Almost no difference between 16 bit and 24 bit?

[theprodguy]

I bounced a 24 bit and a 16 bit wav of the same project. Imported into another project, inverted the polarity of one of them (or inverted the phase, as some erroneously say). They canceled completely.

 

Same thing between a 16 bit dithered file and a non dithered file. This doesn't surprise me as much, however, I was expecting some differences.

 

Is it normal?

[fisherking]

uh...what? just curious, what are you trying to accomplish?

24bit definitely sounds better than 16bit...more information. check reverb tails, definition...etc.

 

you seem to be doing a lot of work to what end?? make some music (better use of time IMHO).

 

[theprodguy]

Trying to understand why things work the way they do is a pretty good use of time IMHO.

[fisherking]

sorry, did not mean to be rude. just wondering what you're trying to accomplish. a little research online, and you'll know about the discernible differences between 16bit and 24bit recording. so was just suggesting you could be making music in the same time...

 

whatever u need to do, enjoy it. that's what this is all about...

[theprodguy]

a little research online, and you'll know about the discernible differences between 16bit and 24bit recording.

I'm not talking about recording at 24 bit, though.

 

Well, after doing some more reading I'm kinda confused... some people say that listening to 24 bit audio makes a difference, others say it's a myth (http://www.head-fi.org/t/415361/24bit-vs-16bit-the-myth-exploded). It's strange though, because everyone says that the CD standard is old and wants higher quality music.

[Eriksimon]

The most important difference is in dynamic range. This means that for most pop-, dance- and rockmusic, which tend to have dynamic ranges of less than 12 dB sometimes, the difference doesn't matter much. 16 bit has a theoretical dynamic range of 96 dB, where 24 bits has a th. range of 144 dB. In reality this means that you can have very detailed sound at low levels with 24 bits. where 16 bits would suffer (which is (also) why there's such a lot of compression used, just to stay in the upper regions of the dynamic range of 16 bit.)

So, the differences really count in very dynamic music, such as jazz, classic, and some popmusic.

Another big difference is in recording: 24 bits can be recorded at much lower levels while still "out-dynamic-ranging" 16 bit, meaning that the risk of clipping inputs is much smaller to nonexistent at 24 bit.

[theprodguy]

Thanks for answering.

 

Yes, I understand the benefits of recording at 24 bit.

 

But for playback, someone says there are audible differences between 16 and 24 bit audio, others say nobody can hear those differences. In all my tests, also from other people, files cancel out almost completely (some noise at about -70/80 dBFS), yet someone claims he can tell them apart. I'm wondering how it is possible if the two files are basically identical.

[Eriksimon]

16 bit and 24 bit are not "basically identical".

[theprodguy]

When did I say they sound the same? I quoted someone else who said that and you didn't quite understand it. In fact, the post never stated that "you cannot hear the difference because all music has a dynamic range of only 12 dB these days".

 

He said that "you can't [hear the difference between a 16bit commercial recording and a 24bit Hi-Rez version], it's not that you don't have the equipment or the ears, it is not humanly possible in theory or in practice under any conditions".

 

But if you're only going to quote what you want and don't answer to any question I'm asking, maybe you shouldn't post at all.

[Eriksimon]

When did I say they sound the same?

 

The title of the topic contains the phrase "almost no difference." Should I have written "almost sound the same" then?

[theprodguy]

The title of the topic contains the phrase "almost no difference." Should I have written "almost sound the same" then?

The title of the topic is a question indeed.

 

So, in case you wanted to answer, my question is, if a 24 bit file and a 16 bit file cancel out completely when inverted, or however leave so little noise that it's not hearable in any environment at normal listening levels, am I not right supposing that they sound exactly the same except for the noise (that you can't hear in any way)? If so, how could a person guess right 9 times out of 10? There has to be something that makes the 24 bit different...

[Eriksimon]

Yes, 8 extra bits.

[theprodguy]

Sure, and what for? Any difference between the two waveforms would be evident when phase-inverted... If they cancel, they're the same. So the sound isn't benefitting at all from those 8 bits.

[steveo12ax7]

Makes sense to me that 24 bit imparts most of its benefits in the recording stage, and not after the audio has been digitized. There is more than enough headroom in 16 bit to play back audio, as long as it has been well recorded.

 

Unless there is a problem with your null tests, I would tend to agree that if they cancel, they are identical. I don't really have a hard time believing that they would cancel either - like I said, if the audio is well recorded without noise floor issues there isn't any reason why 16 bit wouldn't reproduce them just fine.

 

I'm a bit more skeptical about the dither tests though. I've surprised myself by being able to pick out dithered clips vs truncated ones reliably so anecdotally I believe there is a difference.

[shivermetimbers]

Using 24bit over 16 bit will allow for lower amounts of noise distortion when processing the audio,

One example would be compression and different bit depths.

 

Even if the 24/16 nulled out signal resulted in some sound, it would not directly reflect the quality of the signal.

You have heard that using the highest bit rate and bit depth will produce the best results. However, you also heard that the extra storage doesn't outweigh the benefits.

 

You can use analyzers to detect the amounts of distortion present in each signal, but your ears are not geared to judge better from worse.

 

Stick with the 24/44.1k method and don't over think it. You have more important things to do with your life.

[Matt Mayfield]

I bounced a 24 bit and a 16 bit wav of the same project. Imported into another project, inverted the polarity of one of them (or inverted the phase, as some erroneously say). They canceled completely.

 

Same thing between a 16 bit dithered file and a non dithered file. This doesn't surprise me as much, however, I was expecting some differences.

 

Is it normal?

 

Yes. Bits are like digits.

 

1.2 is a number with 2 digits of precision.

1.2000000 is a number with 8 digits of precision.

 

If you convert the number 1.2 from 2-digit format to 8-digit format, you get 1.2000000.

 

If you subtract 1.2000000 from 1.2, you will still get 0, even though you converted one of them to a different "bit depth."

[Eriksimon]

 

Yes. Bits are like digits.

 

1.2 is a number with 2 digits of precision.

1.2000000 is a number with 8 digits of precision.

 

If you convert the number 1.2 from 2-digit format to 8-digit format, you get 1.2000000.

 

If you subtract 1.2000000 from 1.2, you will still get 0, even though you converted one of them to a different "bit depth."

 

 

I think the difference here is accuracy.

In math (= in theory), 1.2 is as exact as 1.20000000000000, literally AND figuratively there's no difference between the two.

In measurements (which an audio file is: a collection of measurements) on the other hand:

1.2 means anywhere between 1.15 and 1.25 (or ± 0.1)

1.20 m.a.b. 1.195 and 1.205 (or ± 0.01)

1.2000000 m.a.b. 1.19999995 and 1.20000005 (or ± 0.0000001)

 

Same principle applies to bitdepth: the more bits, the more accuracy, the better signal reconstruction.

[theprodguy]

Yes, this is actually something I don't get.

 

I've read in various sources that in 16 bit audio, the information in the top 96dB of dynamic range is just as accurate as in a 24 bit file, and the only difference is the dynamic range.

 

From the first link I've posted:

So, 24bit does add more 'resolution' compared to 16bit but this added resolution doesn't mean higher quality, it just means we can encode a larger dynamic range. This is the misunderstanding made by many. There are no extra magical properties, nothing which the science does not understand or cannot measure. The only difference between 16bit and 24bit is 48dB of dynamic range (8bits x 6dB = 48dB) and nothing else. This is not a question for interpretation or opinion, it is the provable, undisputed logical mathematics which underpins the very existence of digital audio.

 

Is this technically correct? I don't understand, then, why the 24 bit file has 2^24 (16,777,216) steps in which that range is divided, compared to the 2^16 of the 16 bit file. Isn't this supposed to represent any level more accurately?

 

From Wiki

The number of possible values that can be represented by the bit depth can be calculated by using 2x, where x is the bit depth. Thus, the resolution of a 16-bit system (2^16) is 65,536 and a 24 bit system (2^24) has a resolution of 16,777,216. PCM audio data is typically stored signed, in two's complement, so a 16-bit audio sample represents a decimal number from -32,768 to 32,767 and a 24-bit sample represents a decimal number from -8,388,608 to 8,388,607.

 

Thanks for your answers in advance.

[Eriksimon]

Have you noticed, about the dynamic ranges of 16 vs 24 bits audio?

96/144 = 16/24 That's no coïncidence.

 

The higher bit depth is not used for greater accuracy over the same range, it is used for the same accuracy over a greater range.

[theprodguy]

Hmm, ok. But back to the example of digits.

 

An 8 digit value is more accurate than a 2 digit value and these values can be in the same range. For example, 1.2 and 1.2340000 can be both in the same range but one is represented with more accuracy than the other one.

 

If the same accuracy is being used, just on a higher dynamic range, why do the possible values increase exponentially (from 2^16 to 2^24)?

[Eriksimon]

because bits are exponential, as are decibels.

 

I think this wiki explains it well:

http://en.wikipedia.org/wiki/Audio_bit_depth

 

In short: as long as your recording is well within the dynamic range of the 16 bits convertors (which is in practice smaller than in theory), it will also be in range of the dynamic range of 24 bits convertors, so differences between the two files will be minimal. The real differences will show if you record a wider dynamic range than 16 bits can proviode, then you will hear a difference between the roiginal 24 bits file and the downsampled 16 bits file. The clearest difference is in the quiet(-er) bits.

[theprodguy]

But when I do bit truncation, what happens? If I truncate a value, I'm still in the same range as before, only approximating it...?

[Eriksimon]

Don't do bit truncation.

Why not? Because it's boring!

Tickle some keys instead!

Really, just believe everybody: 24 bits is better than 16! Why? More is more. Why? Because the actual 21 bits range it physically manages (material aka real world limitations), provides a dynamic range that matches that of the human ear. Now all that is needed is a major cultural "retrovolution" that brings back really, really quiet bits in all genres of modern music. Why? Because we (finally) can! If the standard for digital audio (redbook CD) is upped to 24 bits... the simple fact is that 16 bits PCM is still what all audioplayers use, so feeding it a 24 bits file is useless, and producing a 24 bit file is, except for a few audiophiles with insanely expensive high end 24 bits capable HIFI systems, a very niche thing still.

I really think a 24 bit standard (like DVD-A) would be a big step forward towards a more natural sound, with much more expressive possibilities. And less ear damage, too.

[David Nahmani]

I've read in various sources that in 16 bit audio, the information in the top 96dB of dynamic range is just as accurate as in a 24 bit file, and the only difference is the dynamic range.

That's correct. The more bits, the more dynamic range.

 

Here's how it works:

 

Let's say you're sampling audio at 48kHz. That means every 48,000th of a second you measure the amplitude of your analog audio waveform and store it as a digital value called a sample. Now depending on how many bits I give you to encode the value of a sample, you're going to make an error when measuring the value: if I gave you a ruler graduated in feet and asked you to measure my height in feet and give me a number in feet, your error would be greater than if I gave you a ruler in inches and asked you a number in inches.

 

So every time you measure the analog waveform you're making a rounding error. And the more bits you use to describe that value in the digital world, the smaller the rounding error.

 

Now think of the sample as:

 

sample = analog value + rounding error

 

So if I gave you a ruler in feet and asked you to measure me, and I know I really measure 5 feet 9 inches, and you tell me I measure 6 feet, you're giving me my real height plus a mistake of 3 inches, that's 5 feet 9 inches + 3 inches.

 

Now picture the digital signal composed of all the compounded mistakes. For example if you measured 3 people, for me the mistake was 3 inches, maybe for the 2nd person the mistake is -2 inches and for the 3rd person it's +4 inches. That gives you a new series of a values, which can be used to display a new waveform.

 

→ That new waveform is what your A/D converters are summing with your analog signal.

→ That new waveform is made of unwanted, compounded calculation errors.

→ That digital noise is called quantization noise.

→ The amplitude of the quantization noise determines the dynamic range available for your signal.

 

As you can imagine, the greater the amplitude of the noise (the less bits you use), the higher the noise level, therefore the less dynamic range you have.

 

Hope that helps?

[theprodguy]

So every time you measure the analog waveform you're making a rounding error. And the more bits you use to describe that value in the digital world, the smaller the rounding error.

 

But then why is it correct to say that "the information in the top 96dB of dynamic range is just as accurate as in a 24 bit file"? Isn't that also subject to rounding errors? I don't know how to explain myself better, so I'll quote a post from Head-Fi.org

 

as we increase the bit depth we not only increase the dynamic range [...] but we do increase the accuracy of the audio. [...] with a 16bit word length we can express ~65,000 different values, which means we express ~65,000 different levels of amplitude. I hear you ask, but what happens if the analog signal falls between one of the these ~65,000 values? Well, that's where 24bit world length come in, it can express over 16 million different levels of amplitude. So as you can see as we increase the bit depth of the audio we increase the dynamic range but we also increase the accuracy of the digital waveform

 

This to me sounds right but in contradiction with the statement that "the information in the top 96dB of dynamic range is just as accurate as in a 24 bit file, and the only difference is the dynamic range".

[David Nahmani]

But then why is it correct to say that "the information in the top 96dB of dynamic range is just as accurate as in a 24 bit file"?

 

Think of increasing bit depth as lowering the noise. The loud stuff is just as accurate, but the softer stuff becomes audible all of a sudden.

 

dynamic range = clipping level - noise level

 

more bits = less noise = more dynamic range

 

less bits = more noise = less dynamic range

[theprodguy]

But if in 24 bit audio you can express 16 million values of amplitude and in 16 bit you can express 65000 values. Let's say in 24 bit audio you have a value that equals say decimal number 23456. When you truncate to 16 bit, you're going to approximate that value. So, that value is not the same in 24 and 16 bit audio...

[David Nahmani]

But if in 24 bit audio you can express 6 million values of amplitude and in 16 bit you can express 65000 values. Let's say in 24 bit audio you have a value that equals say decimal number 23456. When you truncate to 16 bit, you're going to approximate that value. So, that value is not the same in 24 and 16 bit audio...

Correct. Both values are (sample = analog value + rounding error) only the rounding error is larger in 16 bits.

[theprodguy]

See where I get confused? If this approximated value is within the first 96dB of dynamic range, you can't say anymore that those 96dB are just as accurate in the 16 bit file. Because there's larger rounding error...

[David Nahmani]

You're playing an acoustic guitar in a room, and someone is sitting next to you listening to you. The room is completely silent and he can hear your guitar very, very well. That's the equivalent of a 24 bit audio recording.

 

Now I sit next to the guitar player and softly start blowing while pursing my lips: "pfffffffffff". You can barely hear the white noise I'm making. That's the equivalent of the 16 bit audio recording.

 

Now the guitar signal hasn't lost any clarity or precision or detail, it's above the noise threshold so it's still the exact same guitar sound. Only in the 16 bit recording if the guitar player plays too soft or start playing really really pianissimo you will no longer hear what he's playing because of me blowing air, vs in the first case you'd have still heard it.

 

Does that help?