Tight 16 vs. Sloppy 24 bit

Discussion in 'Sound Science' started by Avatar86, Sep 6, 2018.
Page 1 of 2
  1. Avatar86
    Although the title might get misread ^_^ I have a notion to share.

    It feels to me as, if a 16 bit song (44.1 kHz) is played in both a player and on a system that is set to 16 bit (44.1 kHz). The sound will be Tight/Attack/Rap/Snappy (in a good sense).

    And if the player encoeds the song in 24 or 32 bit (44.1 kHz), the sound gets Sloppy/saggy/dulled/muffled.
    But the sound also becomes... deeper/more rich/wider/more colorful


    The reason I want it "Tight" is that I listen to a lot of electro. And a main drive in this genre is generally the digital exactness.

    I'm aware that this is a new form of Audiophile. Otherwise the traditional audiophile listens to soft recorded music on Tidal or SACD, where 24 bit realy can give the song more... room.

    Now, I'm a novice at this.
    But I like to share this view with you now. And maybe in the future there will be a higher demand for Hi-Res Dubstep ^_^
    And these guys don't want a soft tune in jRiver or Audivana. They want a kick that blasts the door out!

    A. Try VLC.
    1. Settings: S/PDIF on, no peek limiter, no audio resampling and output modules: from memory.
    2. use Mac (sorry I was pro PC before).
    3. Set MiDi settings to 16/44.1 (or whatever the track is)
    4. Turn off WiFi
    5. Full Volume (full Bitrate).

    B. Use QuickLook in finder (actually a good player!). Scroll by holding the mouse over the track and sliding right on the touch pad.

    Good luck!

    P.s. I might be wrong... who knows. :)
    Last edited: Sep 6, 2018
  2. bigshot
    Why should a noise floor even more below the threshold of audibility sound more "sloppy" than a noise floor below the threshold of audibility? I'd expect them both to sound exactly the same. And the experiments I've made bear that out.
  3. castleofargh Contributor
    @Avatar86 2 problems with this:
    1/ even to this day there is no clear evidence that in typical listening conditions, a listener can tell 16 from 24bit. so your claim would require some proof that you can actually tell when you don't know in advance what's being played.

    2/ a 16bit track you set to output at 24bit will simply have extra zeroes for the extra bits in each sample. which doesn't change the amplitude of the sample. not exactly something that would cause audible change in sound. in fact, if you can send a bit perfect signal(no volume process going to 32 or 64bit then back to what you set for the output), the most likely result should be no difference at all.

    so maybe you imagine the difference, or maybe it reveals another issue we don't know about on your system. but I sooooo very much doubt that on all playback systems, 16bit is "tighter" than 24bit.
  4. theveterans
    A DAC's filtering/oversampling system might make you perceive that difference. If the DAC's filter oversamples to 24 or 32 bits then you're hearing the filters that are built-in to the DAC. However, feeding that same DAC with 24 bits, the filter will get bypassed and you are playing in NOS mode for that DAC which might make you perceive a sound difference.
  5. gregorio
    Your player and your system are set to 16/44.1. If you then set your player to 24/44.1, it will do as Castleofargh stated and add 8 zeros to each sample and then pass those samples to your system. But, as your system is still set to 16/44.1, it will remove those 8 zeros just added by your player and the end result should be absolutely bit identical to what you started with! In other words, the digital output from your Mac should be absolutely identical regardless of whether your player is set to 16 or 24bit. Even if you're changing both your player and your system to 24bit, then the end result (analogue) output by your DAC will still be identical. 16bit or 16bit + 8 zeros will result in an absolutely identical analogue output from your DAC.

    There are only 3 possibilities:
    1. There is a serious programming error in your player. It would have to be a serious error and is very unlikely because padding a 16bit stream to 24bits is trivially easy to program.
    2. When in 24bit mode there's some inadvertent change/difference in the settings. A slightly different volume setting or some additional audio processing, such as audio compression for example.
    3. There is no actual difference in the output of your DAC, the difference you're hearing is caused by your perception (IE. You're effectively imagining a difference).

    1. Firstly, according to the OP, his system is set to output 16/44.1, so regardless of whether his player is outputting 16 or 24bit the DAC will always be receiving 16/44.1. Secondly, even if the OP were also changing his system to output 24bits, still it would make no difference to the DAC's oversampling/filters. I think maybe you're getting confused, oversampling increases the sample rate by some multiple of the input sample rate and then the DAC applies the reconstruction filter for that oversampled rate, which in both cases will be the same oversampled rate and filter because in both cases the OP is outputting 44.1kS/s, bit depth has nothing to do with any of this and is irrelevant. The DAC chips used in today's consumer audio devices are pretty much exclusively 24bit (or 32bit float in a few cases). So a 24bit input is processed (oversampled and then filtered) as is, and a 16bit input is padded with 8 zeros and then also processed at 24bits.

    2. If this statement were true, then the DAC would always be NOS and filterless, regardless of whether you're inputting 16 or 24bit.

    Last edited: Sep 7, 2018
  6. theveterans
    1. The DAC will receive whatever the system output is. If it’s 16 bits 44.1 then it doesn’t matter if you play DXD, it would still be downsampled to 16 bits 44.1 before it goes to the DAC

    2. On Schiit R2R DACs except Yggdrasil this is true. If you feed a 24-bit / 192 KHz or 176KHz output to the DAC, you’re bypassing its filters already. So if you set the output to it then all of 16-bits will be upsampled to 24/192 by adding zeros before it goes to the DAC. Other DACs such as Chord DACs its filters oversamples much higher than 192 KHz.
  7. 71 dB
    That one you got right. Based on your message you have a long way to understand digital audio. Trust me, many assumptions you have about digital audio are wrong. Maybe the comments you have been given here have already exposed some of them?

    How about Low-Res Dubstep? What if 8 bit Dubstep is even tighter? That's the first question you should have asked yourself when coming up with your theory. In other worlds, what is the optimum bit depth in regards of tightness? The second question to ask is if this "tightness/sloppyness" is a real property of the sound or something your head makes up, because in audio our heads make up a lot of things (placebo effect) and you can't trust your ears 100 %.
  8. castleofargh Contributor
    you're mixing everything up here. going from from 16/44 to 24/44 is not upsampling as the sample rate is the same 44.1khz! resampling, upsampling are terms reserved for the act of changing the number of samples. but when you change the bit depth, you don't resample, you add "slots" for the extra bit values into each already existing sample. the number of samples doesn't change.
    OP talks about the difference between playing 16/44 or 24/44 and clearly mentions not touching the sample rate before it goes to the DAC.(then the DAC does its thing, but it shouldn't be relevant irrelevant if it's twice the same operation on both 44.1khz signals right?).

    maybe your confusion comes from how some oversampling methods involve padding with zeroes too? but it's a different padding done somewhere else with a totally different result. it's based on how if you add zeroes in one domain, your end up oversampling in another. so that is often used in a time vs frequency relation.
    but once again, turning a 16bit signal into 24bit is something else entirely. you take one sample, it has a number of 0 and 1 for general code and information, and also 16 1 or 0 expressing an amplitude value. going from 16 to 24bit is adding extra zeroes in each sample so that the final number is a 24bit value instead of 16bit. but the 16bit signal didn't have any value allocated for the extra 8bit of a 24bit signal(duh), so you just add 8 zeroes. which is telling the DAC exactly the same, in a R2R design it would mean: "keep the last 8bit OFF". in a modern design it's giving the same 16bit amplitude.
    you can go to 24bit and back to 16bit a billion times in a row without altering any data so long as you don't dither when going to back to 16.

    PS: for other people who might conclude from what I just said, that playing a 16bit file and setting the output to 24bit on the computer is useless, here is why it's often not useless: volume control on the computer.
    I simplify the following example by removing 2steps(volume control being handled at 32 or 64bit, and dither) but the general notion still mostly stands. let's say I lower the volume on my computer(because it's easy to set the listening level that way). let's say I lower the volume to -18dB so 18/6=3 that's about a 3bit attenuation so let's pretend it's exactly 3bit. the loudest encoded signal is now 3bit lower. and so are all the sample values. if I do that with my 16bit file as it is, the lower 3bits are lost for all samples. it's not a big deal because they're defining the quietest stuff on the file and it's likely to just be noises instead of actual music content. but still I lost 3bit. I'm now listening to 13bit music in my oversimplified situation.

    now the same thing with my 16bit file and the signal output set to 24bit on the computer. my samples have all the unused slots for the extra 8bits. they're zeroes, but if I attenuate my volume by 3bit, instead of getting rid of the last 3bit of each sample, I can now move their value into the extra 8bits. so my signal is still 16bit, but 18dB quieter. I usually can't hear the difference with music and my usual listening levels(because sadly real world situations rarely allow for even 16bit=about 96dB of fidelity). but it's reassuring to know that we have some extra bits at no cost. I just set the computer to the highest bit depth my DAC can handle, and forget about it.

    PS2: if I said something stupid, while trying to be too simple to be real, please let me know.
  9. Arpiben

    You may have confused bit depth (16 bits / 24 bits / 32 bits) with sample rate or sampling frequency ( 44.1kSps/kHz,48 kSps/kHz,88.2 kSps/kHz, 96 kSps/kHz,etc...)
    kSps means kilo Samples per second. It is the unity for sampling rate.
    Sampling rate may be expressed also as a sampling frequency rate in Hertz (Hz)

    As posted by others, changing the bit depth only (bit zero padding), should have absolutely no impact at your DAC analogue output.

    Depending on architecture, DACs may have different filters, digital and analog ones.
    Digital filters in DACs are mainly use for interpolation (calculating the missing points/data when upsampling) purposes.
    In principle, there is an analog LPF (Low Pass Filter) at the end of the DAC architecture.

    Let's imagine that you change your frequency sampling rate, at digital player level, from 44.1kHz to 88.2kHz ( upsample by 2).
    For this action, your digital player needs to:
    • insert a zero value sample in between each original sample data ( zero sample padding)
    • interpolate a value for the previous zero samples added
    • correct the amplitude/level according to interpolation method used (linear, LPF,etc...)
    At this stage, it is already enough for having differences at DAC's output.
    The upsampled digital data will now enter DAC at 88.2 kSps...

    Assuming that your DAC is upsampling up to a maximum of 2048 times ( Chord for example).
    Since you already upsampled by 2 at digital player, your above DAC will 'only' upsample the input by 1024 (instead of 2048) using different stages and interpolating methods.
  10. Arpiben
    We wrote at same time and your post arrived before.
  11. Dodeca
    I found this article tremendously helpful in understanding the science, history, and development of sample rates: https://sonicscoop.com/2016/02/19/t...rates-when-higher-is-better-and-when-it-isnt/

    Also, coming from a pro audio vs. audiophile background gives me a solid respect for the fact that 16/44.1 is, when blind A/B tested, is indistinguishable from higher sample rates. Same with most 320k MP3s. 24bit can be useful when mixing/producing, although lots of folks (myself included) just work in 16/44.1 (or 48). Saves processing power and disk space, good plugins do their own anti-aliasing/upsampling.

    My understanding is that if the same DAC sounds different at 16/44.1 and 24/96 (or 192) it's because of shortcuts taken in the engineering of the DAC, not a result of sample rates.
    Speedskater and sonitus mirus like this.
  12. gregorio
    1. Good article, thanks, I don't recall having seen that one before.

    2. 24bit makes no difference when mixing/producing. Everything is loaded into a 64bit mix environment and processed at 64bit in pro audio DAWs anyway. 24bit is only useful for recording, due to it's increased headroom and it's typically left at 24bit until it's loaded into the DAWs mix environment. It's therefore not actually possible to "work in 16/44.1 (or 48)" in a pro audio DAW! Due to recording at 24bit, the audio file format of these raw tracks is typically 24/44.1 or 24/48, although again, they effectively become 64bit inside the DAWs mixer, so you've effectively got 64/44.1 or 64/48 and then at final mix bounce down that becomes whatever you choose (16, 24 or 32float).

    3. Essentially yes. Technically, 192kS/s should sound worse, due to the potential issue of IMD and because it's too fast a rate to effectively apply the reconstruction filter. Filter rejection is only about -80dB using 192kS/s instead of -120dB with lower rates. I say "should sound worse" but in practise it's doubtful alias-images at -80dB are audible. If 192kS/s actually sounds "better" to someone, it's because either they are getting IMD and like it or because the DAC has particularly shoddy filters at lower rates.

  13. Arpiben
    Music technology/science is indeed truly multidisciplinary (some would say transdisciplinary). It incorporates aspects of the traditional disciplines of music, engineering, physics, psychology and computer science and, of course these disciplines themselves incorporate others such as mathematics, biology and chemistry.

    Dealing with history, your nice paper could have also mentioned: Fourrier, Shannon, Gabor and many others.
    Hereafter the links to Dan Lavry white papers:

    Despite manufacturers trend and propaganda, upsampling or upscaling up to crazy rates does not change at all resolution or timing accuracy of the original file.
    Your paper summarizes it quite well while keeping it simple.
    muza_1, gregorio and sonitus mirus like this.
  14. sonitus mirus
    I enjoyed the read. The article is nearly verbatim to what @gregorio has continued to profess in these forums since as far back as I can recall.

    Here is Justin's website.

  15. Avatar86
    Wow! Nice to see all the enthusiasm around this subject :)

    Still, it's nothing new...

    I've been reading this rap for three years. I'm still not convinced.

    It's the same with people who say that editing existing songs in a DAW, SHOULDN'T give any quality drop when bouncing.

    Still there is a loss in tightness, depth and detail between the original file and the edited file. (I have tried Pro tools, Logic, Audition, Audacity, FL. With different Mac's and PCs. With and without ext. Sound Cards. Even in a Pro studio!!). I've been experimenting with this for 6 years!!!

    Kind of tired of the "you can't hear it"-sh*t.

    No one gives a damn to even try for themselves. ("No your wrong, I haven't tried, but it shouldn't, so I don't know... but you are wrong!"). :xf_mad:

    No one cares for Good SQ in Electro, EDM, Glitch Hop music... where I see a great win in it.

    I'm before my time! :)

    Hope you don't get to offended of this.
    It's just my opinion :) and my strong belief, that 16/44 sound gets tighter on straight 16/44 playback.

    Let time tell right :wink: who knows, I might be wrong... so might you.

    Thanks for your time and detailed explanations.

    Good luck.
    Last edited: Sep 11, 2018
Page 1 of 2

Share This Page