[bigshot]

The funny thing is that the music most HD audio nuts are listening to at these astronomical bitrates is stuff that was considered ephemeral rock music back in the 60s and 70s, so it was recorded quick and on the cheap in the first place. Just today I had a guy on another forum tell me that he didn't understand how digital audio worked and didn't care. He just uses his ears. But whenever he talks about music, it's always in a high data rate format. I think his ears must tell him to pick big numbers.

 

[71 dB]

I think his ears must tell him to pick big numbers.

Not his ears of course, but his mind.

 

[castleofargh]

admittedly, the less you understand how digital audio works, the more likely you are to imagine that bigger is always better. the mistake at least seems logical. the guy who can't visualize audio not being analogue is always going to wish to fill in those "holes" in the music. or have smaller stair steps, or whatever faulty way he uses to try and force digital into a faulty analog representation.
if I knew nothing about digital audio, I believe I would have such ideas.

 

[bigshot]

admittedly, the less you understand how digital audio works, the more likely you are to imagine that bigger is always better.

It also helps to not believe that digital sound isn't "perfect". When you believe that everything is flawed in some way, then any kind of improvement of those flaws must result in better sound. It's a carry-over from the analogue era where every format and component added its own noise, coloration and distortion. The theory back then was to strip back to just the essentials, split components into as many specialized individual boxes as possible, and avoid anything that alters the signal, even if it is correcting imbalances. None of those things are necessary any more.

 

[Oncisounds]

Hi, bit late to this whole party here but a very interesting article! However I have a question if I may and they might sound very silly so I apologise in advance however;

You said that the greater the bit depth the less quantisation errors right? The correction for quantisation errors is to introduce dither which in turn introduces noise into the audio recording in order to get a true a reading of the waveform as possible, this means the less bit depth the more dither and in turn the more noise. With that I figured the higher bit depth the better not necessarily for dynamic range but more so for less noise being introduced. I figured noise would be present in audio at all dB or is it only at the noise floor that this noise is present?

Sorry again if it sounds like a silly question still trying to understand this all.

 

[castleofargh]

Hi, bit late to this whole party here but a very interesting article! However I have a question if I may and they might sound very silly so I apologise in advance however;

You said that the greater the bit depth the less quantisation errors right? The correction for quantisation errors is to introduce dither which in turn introduces noise into the audio recording in order to get a true a reading of the waveform as possible, this means the less bit depth the more dither and in turn the more noise. With that I figured the higher bit depth the better not necessarily for dynamic range but more so for less noise being introduced. I figured noise would be present in audio at all dB or is it only at the noise floor that this noise is present?

Sorry again if it sounds like a silly question still trying to understand this all.

the noise from quantization is determined by the amount between the correct value and the approximation due to quantization, meaning it's going to be quiet noise in the least significant bit. with dither it's going to depend on the type of dither used. but in general, withing the audio band we're still talking about the least significant bits or even lower than that with noise shaping.

obviously, getting higher bit depth is objectively good for the noise floor. or it would be if quantization noise was among the loudest noises you'd get in your audio chain and listening environment. which might not be the case all that often. you'll need pretty clean gears, very quiet room, and to listen to music loudly, just to possibly make quantization noise the loudest noise of all. and even then, there is no guaranty that you'll detect it. a lot of abx tests between 16 and 24bit suggest we usually don't.
the relevance of increasing bit depth is really a matter of context. as far as audio playback is concerned, the noise floor of 16bit is already pretty cool.

 

[71 dB]

Hi, bit late to this whole party here but a very interesting article! However I have a question if I may and they might sound very silly so I apologise in advance however;

You said that the greater the bit depth the less quantisation errors right? The correction for quantisation errors is to introduce dither which in turn introduces noise into the audio recording in order to get a true a reading of the waveform as possible, this means the less bit depth the more dither and in turn the more noise. With that I figured the higher bit depth the better not necessarily for dynamic range but more so for less noise being introduced. I figured noise would be present in audio at all dB or is it only at the noise floor that this noise is present?

Sorry again if it sounds like a silly question still trying to understand this all.

There is nothing wrong asking when you don't know!

Yes, the less bit depth the more dither noise you need, but the dither noise is constant in level. It doesn't matter if there's silence in the music or if the music is blasting off Dither is always the same. It's like having a fan blowing air while listening to music. The noise from the fan is constant. However, in 16 bit audio dither is so quiet you can't hear it. Dither becomes much more important when we go under 16 bits (8 bits for example). In those cases a very crappy badly granulating distortied sound if replaced with undistorted sound with heavy noise which is much better.

 

[Oncisounds]

There is nothing wrong asking when you don't know!

Yes, the less bit depth the more dither noise you need, but the dither noise is constant in level. It doesn't matter if there's silence in the music or if the music is blasting off Dither is always the same. It's like having a fan blowing air while listening to music. The noise from the fan is constant. However, in 16 bit audio dither is so quiet you can't hear it. Dither becomes much more important when we go under 16 bits (8 bits for example). In those cases a very crappy badly granulating distortied sound if replaced with undistorted sound with heavy noise which is much better.

Ahh okay that makes sense, see in my head the less quantisation the less dither you would need to correct for it hence the less noise in the audio interfering. But now I know it's a constant and always present but basically inaudible it kinda makes a lot of sense, same sort od argument as refresh rate in monitors etc.

 

[gregorio]

Ahh okay that makes sense, see in my head the less quantisation the less dither you would need to correct for it hence the less noise in the audio interfering.

Yep, that is correct. The point/facts you were missing was how relevant and audible that dither noise is. When we record something, the recording venue always has noise, the mics used always add noise, the mixing and mastering processes add noise and then when you reproduce the recording, your amp adds noise, your HPs or speakers add noise and your listening environment has noise. All these sources of noise add different amounts of noise, for example a good amp will generally add very little noise, usually less than 16bit dither noise but in most other cases dither noise is lower or much lower than any one of the other noise sources, let alone all those other noise sources combined. As mentioned, it is sometimes possible to hear dither noise but the conditions required to do so are not realistic. First you have to pick a recording that has particularly low levels of recording, mixing and mastering noise, your system and playback environment have to have very low levels of noise and then you have to play that recording at very loud levels. Even if you have such a recording/s and reproduction system/environment, then the human ear adjusts itself for low noise and for comfort sake one would listen at a lower level than normal rather than much higher than normal. In other words, unless you manufacture the unreasonable conditions specifically to hear dither noise of 16bit, it is always going to be inaudible and irrelevant.

Hope this makes sense?

G

 

[Oncisounds]

Yep, that is correct. The point/facts you were missing was how relevant and audible that dither noise is. When we record something, the recording venue always has noise, the mics used always add noise, the mixing and mastering processes add noise and then when you reproduce the recording, your amp adds noise, your HPs or speakers add noise and your listening environment has noise. All these sources of noise add different amounts of noise, for example a good amp will generally add very little noise, usually less than 16bit dither noise but in most other cases dither noise is lower or much lower than any one of the other noise sources, let alone all those other noise sources combined. As mentioned, it is sometimes possible to hear dither noise but the conditions required to do so are not realistic. First you have to pick a recording that has particularly low levels of recording, mixing and mastering noise, your system and playback environment have to have very low levels of noise and then you have to play that recording at very loud levels. Even if you have such a recording/s and reproduction system/environment, then the human ear adjusts itself for low noise and for comfort sake one would listen at a lower level than normal rather than much higher than normal. In other words, unless you manufacture the unreasonable conditions specifically to hear dither noise of 16bit, it is always going to be inaudible and irrelevant.

Hope this makes sense?

G

Yeah it makes perfect sense I figured that would be the case, hence why the noise floor is so high, as that's the levels where it would start to become apparent. But thank you for expaining it all, it makes a lot of sense that noise from all aspects of recording to processing to digital output will cause noise of similar magnitude if not even more so than that of the dither.

Thanks guys!

 

[Haohua]

thank you for the knowledge.

 

[PanzerIV]

Wow this is one of the most interesting read I've done. I might be almost 10 years late but this is still relevant info that anybody should know. At least now I know my ears aren't wrong as I was like "What I can't hear any tiny difference AT ALL between regular files and those higher bit super rare and expensive Hi-Res albums?!" but noticing any difference would only have been placebo effect. When you have in $CAD so remove 33% value, an headphone of 2700$ (Audeze LCD3), a 2000$ DAC (NAD M51), a 1600$ headphone amp (Woo WA6-SE) and 550$ cables (Audioquest NRG-Y3 + Mackenzie RCA), you should have stuff good enough to hear a difference if there's one as it's already better than what 99.9% of people have at home!

I remember reading somewhere here, that recording techniques at the studio was faaaar more important than the actual Bit/Khz numbers and I totally believe that. It's pretty much imo like those cheap 200$ phones that say it records at 1080p blablabla or huge MegaPixel but with absolutely crapty and tiny integrated lens. Trust me I could record even in 720p or take a picture at 3x lower MP than ur phone with my DSLR camera and professional lens with super good optical glass, and it's gonna look infinitely better. As like anything in life, quality over quantity! People really don't realize that 1920x1080 is only a tiny 2.07MP! Then they think their 40MP cellphone is gonna be Nasa quality lol.

 

[Slaphead]

Wow this is one of the most interesting read I've done. I might be almost 10 years late but this is still relevant info that anybody should know. At least now I know my ears aren't wrong as I was like "What I can't hear any tiny difference AT ALL between regular files and those higher bit super rare and expensive Hi-Res albums?!" but noticing any difference would only have been placebo effect. When you have in $CAD so remove 33% value, an headphone of 2700$ (Audeze LCD3), a 2000$ DAC (NAD M51), a 1600$ headphone amp (Woo WA6-SE) and 550$ cables (Audioquest NRG-Y3 + Mackenzie RCA), you should have stuff good enough to hear a difference if there's one as it's already better than what 99.9% of people have at home!

I remember reading somewhere here, that recording techniques at the studio was faaaar more important than the actual Bit/Khz numbers and I totally believe that. It's pretty much imo like those cheap 200$ phones that say it records at 1080p blablabla or huge MegaPixel but with absolutely crapty and tiny integrated lens. Trust me I could record even in 720p or take a picture at 3x lower MP than ur phone with my DSLR camera and professional lens with super good optical glass, and it's gonna look infinitely better. As like anything in life, quality over quantity! People really don't realize that 1920x1080 is only a tiny 2.07MP! Then they think their 40MP cellphone is gonna be Nasa quality lol.

Now try the same thing with varying levels of MP3 and AAC compressed audio against the original 16/44 - I'll bet you'll get to higher levels of compression before you hear an obvious difference than you thought you would - although I will say that some people are more sensitive to the artefacts that compression brings than others are.

When I say that I don't mean hearing ability, it's just that I believe the standard psychoacoustic model used in compression fits better to some people than it does to others. The psychoacoustic model used in compression is a one size fits all approach, and thus there will be people that lie outside of that which model dictates.

You're absolutely right about recording techniques - you'd be lucky to get to 60dB (or 10 bits) of dynamic range in even the best setup studio. However that's not to say that hi-res audio isn't useful - in fact it's essential in the production phase where the music is mixed and then ultimately mastered as hi-res audio gives you the headroom in which to apply levels, EQ and other effects without introducing audible quantisation errors. However for end point delivery 16/44 is all that's needed, in fact it's more than is needed.

 

[PanzerIV]

Now try the same thing with varying levels of MP3 compressed audio against the original 16/44 - I'll bet you'll get to higher levels of compression before you hear an obvious difference than you thought you would - although I will say that some people are more sensitive to the artefacts that compression brings than others are.

100% Agree with what you said that I took off from this quote, as yes I can totally understand this extra headroom is needed for the mixing phase, but not to us final user listening at it.

I'm just not entirely sure to understand though what you meant in that quote above. Did you mean that I could be surprised how low the bitrate of an MP3 needs to be BEFORE I finally start noticing artifacts? To be honest, I've never noticed even the tiniest difference between (CBR 320) and (VBR 0) but as I found the idea of having much more efficiency by having a file size lower with 0% loss of quality, I converted all CBR320 albums I downloaded or bought, into VBR0. However if I'm lucky enough to get a Flac file, now that I got the equipment to listen at it, I'll just keep them that way as these days storage cost absolutely nothing. It's sad though that Flac file are still very rare and you either must buy the album and sometime they still don't offer Flac, or go on specialized websites. I would never spend money into buying MP3 albums, that's for sure!

 

[gregorio]

[1] ... but noticing any difference would only have been placebo effect. ... I remember reading somewhere here, that recording techniques at the studio was faaaar more important than the actual Bit/Khz numbers and I totally believe that.
[2] It's pretty much imo like those cheap 200$ phones that say it records at 1080p blablabla or huge MegaPixel but with absolutely crapty and tiny integrated lens.

1. To be pedantically accurate, noticing any difference may not ONLY be due placebo effect. The reason for this is that your second sentence is true, recording, mixing and mastering are far more important and there are quite a few cases where the CD version of a recording has different mastering than the hi-res version and therefore it can be relatively easy to tell the difference in some cases. As there is no intrinsic audible difference between CD and hi-res, there is the temptation to reduce the quality of the CD version and justify the higher price of the hi-res version. In some cases where audiophiles believe they're noticing a difference, it is as you say placebo effect but in other cases they maybe noticing a real, actual difference but it's between different masters, not the audio file format.

2. We have to be a bit wary of analogies with video and audio. HD in video (even 720p) is technically AND VISUALLY higher definition than standard definition. When greater than 16/44.1 audio became available, the audiophile marketing world jumped on the "HD" video bandwagon, but in order for the term "HD" (or "Hi-res") to have any meaning they had to call CD (16/44.1) "standard definition". This was effectively a lie/false marketing because CD is NOT anywhere near equivalent to standard definition video. Relative to the capabilities of human vision and human rearing, CD is in fact higher resolution/definition than even the latest 4K HDR video specification! Once we have a video specification that exceeds ALL the capabilities of human vision, then higher definition than that would be pointless and that's the situation we have with CD audio.

[1] You're absolutely right about recording techniques - you'd be lucky to get to 60dB (or 10 bits) of dynamic range in even the best setup studio.
[2] However that's not to say that hi-res audio isn't useful - in fact it's essential in the production phase where the music is mixed and then ultimately mastered as hi-res audio gives you the headroom in which to apply levels, EQ and other effects without introducing audible quantisation errors.

1. That's not always the case, some of the best studios can manage a dynamic range of over 90dB. It's VERY expensive though, you're looking at double-shell construction and other expensive isolation strategies/treatments to get the noise floor of the studio very low and then a particularly good sound system capable of high output levels (and low noise/distortion).

2. We have to be careful here and make the distinction between the "mix environment" and the "audio file format" absolutely CLEAR. Modern professional mix environments are typically 64bit (float), all processing ("levels, EQ and other effects") occurs at this bit depth and therefore even cumulative quantisation errors are way below our ability to even reproduce, let alone hear. However, this mix environment is completely INDEPENDENT of the bit depth of the audio file format we load into it. Whether we load a 16bit file or 24bit file into the mix environment is completely irrelevant as the processing (and quantisation errors) will all be at the 64bit level. So as far as headroom for processing during mixing and mastering is concerned, it's EXACTLY THE SAME with 16bit and 24bit audio files. The ONLY place where a 24bit file format can make any difference is for additional headroom during recording.

G