[WoodyLuvr]

Never run from a debate! Regardless!

Anyway....... I've never really paid attention to bit rates and whatnot. FWIW, I've downloaded stuff from HDtracks("lossless") , in the past, and have been a Spotify premium account holder for 3 years now. Listening to the same song , with my back turned to my laptop , and my wife randomly switching between foobar, and Spotify. I could not accurately tell which one was lossless , or Spotify. Almost every time I guessed , my wife said nope, you're wrong.
And yes , I was literally guessing because I could hear no discernible difference whatsoever.
Thank God, because I haven't downloaded anything from there since, which in turn, saved me some cash.

Who knows. Maybe I'm not a "Critical listener", or have "Golden ears".
I'm content with that though.

The silliness reached a level that simply was just giving me a big headache and nothing else. I hear you on saving money. I too went through a FLAC/ALAC phase a few years back... what a waste of time, money, and memory! Cheers.

 

[Guidostrunk]

So, how exactly is the right way? Lol. And to answer your question. Yes it was multiple songs.

Tell me what the right way is, and I'll give it a whirl.

Did you by any chance use tracks consisting of more than one sample for your comparisons? That would perfectly explain why you were not able to discern HD from the lossy format. Sorry to tell you you did it all wrong.

 

[Whazzzup]

hey lay off my lossless

 

[Guidostrunk]

Lol. I laid off my lossless. Yours is fine.

 

[headdict]

So, how exactly is the right way? Lol. And to answer your question. Yes it was multiple songs.

Tell me what the right way is, and I'll give it a whirl.

According to what we just learned on this thread you have to listen to single samples in order to hear the dither's ineffectiveness with lossy formats (or any 16 bit format actually). With music that lasts longer than a single sample, you won't hear it.
Lossless is for losers, by the way.

 

[castleofargh]

Oh dear, you really missed the point there, again.
You merely highlighted what dither does for long and periodic signals, which we all know already. Duh.

What you keep failing to acknowledge, either deliberately or via some form of denial or mental block is that dither doesn't fix short transient events, it merely reduces correlated quantisation artifacts over time in a statistical manner. You have a basic but only partially formed knowledge of dither: a little knowledge can be a dangerous thing'!

You are assuming it magically converts a low bit depth waveform into a higher bit depth waveform in all cases, but it's a statistical method and can therefore only do that over time. What do you think dither actually is? Think about the mechanism!! It's adding a sub-bit noise to a waveform before quantisation to a lower bit depth to statistically knock some values up to the next integer level, instead of the zero dither option that truncates them all flat.

This statistical method can only work over time and that is why it can't fix the shape of individual wavelets, it merely de-correlates the quantization artifacts. Of course the artifacts are non-linear, what did you think was linear about truncating certain values??

I suggest you learn a little about digital signal processing, the dither technique is not magic, it's a trade off, like a PWM signal it's making sure enough hops between the quantised levels occur over time so the errors appear as random noise rather than correlated noise: it's a trade, not a cure, and in no way makes the waveform more 'accurate', it just hides the error in a pleasant noise.

The very facts that:

1) People can hear the effect of different dithers
2) Many people post process digital sound before it's played

tells us that 16bit is inadequate for many, even it you think it's perfect for your car. Even the dither is wrong for post processing, mathematically TPDF dither is best for accuracy of long and continuos waveforms but many CDs and downloads are mastered with a final shaped dither (because they sound better), which is not ideal for any digital post processing like room EQ etc.

I'm still not sure why you keep pushing 16bit, 24bit is clearly a better format - even if you can't tell the difference - and is widely used outside of audio (DVD videos for instance often use 24 bits), your insistence that we all remain listening to a format that was barely adequate 30 years ago shows that you are living in the past and need to embrace the technology of today, not the 'good enough for me' 16bit format.

We don't all listen to church organ recitals.

nobody is forcing you to use 16bit. the very guys telling you how 16bit is audibly enough for playback are or have been using way more bits on a daily basis for professional reasons. hell, I output my 16bit files to 24bit because my DAC is slightly better that way and I feel less guilty for using the digital volume in foobar. you need 24bit, you want 24bit, use 24bit. who cares? we're not from the bit police.
you think albums need 24bit encoding? we have too many failed blind tests on different scales that suggest the opposite for playback. unlike the links you provide that consistently focus on sample rate and ultrasounds yet somehow seem relevant to you. why? or you could simply try like I suggested you do some times back, to use dither of different types at different bit depth and abx the files or at least try to notice the noise floor. you notice it, you go down a little. you notice it less, you keep going down. and then you don't notice it anymore, you look at the bit depth and it's something like 12 or 13bit maybe 14bit with quiet tracks and the volume level real loud. if you reach 18 or 20bits as your threshold, please let us know how much you had to cheat to get that result.

but whatever your problem is, it's not like anybody claimed that 16bit has more resolution than 24. so what's the big deal?




reading this threat the only thing I notice is that you fail to integrate any new information provided to you.
-the transient thing, nobody has any idea what the hell you're talking about. just in case, the time axis is the other one on your graph.
-the DVD movies... someone pointed it out last time already, you don't care, you're happy to repeat your mistakes using the one example that doesn't work.
-your understanding of dither... I don't know where to start so I wont. I'm scared of the moment when you will realize that band limiting also changes the shape of the signal in audacity. will you then say that Nyquist is wrong and buy a NOS DAC with no filter?
all in all it's a little discouraging.




as to why we're ok with 16 when we could have 24bit? I can only answer for myself. I'm happy that way for the same reason I have "only" a 300dpi printer and "only" a 22 megapixel camera. because those are numbers which already reach the threshold of my senses for the typical use I make of those tools. maybe I lack that little alpha male thing where everything must turn into a competition? maybe I just know that better does exist but what really make something better aren't those specific variables? probably both.

 

[Guidostrunk]

First. Who ever said anything about samples of songs? Certainly not me. My comparisons were of full songs, and multiple songs.
Secondly. I wouldn't call anyone a loser, based on their preference in audio format.

Definitely someone who is down right ignorant , deserves loser status

According to what we just learned on this thread you have to listen to single samples in order to hear the dither's ineffectiveness with lossy formats (or any 16 bit format actually). With music that lasts longer than a single sample, you won't hear it.
Lossless is for losers, by the way.

 

[headdict]

First. Who ever said anything about samples of songs? Certainly not me. My comparisons were of full songs, and multiple songs.
Secondly. I wouldn't call anyone a loser, based on their preference in audio format.

Definitely someone who is down right ignorant , deserves loser status

LOL. You were obviously not following the recent discussion on how dither works and I was just trying to apply what I have learned from it. Everything I wrote is of course utter nonsense.

 

[Guidostrunk]

Fair enough. I know nothing about dither, so I'm out of that discussion lol.

LOL. You were obviously not following the recent discussion on how dither works and I was just trying to apply what I have learned from it. Everything I wrote is of course utter nonsense.

 

[pinnahertz]

Your generalisation from a single example is noted.
Yes, the capability is there would have been sufficient.

It was YOUR generalization, and was incorrect. The capability is there, but is unusable for video DVDs!

Outside the HiFi box 24bit and high sample rates are common place, as they are in any pro-audio gear (obviously), which is often IME cheaper and better than the niche 'audio' gear.

You have a odd (means: wrong) definition of commonplace. Higher resolution in production is not what this discussion is about. It's about the release format into the hands of consumers.

I bought a cheap USB to S/PDIF converter the other day with coax and optical out for a project, you know what, even that goes to 24bit / 96kHz.
Yet here we are, double teaming the heretic who suggests 24bit has benefit over 16bit. Doh!

It's because the heretic doesn't seem to understand application of technology. The fact that 24 bit consumer gear is available doesn't change the fact that it's benefits, if any, are inaudible to the user, are a strong marketing influence only, and simply occupy extra bandwidth and storage space. And it's mostly not available to the consumer anyway. Yes, we work in 24 bits in production. Often at slightly higher bit rates too, but that doesn't mean it makes any sense to release that to the consumer, that he would want it, or could actually even play it. You do know what devices most music is played on, right? Or are we in our own world there too?

 

[bigshot]

The irony of this whole debate is that 16 bit audio even without dithering is for all intents and purposes good enough for listening to music in your living room. Dithering just gilds the lilly and makes the noise floor a little bit lower. I fail to see what kind of transients you're going to be eliminating at -70dB. It doesn't affect a 20kHz waveform either. This argument is pure theory stretched to absurd levels. No practical effect at all.

 

[pinnahertz]

The irony of this whole debate is that 16 bit audio even without dithering is for all intents and purposes good enough for listening to music in your living room. Dithering just gilds the lilly and makes the noise floor a little bit lower. I fail to see what kind of transients you're going to be eliminating at -70dB. It doesn't affect a 20kHz waveform either. This argument is pure theory stretched to absurd levels. No practical effect at all.

Yes!

Small point perhaps but dithering raises the noise floor, not lowers. But doesn't affect transients or 20kHz, and this has long since passed beyond theory, and even more importantly, passed beyond the realm of practical application of theory. In most living rooms you can't tell if ti's 16, 16 dithered (which it likely is), or 24 (which it can't be other than the structure of the data itself).

I think in honor of the death of logic and reason in this thread, you should play a good, loud Caruso acoustic recording for us on the mechanical/acoustic player (with it's own special kind of dither). Point the horn east, perhaps I'll hear it.

 

[old tech]

The irony of this whole debate is that 16 bit audio even without dithering is for all intents and purposes good enough for listening to music in your living room. Dithering just gilds the lilly and makes the noise floor a little bit lower. I fail to see what kind of transients you're going to be eliminating at -70dB. It doesn't affect a 20kHz waveform either. This argument is pure theory stretched to absurd levels. No practical effect at all.

I'm far from an expert in this area and this statement makes me question what I've learned to date. I thought that dither raises the noise floor - ie correlating quantisation errors into random white noise?

I agree though that the noise floor of dithered 16bits is for all intents and purposes inaudible. The only question I have in my mind though is why dither at all at 16bits when the dithered noise floor is inaudible, which by definition would mean that the audibility of quantised errors from undithered 16bits quantised would be even lower?

 

[Don Hills]

... You see this is where I'm disagreeing with you, dither can only work over a number of samples because it's a statistical method.

I suspect the problem is that you haven't reviewed many digital waveforms close in and seen how much each point counts. Take the old Genesis 'Lamb Lies Down' album, very dynamic, plenty of very quiet parts where dither is really important for 16bit due to the inherent lack of resolution. The 'perfectly accurate' you speak of is a statistical measure than cannot apply to the exact shape of one-off transient events, only in general will the waveform be considered accurate. E.g. in a periodic waveform you can look at a dithered digital cycles, superimpose all the cycles and averaged together they are indeed accurate, but individually they still have to obey the quantising levels. If therefore you only have one of them you have no idea of knowing the true shape. ...

How many samples does it take to capture the shortest transient that will fit within the Nyquist bandwidth?

 

[Strangelove424]

I'm far from an expert in this area and this statement makes me question what I've learned to date. I thought that dither raises the noise floor - ie correlating quantisation errors into random white noise?

I agree though that the noise floor of dithered 16bits is for all intents and purposes inaudible. The only question I have in my mind though is why dither at all at 16bits when the dithered noise floor is inaudible, which by definition would mean that the audibility of quantised errors from undithered 16bits quantised would be even lower?

It just comes down to what sounds more pleasing if you were theoretically going to hear it. Undithered errors would take the form of distortions that correspond to the signal, which is generally thought to be displeasing when compared to the homogeneity of white noise. We tend to associate white noise with analogue and randomized distortion with digital artifacts, and dithering is often thought of as a way of making the noise floor of digital sound more like traditional analogue. It's not audible, but in theory that's why distortion is traded for noise.