# Why would 24 bit / 192 khz flac sound any better than 16 bit / 44.1 khz flac if both are lossless (if at all)?

Discussion in 'Sound Science' started by thesuperguy, Mar 15, 2014.

1. I wonder if you are mixing up sample rate and bit depth.

A true 24/192 file is not oversampled but simply recorded with a word length of 24 bits and a sample rate of 192. This means it has a dynamic range of 144 dB and the highest possible frequency is 96 kHz (1/2 Fs)
CD players use oversampling (8 times)
If you don't, the first alias starts right after 1/2 Fs (22 kHz)
You need a very sharp (brick-wall) filter to get rid of it.
If you over-sample 8 times (353 kHz), the first alias is at 176 kHz so you can use a very smooth filter.

To the best of my knowledge DSP doesn't profit by increasing the sample rate.
In fact a lot of DSP chips are limited to 48 kHz and the more expensive to 96 kHz simply because the higher the sample rate the shorter the time one have to perform the calculations.
In case of 44.1 you have 0.00002 of a second to do all the calculations.

DSP does profit by increasing the precision. Doing all calculations with 32 or better 64 bit keeps down the quantization error.

As DACs only accept integers, as DSP uses a precision much higher that 16 or 24 integer, in the end any DSP action has to be dithered.
Here a 24 bit recording has the advantage as with  16 bits the dither is at -96 dBFS but with a 24 at -144 dBFS
You have to play FFF loud to make something at -144 audible if possible at all.

To sum up, DSP is about doing calculations
This has nothing to do with sample rate except that sample rate might be a limiting factor (time)
This has all to do with precision, the more bits the better.

2. well the 24 bit has better signal to noise when recording.  I wish it was 40 bit like the digital audio workstation programs (but for a different reason).

3. Ah okay, I am not the most knowledgeable in this subject. So I am assuming we are talking about how 24 bit has a higher dynamic range so therefore it should theoretically clip less when EQing if I were to exceed that 144dB threshold? I don't listen to my music very loud, usually just 8 or 9 o'clock low gain on my matrix mstage hpa2 classic (the dial starts at 6 o' clock). Most metal songs I listen to had an average dynamic range of 15-20dB on my music analyzer plugin with around 30dB peaks. The highest dynamic range I've gotten out of a track was around 52dB peak for a classical piece. Neither of them are exceeding the threshold of a 16 bit music file if I am not mistaken. I think maybe compression has something to do with the clipping and a song that uses compression will have a greater chance of clipping even if it does not surpass 96dB. And where do I find a peak meter? My music player is jriver which allows plugins

4. well it isn't clipless. back when protools and nuendo came out The mixing enviroment was 24 bit. it didn't take long for then to change the mixing environment to 32 bit because things clip too easily when mixing inside the program. A few years ago the environment went to 40 bit.

when someone likes me mixes, we create the enviroment. Nothing really has changed here compared to the days of an analog desk and tape drives.
Then it goes to the master gain shredder (mastering engineer).

Sound card using consumer audio drivers have about 6 db of headroom on the master. On asio drivers there is no extra headroom.

btw when the audio music is transferred to cd media,  they bring it out to analog and clip the input of a modified converter. The signal is more treated as a percentage in a deflection scale than a audio signal.

oops I need to add some more:

ok I don't know where you are measuring, but any readings you are getting after conversion will vary depending on your converter.
btw those analyser plugins are not intended for measurements. They are not accurate. maybe you should try a piece of pro audio software. This one was originally written by Sony: http://www.magix-audio.com/us/spectralayers-pro/?utm_source=sonycreativesoftware&utm_medium=referral&utm_campaign=redirect&lang=en&prdt=spectralayerspro

5.
1. Generally that's true, although there are some DSPs out there which do. Some non-linear DSPs such as some compressors and modelling plugins for example. Also, it very much depends on the programming of the DSP and any design considerations/limitations. I have come across DAW plugin DSPs which sounded better operated at 96kHz than at 44.1kHz, simply because that was the sample rate which seems to have attracted the most programming effort. That doesn't appear to be quite as common today as it was a few years ago and probably doesn't affect many consumer DSPs.

2. True but only up to a point. In DAWs, the cumulative quantisation error of successive DSPs can theoretically be an issue but we'd have to use quite a few dozen DSPs to get to that stage with 24bit.

1. 24bit doesn't have a signal to noise ratio, it just has a lower noise floor than 16bit. This lower noise floor does not directly affect the SNR when recording, as the SNR is defined by the mics, mic-preamps and the SNR of what's being recorded and where.

2. You could make it 512bit if you want, it still won't affect the SNR though.

3. Why are you coming here and just making stuff up? ProTools was 16bit when it came out, then it's mixing environment changed to 56bit about 20 years ago (Protools TDM), the cheaper LE versions of Protools started at 32bit float and all today's versions are 64bit float. Nuendo started at 32bit float and is now 64bit float. No current DAWs I'm aware of have a 40bit mix environment.

4. No they don't! I'm sure the odd person might have done what you're suggesting, years ago, but it's extremely atypical today!

1. No, you're looking at this backwards. 0dBFS is the maximum level a digital audio file can have and this is exactly the same maximum with 24bit as with it is with 16bit. The added dynamic range of 24bit is at the quietest end of the scale, not the loudest. So you are just as likely to clip at 24bit as at 16bit when EQ'ing. Commercial music releases are optimised to 0dBFS (or just fractionally below), so you will need to lower the (digital) level of the file before adding EQ, otherwise you are likely to exceed 0dBFS (clip).

2. No, you're not mistaken and those figures look pretty representative. CD has a theoretical maximum dynamic range of 96dB (-96dBFS to 0dBFS), so your classical recording is actually using about 9 of the available 16bits, meaning you could lower the digital level of your 16bit file by about 40dB before you start hitting the digital noise floor.

G

6. "

1. 24bit doesn't have a signal to noise ratio, it just has a lower noise floor than 16bit. This lower noise floor does not directly affect the SNR when recording, as the SNR is defined by the mics, mic-preamps and the SNR of what's being recorded and where.
"

Obviously you don't realize ADCs still fallow the rules of analog up till the threshold of the digital domain.
So the ADC converter chips have a signal to noise product depending on signal level.
Of course you can look at most ADC chips and they have them stated in their data sheets.

3. Why are you coming here and just making stuff up? ProTools was 16bit when it came out, then it's mixing environment changed to 56bit about 20 years ago (Protools TDM), the cheaper LE versions of Protools started at 32bit float and all today's versions are 64bit float. Nuendo started at 32bit float and is now 64bit float. No current DAWs I'm aware of have a 40bit mix environment.

the audio is in 40 bit. the program itself might be 64 bit but the digital audio is 40 bit.
Obviously, you don't realize this.

btw I'm still out there in the industry.

I just don't work at that publisher anymore.

7.
1. Why "obviously"? ADCs must follow the laws of physics and due to thermal noise, this resultant noise floor creates a theoretical limit of about 21bits of dynamic range. In practice most pro ADCs achieve considerably less dynamic range than 21bits. However, this is irrelevant because that's still roughly 100 times more dynamic range than can practically be achieved in real world acoustics (noise floors of recording environments) or in the analogue chain prior to the ADC (mics/preamps). This is where the SNR is defined, not by the ADC, as already mentioned!

2. You're right, I obviously don't realise that, just as I don't realise that the earth is flat or that the moon is made of cheese! Just repeating a figure you have made up does not make it more convincing! The digital audio file/s themselves are usually 24bit or 16bit (or occasionally 32bit float) and is independent of the mix environment within the DAW, the bit depths of which I have already given you for the two DAWs you mentioned. 40bit is a number you've just made up! If you wish to prove you're not just making it up it's simple, just provide the information from Stienberg or Avid/Digidesign.

3. There is no way that even an apprentice audio engineer would not know points #1 and #2 above, even after just a few months in the industry, let alone 25 years! Which means either you're lying/trolling or, as mentioned, you are deliberately trying to mislead and have some other role in the industry which is not that of an audio engineer, of which you obviously have little/no actual knowledge.

Lying about (or at least misrepresenting) your position in the industry and therefore your level of knowledge/understanding is troubling. More troubling still, is that you're just making stuff up and passing it off as fact. This is troubling because this is the science forum, do you honestly believe we're all dumb enough to just swallow whatever BS you fancy making up? Even if you do believe we're all dumb enough, what do you hope to actually achieve? Is this just some rather sad attempt at self-aggrandisement or are you leading up to a dodgy marketing strategy of some product you're trying to sell?

G

8. Maybe this discussed a while ago but I could not find.

We know that our devices which is in audio reproduction chain are not perfect. Any of them. So some sound gets distorted.

This is a straight (maybe a bad) example but I am sure you will get this:

24 bits for not in perfect system may be redundant for us for listening purposes but in real world?

Bits are just example. Consider the amount of information stored in files. According to Nyquist theorem we dont need more than 40khz for listening but does not that applies for a perfect situation with zero entropy?

9. I agree most of Dr. Techno's posts seem either ill informed or out of date.  I also wonder if English is not his native language and we have a language barrier.  If so this is more a case of miscommunication instead of something nefarious.

Digital mixers (as in hardware mixers) would commonly run 40 bit float.  Especially those that used SHARC processors for various DSP effects. Perhaps that is what he has in mind.

I am all for calling out bad info in the Sound Science forum, but maybe we should try a bit less confrontational approach to find out if we are hearing what Dr. Techno is actually saying first.

A few examples of current hardware digital mixers that do the DSP in 40 bit float:

http://ams-neve.com/88d/

http://www.fullcompass.com/prod/254390-Midas-M32

https://www.bhphotovideo.com/c/product/791831-REG/Behringer_X32_X32_32_Channel_16_Bus_Total.html

http://www.stagetec.com/en/audio-mixing-consoles/crescendo/specifications.html

http://www.soundcraft.com/products/vi7000

10. you forgot all those semi-pro interfaces (firewire,usb,thunderbolt) that has mixer built in them too (granted they are mostly used during tracking for direct monitoring).

I can't upload a datasheet. but here is one  of the converter chips that is out there and its "analog characteristics"

pg 7 of AK5394AVS which is here:  https://www.akm.com/akm/en/file/datasheet/AK5394AVS.pdf

11. btw you guys assume too much

12.
We don't need to guess what he has in mind, he was very specific. He not only talked about DAWs (rather than hardware mixers) but two specific DAWs, ProTools and Nuendo:

"back when protools and nuendo came out The mixing enviroment was 24 bit. it didn't take long for then to change the mixing environment to 32 bit because things clip too easily when mixing inside the program. A few years ago the environment went to 40 bit." - None of this is correct!

G

13.

If you got to harp on things I am recalling from memory, then you need to prove I'm wrong, or be quiet. Not complain or make statements you can't back up.

14. well you've got lots to learn then

15.
"The Pro Tools mix engine has traditionally employed 48-bit fixed point arithmetic, but floating point is also used in some cases, such as with Pro Tools HD Native. The new HDX hardware uses 64-bit floating point summing." - Wiki

"Mixing level scaling stores 48-bit results using a 56-bit accumulator for maximum precision." - 1999. Pro Tools 5.0.1 Reference Guide (page 283 explaining the TDM mix architecture, introduced in 1994).

"Steinberg has labeled Nuendo its "Media Production System" and with good reason. Open the program, click "New Project," and there's a list of various application templates ranging from Pro Logic Video Mixdown and 24/96 DVD 5.1 Authoring to 32-bit Stereo Master and Audio/MIDI Music Production." - Mix magazine review of initial Nuendo release.

OK, I've provided some back up for my statements, now it's your turn! Provide corroborating info that Nuendo ever had a 24bit mix environment, that Pro Tools or Nuendo ever had a 40bit mix environment. Of course, you won't be able to do that! Even a relative newbie recording, mix or mastering engineer knows that Nuendo is a host based platform (and can therefore can't be anything other 32 or 64 bit mix environment), let alone someone who's been in the industry for 25 years! It's also inconceivable you wouldn't know the basic architecture of the industry standard DAW software (Pro Tools). Likewise, one might have to explain to a 1st year music technology student that at least 3 of the 24bits cannot be anything other than thermal (Johnson) noise and therefore that a 40bit file format with an additional 16bits of thermal noise is ridiculous but one would be shocked to have to explain that to practising professional, let alone a 25 year veteran!

For this (and several other reasons), your pre-emptive excuse: "I am recalling from memory" is hogwash. There's only one way that someone professing to be a highly experienced professional audio engineer could "recall from memory" something which so obviously flies in the face of basic digital audio theory, and that's if they have no understanding of basic digital audio theory and must therefore be lying about being a highly experienced audio engineer! What I find baffling is that even after your lie was exposed, you continue to defend your "facts" as not just made up and challenge those of us who have in fact been pro audio engineers for 25 years (or more). Regardless of why you decided to post made up "facts" in the science forum and regardless of whether  you're willing to publicly admit it, you (and many/most of us) know that you made them up and you must surely realise that continuing to defend those "facts" will achieve nothing besides digging a deeper hole for yourself!

My apologies to others if my challenging of this member seems overly harsh but there's already way too much made up BS from marketers/retailers in the consumer audio world, without having to waste time dealing with someone just making up BS for self-aggrandisement and/or the fun of misleading others!

G

sonitus mirus likes this.