24bit vs 16bit, the myth exploded!

[sander99]

Digital clipping


This PCM waveform is clipped between the red lines

In digital signal processing, clipping occurs when the signal is restricted by the range of a chosen representation. For example, in a system using 16-bit signed integers, 32767 is the largest positive value that can be represented. If, during processing, the amplitude of the signal is doubled, sample values of, for instance, 32000 should become 64000, but instead cause an integer overflow and saturate to the maximum, 32767. Clipping is preferable to the alternative in digital systems—wrapping—which occurs if the digital processor is allowed to overflow, ignoring the most significant bits of the magnitude, and sometimes even the sign of the sample value, resulting in gross distortion of the signal.

Directly bouroughed for fair use EDUACATIONAL PURPOUSESSES: https://en.wikipedia.org/wiki/Clipping_(audio)

TheRe yUh gO.
duh doi.

weapons drawn: wIkiPeDiA

Indead the more you post the clearer it gets you understand practically nothing of all this stuff (or are deliberately playing the fool).
A digital signal can clip (or wrap if you like) yes. But that has nothing to do with the bit depth.
Your drawing suggests that you think that for example 24 bit audio is converted to 16 bit by removing the 8 most significant bits. That is nonsense. One correct way would be to remove the 8 least significant bits. And maybe apply dither to the result. In general: of course you convert - or produce a new track - in such a way that the result fits in 16 bits without clipping, which is always possible. And as explained by G the only thing that changes is the noise floor, and because the noise floor of a 16 bit signal is not audible except when putting the volume ridiculously high it doesn't matter. (And even if you set it ridiculously high you probably still won't hear the noise floor because your hearing will adapt to the loud levels and become - temporarely if you are lucky - less sensitive).

I have 800,000 songs of space on my 32GB phone SD card.

But this thread is about 24 bit versus 16 bit uncompressed PCM. It doesn't say anything about massively compressed MP3. Indeed, if that is what you are talking about that is something else entirely. And that could sound bad yeah. So you are barking up the wrong tree here, it is not the fault of 16 bit uncompressed PCM/WAV.
Like @chef8489 already said.

 

[Lazysnakes]

Indead the more you post the clearer it gets you understand practically nothing of all this stuff (or are deliberately playing the fool).
A digital signal can clip (or wrap if you like) yes. But that has nothing to do with the bit depth.
Your drawing suggests that you think that for example 24 bit audio is converted to 16 bit by removing the 8 most significant bits. That is nonsense. One correct way would be to remove the 8 least significant bits. And maybe apply dither to the result. In general: of course you convert - or produce a new track - in such a way that the result fits in 16 bits without clipping, which is always possible. And as explained by G the only thing that changes is the noise floor, and because the noise floor of a 16 bit signal is not audible except when putting the volume ridiculously high it doesn't matter. (And even if you set it ridiculously high you probably still won't hear the noise floor because your hearing will adapt to the loud levels and become - temporarely if you are lucky - less sensitive).


But this thread is about 24 bit versus 16 bit uncompressed PCM. It doesn't say anything about massively compressed MP3. Indeed, if that is what you are talking about that is something else entirely. And that could sound bad yeah. So you are barking up the wrong tree here, it is not the fault of 16 bit uncompressed PCM/WAV.
Like @chef8489 already said.

hmm. I can see that.
still



with loud enough difference, you can hear the difference in higher frequencies, and with loud enough 1db to 155db changes in music.
if you aint into 155db car audio, your loss bud. the human ability to recognize the difference, and the complete and totally irrelevant idea that the hardrive space it takes up is ridiculous with 1-32 terabyte hard drives, by 2025 to be replaced with 32 terabyte drives shackles with 16bit in hi-res is pointless. might as well have 24 and even 32 bit with over 700khtz, whatever-the-spec.

in otherwords, humans can hear the differnce, probably, but regardless, we should not dismiss 24 or 32bit. it should be embraced as standard or the future of car audio, home and public speaker audio (including wave clipping in theaters with explosions) and used for self-increasing audio volume changes and sampling for dog training and child training (10-40khtz) etc.

 

[chef8489]

He has been messaging me and still doesn't get it.
"
I don't understand why you are saying this. I have a 32bit file and a 16bit file, (no 24) the 32 bit sounds louder without distorting, period.
the 16 distantly lacks the punch.
you could master it and then rack the volume up higher, but then the intro will sound way too loud."

"favorite.
with loud enough difference, you can hear the difference in higher frequencies, and with loud enough 1db to 155db changes in music.
if you aint into 155db car audio, your loss bud. the human ability to recognize the difference, and the complete and totally irrelevant idea that the hardrive space it takes up is ridiculous with 1-32 terrabyte hardrives, by 2025 to be replaced with 32 terabyte drives shackles with 16bit in hi-res is pointless. might as well have 24 and even 32 bit with over 700khtz, whatever-the-spec."

 

[sander99]

@Lazysnakes:
That video you posted is exactly the kind of marketing bull that we like to warn people for here in the Sound Science forum. Like someone before me nicely put it: MQA is about unfolding dollars from your wallet.

 

[bigshot]

I'm sorry, I've had my allotment of Dunning Kruger for this month. Have a nice day, troll. Same old tricks I see.

 

[castleofargh]

@Lazysnakes
I have tried to understand your posts and failed. Your reasoning often doesn't seem to make sense based on my understanding of digital audio and decibel.
You need to make an effort so that we can have a chance to follow your explanation. but my best guess so far is that you have some profound misconceptions about how things really work.
When you have an amplitudes encoded in PCM, if you have 3bits, 111 will be 0dB(the loudest amplitude you can record). all the other permutations will be for lower amplitude values. Now if you have 8 bit PCM, 11111111 will be 0dB. With 8bit we have many more permutations possible, allowing to encode much quieter signals. But the maximum any given bit depth allows is always referred to as 0dB! Then the DAC might turn that 0dB amplitude into 2volt, or maybe 2,5volt, or maybe closer to one volt for portable devices. There is no rule about that.
0dB is only relevant as being the highest amplitude in digital, and it's an important reference for the sound engineers who want the music to be recorded at less than 0dB(to avoid clipping). and it's important for us if we use some EQ or other DSP to boost a signal already close to 0dB. That's when clipping would occur and it would be bad. Which is why most EQs have a global gain slider to keep our EQ from actually boosting the signal in the digital domain, as it's not something we can do. We don't have bit permutations to go above a line of ones AKA 0dB.

It's fair to consider that bit depth defines how quiet the quantization noise will be(assuming no dither). And really it doesn't do much else. Bit depth most certainly does not decide how loud a signal will be, as again, no matter how many bits you use, the maximum will still be the 0dB digital reference. It can seem arbitrary but that's how digital audio is handled.

 

[sander99]

if you have 3bits, 111 will be 0dB ... if you have 8 bit PCM, 11111111 will be 0dB.

The core of your story is correct. 0 dB is the max value.

However to be completely correct:
For PCM the samples are coded in two's complement, see link below. In that coding scheme 011 and 01111111 would be the respective maximal positive values,
and 100 and 10000000 would be the minimal (or max negative) values.
Actually 100 (-4) and 1000000 (-128) are 1 larger in absolute value than 011 (-3) and 0111111 (-127), so I assume they would normally never be used in a 3 and 8 bit digital signal respectively.

https://en.wikipedia.org/wiki/Two's_complement

 

[castleofargh]

The core of your story is correct. 0 dB is the max value.

However to be completely correct:
For PCM the samples are coded in two's complement, see link below. In that coding scheme 011 and 01111111 would be the respective maximal positive values,
and 100 and 10000000 would be the minimal (or max negative) values.
Actually 100 (-4) and 1000000 (-128) are 1 larger in absolute value than 011 (-3) and 0111111 (-127), so I assume they would normally never be used in a 3 and 8 bit digital signal respectively.

https://en.wikipedia.org/wiki/Two's_complement

I admit that I was thinking about physical R2R gates in a DAC, because it's the easiest way to model(at least in my mind) the passage from binary(switches) to actual voltage amplitude.
Never hesitate to point out when I'm full of crap, it benefits me too.

 

[bigshot]

I guess when we respond, it gives them attention. Their name is mentioned. They get email notification that someone has replied. They see multiple people take time out of their day to explain the same idea in a range of ways from simple and direct to complicated and full of footnotes and numbers. It isn’t to their benefit to understand because the second they acknowledge it, the attention stops. The internet brings out interesting behavior in people.

 

[old tech]

I think part of @Lazysnakes issue is that he references that pseudoscience Hans beek you tube videos as a source.

 

[Lazysnakes]

I don't see a reason, if even humanly indistinguishable, to limit audio from 16bit instead of hi res files. With terabyte and exabytes of storage there is no reason to waste time mixing and tuning, when you can simply store thousands of 700mb 5 minute songs with little impact on your drive.

In a previous edit I made a serious mistake. There are micro sd cards right now eith 500GB of storage. This is physical storage.
You can then use arithmetic or algorithm coding to compress files further a d depending on the genius that invents the code, this can be quite substantial. Lossless compression isnt the same as a file that is readable quickly,, but as it continues to advance and distantly is more advanced the the oldness of this forum O.P. post I'd say handicapping your files if you produce them yourself, not necessarily buying 24bit, but production should still be 24 bit depth to 32 bit depth.

Compression using 2025 algorithm coding or even 2020 vs the O.P. date, should be enough to store millions of songs or thousands of songs 50MB to 125MB in length or even 32 bit depth or about 500 MB in physical length on mobile SD cards with 500 GB of storage or even just 32MB. This storage power makes any argument for 24 bit depth being pointless and limiting your library invalid.

And I certainly don't understand why one would say the files are damaging to audio quality.

 

[bigshot]

This is going in circles. Here is an article that goes into great detail about how 16/44.1 is all you need to accurately reproduce everything human ears can hear.
https://web.archive.org/web/20200426202431/https://people.xiph.org/~xiphmont/demo/neil-young.html

 

[dazzerfong]

All micro sd cards are limited to computer chip space of only about 32kbytes. They have no physical storage beyond that, your 32gb sd card can store songs compressed, which can then be read later.

I don't see a reason, if even humanly indistinguishable, to limit audio from 16bit instead of hi res files. With terabyte and exabytes of storage there is no reason to waste time mixing and tuning, when you can simply store thousands of 700mb 5 minute songs with little impact on your drive.

I have no idea why you pulled up allocation unit size - that has zero bearing on the discussion here. It also isn't limited to 32 KB: it's adjustable when you format it. You select a size depending on if you have many small files or some big files. Also, compression isn't what's making storage a non-issue: there's a limit to how much you can compress (research entropy) and SD cards just keep getting bigger.

No-one's arguing against keeping large files. They're just arguing for the audibility of it over a standard Redbook file.

In other words, sure it doesn't really harm it, but you can't hear the difference anyway. So while on my NAS I have a 24-bit 192 kHz HDTracks version, on my DAP which has 500GB I'm gonna scale it down to 320 kbps so I can have 20,000 songs on it.

 

[Lazysnakes]

I have no idea why you pulled up allocation unit size - that has zero bearing on the discussion here. It also isn't limited to 32 KB: it's adjustable when you format it. You select a size depending on if you have many small files or some big files. Also, compression isn't what's making storage a non-issue: there's a limit to how much you can compress (research entropy) and SD cards just keep getting bigger.

No-one's arguing against keeping large files. They're just arguing for the audibility of it over a standard Redbook file.

In other words, sure it doesn't really harm it, but you can't hear the difference anyway. So while on my NAS I have a 24-bit 192 kHz HDTracks version, on my DAP which has 500GB I'm gonna scale it down to 320 kbps so I can have 20,000 songs on it.

Why? It's like arguing that lossy compression is okay or a good thing even if-even then, 99% of audiophiles can still hear no difference?
Why stop there? Why not cap your highs to your hearing limit? If your 35 just cap your songs to 16khtz?

It does not go along with what the same people argue for lossless files.

And I standby the idea, although I concede the argument, some people can still use the difference in signal.

 

[dazzerfong]

Why? It's like arguing that lossy compression is okay or a good thing even if-even then, 99% of audiophiles can still hear no difference?
Why stop there? Why not cap your highs to your hearing limit? If your 35 just cap your songs to 16khtz?

It does not go along with what the same people argue for lossless files.

And I standby the idea, although I concede the argument, some people can still use the difference in signal.

Lossy compression is a good thing for convenience. The folks who invented the principles of lossy compression were geniuses. What's achieved now with lossless codecs is a good compromise between portability and hitting the threshold of audibility.

What you're using now is called the slippery slope fallacy. Don't. The reason that's a bad idea is that it means that you'll have to re-encode all your files as your hearing ages.

I personally vouch for lossless files because I'm a data hoarder, but day to day I don't use said lossless files. FLAC doesn't universally support replay gain across all my devices whereas MP3 gain works no matter what.

Difference in signal between 16 bit and 24 bit? You literally cannot. It's like saying what's the difference between 1010000000000000 and 101000000000000000000000. Answer is 8 zeroes that were never going to be used in the first place anyway.