.
 
 
 
 

JohnEgan
 
...
 
However, that aside, as I understand bit depth now, possibly wrongly, since my audio interface is "only" capable of 24 bit depth A/D, conversion, so as you mention a 119 dB analog dynamic range, so for e.g. the A/D conversion at a point in time determines the analog level amongst the 119 dB range in 16.7 million possible steps,  or into 0.0000007 dB  steps of accuracy between quietest and loudest levels? (32 bit ~ 4 billion steps, if you had an AI capable of this?).
 
....  But more often does not sound as good as its not capable of reproducing the smaller nuances created with effects or VI's you hear or sense when listening to your 96K/64/32 bit depth (or actually "only" 24 bit depth from AI) mix or master wave files on studio monitors, as opposed to your processed 44K/16 bit depth production in your car or elsewhere. I find the final "bit rate" used for MP3 significantly affects the final sound quality, possibly more so than a lot of the other things along the way.  
....

Again, a couple of different concepts got conflated here:
 
The 16-bit (96dB) or 20-bit (120dB), or 24-bit (144dB) capability of the converter gets stretched out over the real-world noisy analog circuitry, with is able (on a good day with the wind at its back) able to give about 120dB of dynamic range.  The signal-to-noise ratio of a typical condenser microphone will be in the 84dB range, so its noise floor isn't exactly plumbing the depths of what a converter can do.  
 
The MP3 'bit rate' that you mention is a completely different thing than the bit depth and sampling rate of pulse-code-modulated wave files, and it's absolutely reasonable that you'd hear reproduction problems with MP3's that are encoded at less than 320kbps.  Keep in mind that the whole MP3 strategy is to 'throw away' significant portions of the musical content while minimizing the acoustic harm it causes.  A regular stereo CD wav file is 1141kbps equivalent bitrate, while the maximum (and least harmful) MP3 bitrate is 320kbps. 

mister happy
Unless you conduct a dither listening test... then it turns out nobody can hear the prescence or absence of dithering in a 16bit file even if the playback levels are calibrated to -23DBFS pink noise = 105dBSPL C weighted at listening position.

I agree, words are useless without some tests. So:
http://www.azslow.com/files/NoDither2.wav
http://www.azslow.com/files/WithDither2.wav
 
Before anyone start to wonder why I have uploaded 2 files with silence...
You will have to drive your card really hard (or chain 2 buses with +6 dB in Sonar) to hear something.
But once you manage to get normal level of the sound, the difference will be obvious. And that is the whole point:
clearly demonstrate what dithering does, without "hi end" equipment, "golden ears" or any cheating. Both files are exported from Sonar, 16bit 44.1 kHz from the same just recorded n00b clip (with prior lowering the clip gate). One with and another without dithering.
Initially I have made clips with ~10dB higher level. With studio headphones no blind test required to hear the effect, but with $5 phones in my notebook the effect was not so obvious.
 

1.  Uh, you should avoid upsampling audio, because you are adding something that isn't there, not making it somehow (ha ha) "better".
 
2.  When is the last time you performed a frequency-based hearing test?  Unless you are in your twenties, I doubt you can hear 18kHz, and there's a high probability you can't get to 13kHz, especially if you've spent a decade or two rockin' out live.  If you are extended-range-deaf (likely true), you fool yourself going back to the notion of "better".
 
3.  Most everyone listens to [and sometimes actually buy] mp3's.  64 bit cobbled down to an mp3 accomplishes ... what?
 
4.  24 bit 48kHz - in my fantastically self-over-rated opinion - is all you need, a good compromise between disk space, cpu load, artifact avoidance, and what you and your patrons might actually be able to hear. 

Hey, thanks to the OP for asking the question and thanks for all the experts that chimed in.
 
I thought I understood some of this, but I think I owe myself a refresher course on Digital Audio
 
I hope these discussions don't get erased or lost if the cake servers go away eventually. I would hope they could find a more permanent home somewhere on the "Intertubes."
 
Take care,
h

hbarton
Hey, thanks to the OP for asking the question and thanks for all the experts that chimed in. 
 

Yea thanks, I appreciate the "feedback" I hesitate to ask sometimes for all the info my brain has to now process, LOL.
These recent Izotope posts also help,
 
https://www.izotope.com/en/community/blog/tips-tutorials/2018/01/understanding-audio-signal-flow-in-a-daw.html?utm_medium=Email&utm_source=ACS&utm_campaign=2018+Hi+Five&utm_content=Jan+13&utm_term=All
 
https://medium.com/@dannyanthony/the-advantages-of-a-floating-point-system-in-a-daw-b45fd9a4efc2
 
cheers
 
 

.

mister happy

azslow3
...words are useless without some tests. So:
http://www.azslow.com/files/NoDither2.wav
http://www.azslow.com/files/WithDither2.wav
Before anyone start to wonder why I have uploaded 2 files with silence... ...

This is a specious example that may lead innocent bystanders to draw conclusions that have no practical application.

That is an example how dithering affects the sound. Nor more, nor less.
Sorry to say, but what you provide are just WORDS. Without mathematical, physical aural explanation. You just claim: 15 bits are sufficient for any sound at any fidelity level.
 

A test I like to use for this example is a cymbal crash followed by its complete tail to silence. You can either calibrate your playback as I suggested at the onset or just turn the levels up so that the cymbal's crash gives you a headache. Then you can listen intently for the final bit fluttering in the silence. Set up a pair of files and take an ABX test using one the ABX testing apps. When you are done you will never worry about dithering again.

If you have done that and do not worry, fine!
You "like" testing with cymbal. That is the worse way to test any noise is audible: loud initial sound and the whole sound is just a noise by design. Take a gun and shoot it into air every time before listening music, you can save a lot of money on your equipment then
 

.

If you listen to your musical content at common playback levels, For example; a playback level calibrated to -23dBFS pink noise = 72dBSPL C weighted at the listening position then the least significant bit in the digital stream will be fluttering well below the SPL threshold of hearing. 
 
That is very simple math.

That is not math at all. That is proposed loudness (only) condition for test.
 
Math is in fact very simple:
* 16 bit without dithering at all effectively leave 15 "true" bits. That means ~-90dB SNR
* 16 bit with simple dithering keep all 16 bits meaningful. So ~96dB SNR.
Without doing anything fancy, Dynamic range is equivalent to these numbers. Fancy dithering algorithms claim perceived dynamic range increase up to 120dB.
 
Is that level of noise always meaningful and can be spotted? Definitively not. There can be at least several examples what can make it not noticeable by definition (any from the list):
* calibrated listening environment in which that level is under ATH
* when the environment has background noise comparable with SPL from this noise
* when the signal already has more noise then that, so the content is "noisy" by design. Including f.e. drums and "vintage warm" processors or emulations (tubes, tapes)
* when the signal is overall maxed and dynamic range is overall small
 
But you claim it NEVER make any sense. And that part is questionable:
* modern equipment can technically reproduce such level, unlike f.e. 64bit AD/DA
* there are many SoftSynth which produce "crystal clean" signal. They sound "cold" and "airy", with tiny "noise" component".
* hobbyists rarely achieve perfect mastering and so to not squash things completely leave quite big dynamic range in the final file. Also some classic CDs do not "ride" the volume by intention, there can be quite tracks. All that move "listeners" to turn clockwise the volume knob on Hi-Fi/End amplifier, effectively moving the noise level away from "callibrated" listening conditions.     
 
The math is just as simple if you listen to your music more loudly.
 
So, using your words... Assuming you have done "everything right" in recording and mastering your cymbals, I completely agree that dithering is not important.
 
But for us, n00bs, assuming that is just a check-box in the export dialog, it is good to set it. But avoid it during intermediate processing since that make no sense and can hit hard our low-end computers.