Digital Audio Explained

[@confusionprice1422]

This is the only video I could find that explains the difference between the different bit depths, thank you

[@tomraj]

I'am scared...

[@sltechnology2273]

This was the only the video I could find to realize the diffrent between sample sizes. After 11 years, huh?

[@thefauvel7558]

Higher sample rate isn't always better quality. There is no point in 192kHz for example, unless we were able to hear more of the spectrum, because you don't benefit from it.
Signal sampling is a bit different than resolution, so is bit depth, which only affects the noise floor.
The conversion between continuous signal to discrete signal and back to continuous does not make the wave different, or jaggy, nor creates stairsteps, therefore the information in-between samples is not lost in any way, in fact it is reproduced perfectly.
The sample points(discrete signal) are the path an analog wave will go through(exactly like in the original source).

[@thefauvel7558]

As long as we are above Nyquist we will get perfect reconstruction of the original wave. Quantization will, of course, create distortion later on.

[@thefauvel7558]

+CaesarGamer The waveform is not affected by the bit depth.... 24bit is unnecessary.
The only thing bit depth will affect is the noise floor and with shaped dithering and 16bit the noise will be very much reduced. 24 bit will generate noise that is well below what the human ear can hear, just like 16bit.

[@thefauvel7558]

This video is quite misleading

[@thefauvel7558]

+MrAudioSoundImages I don't see where digitizing is producing stairsteps. The samples(discrete signals) are merely values at a determined point(resembling pixels), the data in-between samples is not lost(it's not constant either), it's just undefined before playback.
You are correct when you say higher sample rates have a benefit(not for audio fidelity), but 96kHz or 192kHz is honestly useless and it will take up space more than anything else.
While your video is very intuitive and good, it lacks the emphasis on how the wave is reconstructed, D-A, and the sampling process, which may cause a lot of confusion.
It might look like, because of the lack of emphasis on those matters, for an average Joe, that a digital output will never be able to produce a wave perfectly(it's going to be all jaggy and squared) unless you have infinite sample rate, which is fundamentally not true, because the samples are discrete signals, rather than continuous (it does not have a value defined at each and every point/not a continuous time function).

[@thefauvel7558]

+MrAudioSoundImages Your video seems to imply that you will get better audio reconstruction and higher details, therefore fidelity, if you crank up the sample rate and bit depth. That is simply not true if what you are trying to record is audible frequencies and you are above Nyquist frequency. You aren't gonna get more details in your track at audible frequencies if you have 48kHz for example. At 44.1kHz all the human hearing spectrum is covered and the waves are reconstructed perfectly, all according to Nyquist theorem, which also explains why you don't need to have infinite sample rate to get the perfect reconstruction of a sound wave.
Higher sample rates have other benefits, but so long as you are above Nyquist you are not going to get better details by having higher rates.

[@Illogical.]

1:02 wrong number on screen.
Also, any number of bits technically could be used. But multiples of 8 are easier to work with, and more accessible.
Also, and this is just me brainstorming, it probably would be reasonable to use the difference between the current slice and the previous slice instead of the full value of the current slice, since adjacent slices will pretty much always have values that are pretty close to each other. This would in some cases remove data.
I do not know enough about audio hardware to know to what extent this would work before it would start losing significant amounts of data.
Also, with this relative approach, using the first few bits of the values as how much to leftshift the lower bits by will allow for a better range, at the cost of lowered accuracy across nearly all ranges. But that likely wouldn't be a big issue in most cases. (It's slightly more complicated than I explained, cause there are several ways to take care of the sign bit.)

[@dylanm.3692]

I always imagined it like video, where it's cut up into frames.
In each "frame", there are many packets of data that encode an individual sine wave's frequency and amplitude, and these sine waves are added together to get something that would approximate the actual waveform for a tiny fraction of a second. Like how one picture approximates a video for a fraction of a second.
It makes sense in my head, but maybe it would sound godawful in practice, lol.

[@Badassvidsz]

Thanks Mr Audio

[@xardas1500]

thanks for the conceptional explanation.. gives me a breath in of the whole project i'm working on. carry on like this

[@bekhacker]

nice

[@DRAGNIL68]

thanks! this vid helped a lot

[@prabinojha8274]

0:48
horizontal chunks*

[@beakf1]

hello do you have any videos explaining dynamics and noise floors of digital vs analogue. Im trying to learn what the metres in cubase 5 represent.

[@stevenswall]

This is not correct. The sample rate determines what frequencies you can capture, not how closely they represent the real thing. When you draw a sine wave through enough dots that there is only one correct solution, adding more dots doesn't make the line you drew any smoother. The computer needs enough samples to draw the curves, but no more, unless you're trying to capture a higher frequency and need more dots to draw the curve.

[@prep74]

Mr Audio does not understand digital audio - he just regurgitates a pervasive myth (usually spread by vinyl fanboys).
It is not a series of steps but an accurate sound wave curve. Sampling rates do not "smooth" the curve but effect the frequency of the sound which is captured.
Read the article below for an accurate, and properly referenced, explanation.
people.xiph.org/~xiphmont/demo/neil-young.html

[@prep74]

MrAudioSoundImages At least you are open-minded unlike some of the other clips on youtube. The pixelisation analogy is not useful in understanding digital audio - well at least when you are sampling at a Nyquist rate greater than the human hearing range - as it creates misunderstandings such as the steps myth. There are artefacts outside the Nyquist sample frequency range, mainly due to aliasing but these would normally appear as noise and are filtered out by the DAC.
The bit rate only effects the sound floor. CD is 16 bit and is all you need for signal to noise ratio that is superior to an analogue medium. 24 bit is overkill for playback but useful for production (ie mixing and mastering).

[@prep74]

MrAudioSoundImages Yep, agreed, for recording and production you do want more than 16bit. For playback though, it adds nothing. Btw, the link below is good reading. It is from a UK hi fi store that sells both analogue and digital equipment. It's not meant to be a plug for the store but the text under the heading "Slaying Dragons" makes some good points from a practical listeners/users perspective.
www.fwhifi.co.uk

[@prep74]

MrAudioSoundImages I came across this thread in the headhifi forum regarding 16bit vs 24bit. Thought you might find it interesting. Cheers
www.head-fi.org/t/415361/24bit-vs-16bit-the-myth-exploded

[@paulanderson79]

Slight;y over simplified. CD audio, as with all other forms of digital audio, is NOT a sample and hold function. Sample rate and bit depth do not work in this way.

[@frederick98]

typing MISTAKE about 16 bit

[@SAbdullah03]

too much info in a minute and a half

[@zakali3284]

*Comment Removed

[@zakali3284]

i find it too oversimplified as a GCSE student but my friend went a little overboard as he made the comment, i agree but would have worded it different

[@zakali3284]

MrAudioSoundImages MrAudio: i am an extremely big fan of your videos sorry for being an ass and "thank you mr audio"

[@zakali3284]

MrAudioSoundImages really sorry for being an ass back there i guess my friend went a bit overboard. i do understand your intent to educate. me a and my friends thoroughly enjoy waiting for your new videos

[@ramizhossain9082]

Again maths.

[@gunnarliljas8459]

This is highly inaccurate