[TheSonicTruth]

If you record carefully and don’t require a lot of post processing in the
mix, 16/44.1 is more than enough. It’s better than most 24 track analogue
decks are capable of, so any analogue master is going to be more hobbled
than digital ones.

I'm genuinely surprised to be hearing such from you.

 

[gregorio]

[1] I can reliably hear the differences between 24/44, 24/96, and 24/192. Not to mention DSD Files. Dynamics, and noise are pretty consistently what stands out between all these files, less noise on higher quality files, better dynamic range.
[2] DSDs sound more analogue, [2a] "like I am there" than any other format.
[3] If I was you, I would get my ears checked, or invest in better gear.

1. There is no dynamic or noise differences between any of those formats, they are all have identical dynamic range and noise, except DSD which has significantly less dynamic range than any of the others. So how can you reliably hear differences which do not exist?

2. That's a bit of a sad indictment, do you really think DSD sounds as bad as analogue?
2a. I don't understand, if you think analogue sounds "like I am there", why do you think they invented digital audio instead of just sticking with vinyl or cassette tapes?

3. You're the one "reliably hearing" differences that do not exist, so shouldn't it be you getting your "ears checked"?

One misconception though is that the CD is DDD. It is actually DAD as it was mixed through an analog work station.

The CD codes were rather misleading. DDD doesn't mean an album was mixed digitally, the 3 letters ALL refer to the recording of the 3 stages: The first D indicates the audio was recording in digital, the second D indicates the mix was recorded in digital and the third D indicates the master was recorded in digital. So, you can mix and master with an analogue desk/equipment and it's still DDD (rather than DAA) providing you record the analogue mix and master in digital and indeed, this is the case with pretty much all DDD rock/popular music genres up until the millennium.

[1] Thank you for that history, but my question remains: When did higher sample rates and bit depths start to be used in production?
[2] When you record something in digital, then apply processes(EQ, comp, etc) digitally to it,
[2a] it's like a recording of a recording of a recording, and quantization noise and other artifacts slowly accumulate.
[3] That is why 24bit and higher bit depths, and 92kHz and higher sample rates are used in the production process before exporting down to 16/44.1 for consumer mass-distribution.

1. 32kHz, 44.1kHz and 48kHz were the standard for many years, 96kHz started to be widely available to studios in the early 2000s. 20bit recording started to be employed in the early 1990s and 24bit again in the early millennium.

2. You're confusing "recording something in digital" and mixing ("applying processes") in digital. At the beginning of the 1990's (and even a few years earlier) there were digital mixers that processed internally at around 28bits but they weren't particularly common. By the late 1990's professional digital and virtual mixers were 32bit float or 48bit fixed, today they're mostly 64bit float.
2a. Not really. There are no other artefacts besides quantisation noise and that's defined by the bit depth of the mix environment. So with 48bit fixed, the quantisation noise is roughly at -288dB and around -370dB with 64bit float. You literally need hundreds or thousands of processors for quantisation noise to accumulate to anywhere near audible levels. This is independent of the recording bit depth, these quantisation noise figures are the same regardless of whether the recorded audio being mixed is 24bit or 16bit.

3. There are and have never been any 24bit professional mixers as far as I'm aware. Sampling rates are a bit more complicated; 48kHz is by far the most common sampling rate used professionally/commercially, as it's the worldwide standard for film and all SD and HD television. There are few situations where a higher sample rate is beneficial when processing but this is again independent of the recorded audio sample rate and even independent of the sample rate of the mix environment, the processor will internally upsample and downsample if necessary. The only legitimate use for recording at 96kHz or higher sample rates is for some sound effects design purposes and for scientific purposes (recording insects or bats for example). However, 96kHz and 192kHz are reasonably common as recording formats for music products because clients may want to market a "hi-res" version at some stage.

G

 

[Arjey]

Does this mean there is absolutely NO difference between 16/48 and 24/48?

 

[castleofargh]

Does this mean there is absolutely NO difference between 16/48 and 24/48?

The quantization noise rises to 16bit, and depending on the playback gear, you might see some slight increase in distortions. Both should remain inaudible under typical music listening, and normally do, as suggested by most listening tests.

 

[gregorio]

Does this mean there is absolutely NO difference between 16/48 and 24/48?

To elaborate on what castleofargh stated: Yes, there is a difference, 16bit has a 48dB higher noise floor than 24bit, which is defined by dither (noise). This puts the noise floor of 16bit at around -92dB, which is audible under laboratory and mastering studio conditions using very high (uncomfortably loud) listening levels, tests indicate roughly 14dB higher than comfortable levels. However, it's been standard/recommended practice to use noise-shaped dither (rather than standard dither) since before 24bit recording was even available to recording/mastering studios, which reduces the noise floor of 16bit by a further 25dB or so (to roughly -120dB). Therefore, the digital noise floor of 16bit as defined by (noise-shaped) dither is roughly 40dB below what would be audible under laboratory conditions when listening at comfortable levels (and 40dB is 100 times)!

G

 

[TheSonicTruth]

To elaborate on what castleofargh stated: Yes, there is a difference, 16bit has a 48dB higher noise floor than 24bit, which is defined by dither (noise). This puts the noise floor of 16bit at around -92dB, which is audible under laboratory and mastering studio conditions using very high (uncomfortably loud) listening levels, tests indicate roughly 14dB higher than comfortable levels. However, it's been standard/recommended practice to use noise-shaped dither (rather than standard dither) since before 24bit recording was even available to recording/mastering studios, which reduces the noise floor of 16bit by a further 25dB or so (to roughly -120dB). Therefore, the digital noise floor of 16bit as defined by (noise-shaped) dither is roughly 40dB below what would be audible under laboratory conditions when listening at comfortable levels (and 40dB is 100 times)!

G

But obviously there is some benefit to using higher specs on the production side, or else they would not have bothered to start doing it, right?


One will more quickly raise the noise floor if the entire production process - RMM(Recording sessions, Mixing, Mastering) is done at 16/44.1 than at 24/92 or higher before exporting to consumer formats, correct?

Sort of analogous to a wedding photographer capturing the event portraits at the highest raw megapixels he can afford, and his equipment is capable of, before any touch-up processing is done and smaller 8x10s, 5x7s, or wallet-sized copies of the special day's photos are produced.

 

[bigshot]

Does this mean there is absolutely NO difference between 16/48 and 24/48?

No audible difference for the purposes of playing back recorded music in the home. The differences are purely technical and would have no practical application for how you would be using music files.

 

[71 dB]

But obviously there is some benefit to using higher specs on the production side, or else they would not have bothered to start doing it, right?

On the production side the benefit is you don't need to optimize the dynamic range in everything you do and you know the noise floor won't explode to your face no matter how much you process the sound. ===> Clear benefits on the production side, but no benefits on consumer side (apart from "hi-res" placebo effect).

 

[71 dB]

The quantization noise rises to 16bit, .

If one is to create a 16 bit version of a 24 bit file it would be "dumb" to just truncate the 24 bit data to 16 bit and get 16 bit quantization noise. Instead people use dither noise and instead 16 bit quantization noise they get 16 bit dither noise and these two are very different things. Quantization noise correlates with the signal and causes non-linear distortion. Dither noise doesn't correlate with the signal at all and avoids non-linear distortion. Even having more power, dither noise appears less annoying to human ear and shaping the spectrum of dither can make it spychoacoustically much quieter. In fact shaped dither can achieve perceived dynamic range so large that one can call such 16 bit consumer audio total overkill by several bits.

We should also remember that producing music at larger than 100 dB dynamic range is not trivial. Good luck recording a singer at 120 dB dynamic range without noise! The recording boot is not noise free. The microphone pre-amp is not noise free. There will be noise significantly above the 24 bit quantization noise. Ironically this noise acts as dither and helps keeping things linear and distortion free. So, the 24 bit recording may have a noise floor at level not much below 16 bit dither meaning going from 24 bit to 16 bit doesn't mean increasing noise by 48 dB. Using 24 bit allows flexible use of the bits. For example, the four highest bits can be "allocated" to 24 dB headroom while the lowest 4 bits ensure the quantization happens well below the noise floor of the whole recording system.

 

[gregorio]

[1] But obviously there is some benefit to using higher specs on the production side, or else they would not have bothered to start doing it, right?
[2] One will more quickly raise the noise floor if the entire production process - RMM(Recording sessions, Mixing, Mastering) is done at 16/44.1 than at 24/92 or higher before exporting to consumer formats, correct?

1. On the recording, mixing and mastering side yes, there is some benefit, on the reproduction side, no audible benefit at all. For example, during recording 24bit allows for plenty of headroom, which is useful because we obviously don't know what the peak level is going to be before the musician has actually performed but we do know exactly what the peak level of the master is, so we don't need any headroom.

2. If the mixing and mastering were done at 16bit the noise floor would theoretically increase more quickly than if it were done at 24bit but even at 24bit it would increase unacceptably, which is why professional mixers have never been 24bit devices (let alone 16bit). Even the early ones were about 28bit but by the time there were widely used they were 32bit float or 48bit, as I've already explained just a few posts previously.

G

 

[71 dB]

Does this mean there is absolutely NO difference between 16/48 and 24/48?

For consumers placebo effect is all "difference" there is. People may think 24/48 sounds better than 16/48 because intuition says more bits must mean better fidelity, but 16 bits is already a few bits more than we actually need in consumer audio (the threshold is ~13 bits).

 

[Arjey]

So, if I have a bunch of music in 24 bit, and I only use it for listening, no mixing or anything.. Is it safe to say that a can just convert everything to 16 bit (to save some space) and I won't hear a difference?

And a different question.
24 bit can go up to 192khz. I know most people don't here even over 15khz. I personally can hear up to ≈19khz (I'm pretty young, 19 y.o., maybe that's why I hear so high frequencies and hate sibilance).
Then there are low frequencies, people don't really hear 1-4hz as much as feel the vibration.
So my question is, say I have Really good headphones, a Really good source, and a Really good recording.. will I be able to "feel" the difference between 48 vs 192khz? I just want it to sound as natural and realistic as possible.

The real world has unlimited (almost) frequencies, thought we can't even hear them.
But for example: dog whistles. We can't here them, but there where a couple of time I was walking with a friend in a park with his dog, and I was at a distance, and he just blew the whistle (I was looking in a different direction), but for some reason I just know that "something" is different, and I look in that direction. And there he is, blowing a whistle that I can't hear.
Something like a "sixth sense"..
So what I'm asking is, will converting a 24/192 or DSD to 16/48 "remove" that slight sixth sense of airyness or realism?

Sorry for the weird way of putting this question, to which you guys most likely don't have an answer, because it's absurd, and the answer will most likely be: "depends on what you believe.." but still. I think is an interesting thing to think about)

 

[71 dB]

1 --- So, if I have a bunch of music in 24 bit, and I only use it for listening, no mixing or anything.. Is it safe to say that a can just convert everything to 16 bit (to save some space) and I won't hear a difference?

2 --- And a different question.
24 bit can go up to 192khz. I know most people don't here even over 15khz. I personally can hear up to ≈19khz (I'm pretty young, 19 y.o., maybe that's why I hear so high frequencies and hate sibilance).
Then there are low frequencies, people don't really hear 1-4hz as much as feel the vibration.
So my question is, say I have Really good headphones, a Really good source, and a Really good recording.. will I be able to "feel" the difference between 48 vs 192khz? I just want it to sound as natural and realistic as possible.

3 --- The real world has unlimited (almost) frequencies, thought we can't even hear them.
But for example: dog whistles. We can't here them, but there where a couple of time I was walking with a friend in a park with his dog, and I was at a distance, and he just blew the whistle (I was looking in a different direction), but for some reason I just know that "something" is different, and I look in that direction. And there he is, blowing a whistle that I can't hear.
Something like a "sixth sense"..
So what I'm asking is, will converting a 24/192 or DSD to 16/48 "remove" that slight sixth sense of airyness or realism?

1 --- Yes. You can even convince yourself about it by producing that 16 bit version and then upload both the 24 bit version and the 16 bit version on a wave editor such as Audacity, invert the 16 bit version and mix it with the 24 bit version so that you get the difference (make sure your result is 24 bit or higher) and the try to hear the result. You'll notice at you hear nothing unless you turn the volume up crazy and that is the difference signal only. It is totally masked when you listen to the music! That's when you go "of course I can't hear the difference between 16 bit and 24 bit.

2 --- People can only sense low frequencies (such as 1-4 Hz) but around 16 Hz starts the sensation of sound (combination of both). Sampling rate affects the upper limit of frequency range so the difference between 48 kHz and 192 kHz sampling rates is above human hearing range and doesn't matter. 48 kHz is enough and so is 44.1 kHz. The truth is 192 kHz sampling rate is totally pointless in consumer audio. If you study the ultrasonics of bats then you need 192 kHz, but music listening is not about bat science. Whatever you "feel" between 48 kHz and 192 kHz is probably due to placebo effect that affects your perception because you "believe" 192 kHz must sound better. When you have the knowledge of digital audio and what is enough you'll be able to remove the effect of placebo.

3 --- Maybe you knew "something" is different, because you heard the change in the behavior of the dog who did hear the whistle? Maybe dog whistles produce some sound in the human hearing range, say in the range of 10-20 kHz? I have to admit I don't know much about dog whistles. Listening tests suggests that downsampling hi-res audio shouldn't remove anything relevant to human hearing.

According to Wikipedia "To human ears, a dog whistle makes only a quiet hissing sound." and "some [dog whistles] are adjustable down into the audible range."

 

[Arjey]

1 --- Yes. You can even convince yourself about it by producing that 16 bit version and then upload both the 24 bit version and the 16 bit version on a wave editor such as Audacity, invert the 16 bit version and mix it with the 24 bit version so that you get the difference (make sure your result is 24 bit or higher) and the try to hear the result. You'll notice at you hear nothing unless you turn the volume up crazy and that is the difference signal only. It is totally masked when you listen to the music! That's when you go "of course I can't hear the difference between 16 bit and 24 bit.

2 --- People can only sense low frequencies (such as 1-4 Hz) but around 16 Hz starts the sensation of sound (combination of both). Sampling rate affects the upper limit of frequency range so the difference between 48 kHz and 192 kHz sampling rates is above human hearing range and doesn't matter. 48 kHz is enough and so is 44.1 kHz. The truth is 192 kHz sampling rate is totally pointless in consumer audio. If you study the ultrasonics of bats then you need 192 kHz, but music listening is not about bat science. Whatever you "feel" between 48 kHz and 192 kHz is probably due to placebo effect that affects your perception because you "believe" 192 kHz must sound better. When you have the knowledge of digital audio and what is enough you'll be able to remove the effect of placebo.

3 --- Maybe you knew "something" is different, because you heard the change in the behavior of the dog who did hear the whistle? Maybe dog whistles produce some sound in the human hearing range, say in the range of 10-20 kHz? I have to admit I don't know much about dog whistles. Listening tests suggests that downsampling hi-res audio shouldn't remove anything relevant to human hearing.

According to Wikipedia "To human ears, a dog whistle makes only a quiet hissing sound." and "some [dog whistles] are adjustable down into the audible range."

Thanks for the detailed reply! I guess I'll just stick with 48khz, though I don't really hear a difference between 48 and 44.1, but just to be safe) Also most films and many games use 48, and it's easier on the system (if it's old, at least) to keep playing the same sample rate.
Here's some interesting and detailed material on the matter. https://forums.stevehoffman.tv/threads/longish-why-i-prefer-48-khz-over-44-1-khz.347515/

 

[old tech]

Thanks for the detailed reply! I guess I'll just stick with 48khz, though I don't really hear a difference between 48 and 44.1, but just to be safe) Also most films and many games use 48, and it's easier on the system (if it's old, at least) to keep playing the same sample rate.
Here's some interesting and detailed material on the matter. https://forums.stevehoffman.tv/threads/longish-why-i-prefer-48-khz-over-44-1-khz.347515/

I like the SH forum, mainly for finding consensus on the best masterings of a CD or LP.

There are some very knowledgeable people who post in that forum but also a lot of misinformed woo, that 48 v 44 thread is a case in point. A year ago there was a thread comparing 44.1 vs hi res (ie anything from 24 bits to DSD) and the overwhelming response was that no-one could hear a difference if the same master was used under blind level matched tests. What was surprising was that the thread ended up being over 5 pages of good rational discussion but the mods then deleted the entire thread - probably because Steve Hoffman is trying to break back into the industry with AF SACDs, (which is based on a lie that the format itself sounds better than CDs). They lost a lot of good members after that thread deletion.

That forum also seems to attract a lot of vinylphiles, digiphobes and the downright irrational. So take care with information from that site - check the info with the more evidence based sound science forums such as this sub forum or at https://hydrogenaud.io/. Another good resource which contains many blind tests and discussions is at http://archimago.blogspot.com/

With your point about the dog whistle, bare in mind that they vary quite a bit in frequencies - some start at 20khz and others go up to 23 to 25 khz. So it is possible that you could have heard it but I tend to agree more with 71 DB, that your perception was subconsciously more around the reaction of the dog and/or a fleeting sight of the whistleblower in your periphery. You really need a controlled test to be sure.