auxs and 8 returns and all that nice stuff that is possible these days than not being able to do so because of 96khz support).

(*) Only exceptions is when heavily creating music with synths. Sure, finally everythink will be converted to 44.1 khz, but while processing the output of the synth with other plugins, there are really (very small) advantages in the quality of the final rendering when downsampling is the last step and not done immediately before the signal goes out of the VSTi. But I bet nobody is able to hear the difference, it's only measurable and can therefore be neglected. There are many people with another opinion and some even think it helps to use 2.82 Mhz(!!!) (see DSD madness https://en.wikipedia.org/wiki/Direct_Stream_Digital). That's ok for me as long as I am allowed to continue in 44 khz...
(Keeping fingers crossed I didn't read/understand/say something wrong again )

 

Nope, I was pretty much agreeing with your earlier posts!

 

White papers from Lavry Engineering

• Digital Audio: An Introduction
• Sampling Theory
• Sampling, Oversampling, Imaging, Aliasing
• The Optimal Sample Rate For Quality Audio
• Acoustic Bit Correction® Do you need 20 Bits?
• Understanding FIR Filters
• Understanding IIR Filters
• On Jitter
• About the >PiC< ™

http://www.lavryengineering.com/lavry-white-papers/

 

the reason you may want to begin a project in higher sample rate than import files from lower sample rate is because the processing done with vst and hardware unites when working at the high sample rate many times will generate harmonics above the audible rate that still impact what it is the audible range.
personally my preference whenever possible is to work in delivery format () normally 44.1 24 bit) unless you are making a youtube video there is NO reason to have shit audio less than 24 bit these days with FLAC making mp3 obsolete and the cheap storage space available to every human hard rives are dirt cheap flash drives pen drives etc. we should be focusing on a 24 bit delivery more than sample rate. if you work in high sample rate say 192 then at end re-sample down to 44.1 at very last step of mastering nothing is really lost that audible , however no matter what a finished mastered mix at 24bit will always sound obviously shittier and worse objectively if reduced to 16bit no matter how great the dither or noise shaping, combine this with the objectively no need to ever go below 24bit anymore and the problem is actually solved now. ( notice i am not counting the occasionally super rare case of needing to go to 16bit)

 

I actually hadn't thought of that -- since CD production is likely on its way out, the need to down-convert to CD quality (44.1 kHz/16-bit) probably isn't very high. I know music for film is usually 48 kHz/16-bit or 96 kHz/24-bit, and commercially-released .mp3s are still at (below) CD quality, but most people seem to be gravitating to what they keep calling "HD music" nowadays, so maybe the format is finally on its way out. Personally, I still keep in mind that I'll have to master for CD, but by the time I'm ready to put something out again, it's possible that the only music formats around will be vinyl and PONO. And maybe cassette will make a comeback?

 

Has anyone seen this video?