[JaeYoon]

I love you guys.

Dead.....

Here is an audiophile salesman hat. Anyone wanna get started?

You see purchase this Audiophile Dental Chompers!! Bite into your own stereo system and ensure audio nirvana! But wait!!! Before you touch your keyboard and search google for snake oil!

You see Ludwig Van Beethoven bit into his own piano!! In order to hear how it sounds!
Now you can hear music the way the artist intended it to! Just wait! Call now and receive these cool audiophile certified bumpers. You wouldn't want jitter and EMI and that nasty crosstalk to infiltrate your music now!!

Order now for only...wait for it! Wait for it!
10,000 dollars. Yes! If you order now we can perform surgery and connect these expensive audiophile quality cables right into your chompers as well! Buy now right away!! What are you waiting for!!! Go on!! Throw your money!! Did I forget that these chompers can play back HIGH RESOLUTION 24 bit audio and dsd natively!!! Wow you must be pushing that buy button twice!!

Don't forget to buy an ESD proof audiophile certified case for your chompers too!

 

[Don Hills]

Beethoven bit his piano to hear the music, right? Maybe we could come up with an audiophile dental appliance of some sort!

Done that. Clamped the outlet of a PA compression driver between my teeth. Made my eyesight go blurry and my teeth ache.

 

[ev13wt]

I love you guys.

Dead.....

Here is an audiophile salesman hat. Anyone wanna get started?

You see purchase this Audiophile Dental Chompers!! Bite into your own stereo system and ensure audio nirvana! But wait!!! Before you touch your keyboard and search google for snake oil!

You see Ludwig Van Beethoven bit into his own piano!! In order to hear how it sounds!
Now you can hear music the way the artist intended it to! Just wait! Call now and receive these cool audiophile certified bumpers. You wouldn't want jitter and EMI and that nasty crosstalk to infiltrate your music now!!

Order now for only...wait for it! Wait for it!
10,000 dollars. Yes! If you order now we can perform surgery and connect these expensive audiophile quality cables right into your chompers as well! Buy now right away!! What are you waiting for!!! Go on!! Throw your money!! Did I forget that these chompers can play back HIGH RESOLUTION 24 bit audio and dsd natively!!! Wow you must be pushing that buy button twice!!

Don't forget to buy an ESD proof audiophile certified case for your chompers too!

You and me, we are going to make billions.

 

[RRod]

Beethoven bit his piano to hear the music, right? Maybe we could come up with an audiophile dental appliance of some sort!

Tuning fork to the teeth does wonders for PRaT!

 

[OddE]

The formula for dynamic range (without dither) of N bit data is 20*log (2^N * sqrt(3/2)) = 6.0206*N + 1.761.

16 bit => DR = 98.09 dB
24 bit => DR = 146.26 dB
32 bit => DR = 194.45 dB

There's not even amplifiers with such a low noise floor to be able to achieve this kind of dynamic range.

-I am a bit late to the party, but I'd just like to chime in that the Benchmark AHB2 is getting close enough for all practical purposes (and then some!) - they claim THD+N at -118dB or so at rated output. (Which is where you'd need to be at in order to take advantage of the offered dynamic range.)

Obviously, you'd also be in severe pain and your ears wouldn't be anywhere near capable of relaying said range to your brain - but it would be there. (nevermind the horrific distortion introduced by any transducer driven that hard; again, you'd have other priorities. Like shutting the system down, for instance.)

 

[Cutestudio]

Dither increases usable dynamic range of CD about 10-20 dB (because the signal doesn't granulate (doesn't modulate quantization noise), it remains "pure", just accompanied by the dither noise, with can be made very quiet for human hearing by noise shaping. This means you can reproduce sounds at level 100 dBFS and below!).

The original article is faulty in it's primary premise, something I have pointed out before to the blind.
Dither is a statistical method and while your statement is superficially correct there is an important caveat: We are still representing the data in 16bit digital so the quantisation errors are still present.

Logically then it means that any transient HF shape is still locked to those quantised steps, and therefore still the wrong shape. As a tone it's not an issue (it's a statistical method), but as a one off, non repeated shape, it's still quantised to the wrong shape. The question of whether this creates an audible effect can be addressed by the question: "Can we heard the difference between no dither and (the various types of) dither?" If we can, we must be able to also hear the error. My next point however makes clear that this error is also irrelevant.

The A-weighed background noise of a quiet living room is about 30 dB and the loudest peaks of real symphony orchestra is about 110 dB, so in order to reproduce that dynamic range you need about (110-30) dB = 80 dB of dynamic range. So, (optimally used) 14 bits is pretty much how much is needed. On CD there's 2 more bits to increase the dynamic range by 12 dB so we are good.

A good post but this is where theory and practice diverge. This thread is about the LSB, bit0, and the worries that swapping that bit for 8 bits (i.e. going to 24bit) may be required or not: i.e. it's about the details of the tiniest sounds. This theory is inapplicable to modern pop CDs however because they are all mastered in 15 bits, having thrown away bit 15, the MSB.
There's even a very good chance that those loudest peaks of a symphony orchestra will be simply cut off.

So while the discussion of dither and the misunderstanding of it's limitations is indeed interesting; as a subject I feel it's rather irrelevant to modern digital music. 286 pages on the LSB when the MSB went a long time ago: i.e. we've got a bigger problem that 16 bit dither at -96dB, a problem that's almost exactly +6dB in size.

Interestingly I find most of life is like this, the elephant in the room is usually quite invisible. Today there are many elephants.

 

[gregorio]

[1]The original article is faulty in it's primary premise, something I have pointed out before to the blind.
Dither is a statistical method and while your statement is superficially correct there is an important caveat: We are still representing the data in 16bit digital so the quantisation errors are still present.
[1a] Logically then it means that ...
[2] This theory is inapplicable to modern pop CDs however because they are all mastered in 15 bits, having thrown away bit 15, the MSB.

1. Unfortunately, you appear to have completely misunderstood how dither works and it's purpose. After the correct application of dither there is NO quantisation error, nada, none whatsoever, response is perfectly linear down to the digital noise floor.
1a. "Logically then", the rest of what you say on this matter is nonsense because it's based on your misunderstanding of digital audio/dither. There are no quantisation steps in the waveform after conversion and it is NOT the wrong shape, regardless of whether it's a transient or any other waveform shape!
BTW, it's not a good idea to call others "blind" when you yourself obviously have a poor grasp of the issue!

2. I've got no idea where this assertion comes from, it's complete nonsense but at least it's original complete nonsense, I've never heard this one before. Where did you get the idea that the MSB is "thrown away" during the mastering process?

G

 

[71 dB]

The original article is faulty in it's primary premise, something I have pointed out before to the blind.
Dither is a statistical method and while your statement is superficially correct there is an important caveat: We are still representing the data in 16bit digital so the quantisation errors are still present.

Dithering breaks correlation between the signal and noise. It means errors can't be appointed to the signal, but they form dither noise which is a separate entity of the signal. Quantization errors mean noise that correlates with the signal. When you remove the correlation, quantization noise becomes, well just noise present with the signal.

Logically then it means that any transient HF shape is still locked to those quantised steps, and therefore still the wrong shape. As a tone it's not an issue (it's a statistical method), but as a one off, non repeated shape, it's still quantised to the wrong shape. The question of whether this creates an audible effect can be addressed by the question: "Can we heard the difference between no dither and (the various types of) dither?" If we can, we must be able to also hear the error. My next point however makes clear that this error is also irrelevant.

Well, yes. Signal + noise is not the same as signal alone meaning we have "wrong shape". To correct this you need infinite bits, infinite dynamic range in theory. In real life worrying about wrong shapes at this scale (16 bit) is meaningless.

Look at the magnitude spectrum above: I genereted 1000 Hz sinusoid, level -120 dB (amplitude 0.000001) at 32 bit mode. Then I made it 16 bit using shaped dither noise and the spectrum of that 16 bit signal is above! The 1000 Hz sinusoid is clearly there. If I amplify this signal by 50 dB, I can listen to it and hear (just) the sinusoid behind the dither noise. The sinusoid sounds pure, because it is very very pure. What in this spectrum shows quantization errors? Quantization errors (no dither) look like this:

I Generated 1000 Hz sinusoid, level -80 dB at 32 bit. Then truncated it to 16 bit without dither. This is pure quantization error. You see 3000 Hz, 5000 Hz, 7000 Hz etc. (odd harmonics) dominating as you would expect. Below the same sinusoid at 16 bit with triangle dithering:

The noise power is similar to the non-dither case, but the specrum shape is very different, because the dither noise doesn't correlate with the signal (that would cause harmonics).

A good post but this is where theory and practice diverge. This thread is about the LSB, bit0, and the worries that swapping that bit for 8 bits (i.e. going to 24bit) may be required or not: i.e. it's about the details of the tiniest sounds. This theory is inapplicable to modern pop CDs however because they are all mastered in 15 bits, having thrown away bit 15, the MSB.

Huh? Are you talking about loudness war?

There's even a very good chance that those loudest peaks of a symphony orchestra will be simply cut off.

Not on the CDs of orchestral music I have. Stop buying bootleg crap.

So while the discussion of dither and the misunderstanding of it's limitations is indeed interesting; as a subject I feel it's rather irrelevant to modern digital music. 286 pages on the LSB when the MSB went a long time ago: i.e. we've got a bigger problem that 16 bit dither at -96dB, a problem that's almost exactly +6dB in size.

I wonder who has demonstrated most misunderstanding of dither? MSB = + 6 dB of dynamic range. LSB = + 6 dB of dynamic range. They are equally important.

 

[TheSonicTruth]

Dithering breaks correlation between the signal and noise. It means errors can't be appointed to the signal, but they form dither noise which is a separate entity of the signal. Quantization errors mean noise that correlates with the signal. When you remove the correlation, quantization noise becomes, well just noise present with the signal.



Well, yes. Signal + noise is not the same as signal alone meaning we have "wrong shape". To correct this you need infinite bits, infinite dynamic range in theory. In real life worrying about wrong shapes at this scale (16 bit) is meaningless.



Look at the magnitude spectrum above: I genereted 1000 Hz sinusoid, level -120 dB (amplitude 0.000001) at 32 bit mode. Then I made it 16 bit using shaped dither noise and the spectrum of that 16 bit signal is above! The 1000 Hz sinusoid is clearly there. If I amplify this signal by 50 dB, I can listen to it and hear (just) the sinusoid behind the dither noise. The sinusoid sounds pure, because it is very very pure. What in this spectrum shows quantization errors? Quantization errors (no dither) look like this:



I Generated 1000 Hz sinusoid, level -80 dB at 32 bit. Then truncated it to 16 bit without dither. This is pure quantization error. You see 3000 Hz, 5000 Hz, 7000 Hz etc. (odd harmonics) dominating as you would expect. Below the same sinusoid at 16 bit with triangle dithering:



The noise power is similar to the non-dither case, but the specrum shape is very different, because the dither noise doesn't correlate with the signal (that would cause harmonics).



Huh? Are you talking about loudness war?



Not on the CDs of orchestral music I have. Stop buying bootleg crap.



I wonder who has demonstrated most misunderstanding of dither? MSB = + 6 dB of dynamic range. LSB = + 6 dB of dynamic range. They are equally important.

"
↑
There's even a very good chance that those loudest peaks of a symphony orchestra will be simply cut off.
Not on the CDs of orchestral music I have. Stop buying bootleg crap. "

He may be right. My classical CDs are all original albums released in the mid to late 1980s, yet they all have rock DR values: DR11-13.

I *know* that a typical symphony sitting features much greater actual DR swings than what was put on my CDs. *Something* was done to them!

 

[bigshot]

When people speak of the dynamic range of a symphony orchestra, they're talking about how loud it is from a close perspective. Put your ear right next to a trumpet or tympani and you'll hear some pretty loud sound. But in practice, the louder instruments are arranged further back and the recording is made from a little distance to get a more listenable perspective. The actual dynamics of an orchestral recording is in the same range as any other dynamic music- 35-40dB max. It isn't comfortable to listen to music with wildly exaggerated dynamics.

I have one surround sound orchestral recording that has the stupidest dynamics I've ever heard. It's a violin concerto, and they arrayed the band all around the conductor in a circle and miked the conductor's perspective. Someone had the bright idea of placing all the instruments at the same distance from the mic. They put the tympani right next to the violin soloist, so every once in a while there's a massive kettle drum hit that blasts you out of your chair. Unlistenable.

 

[TheSonicTruth]

When people speak of the dynamic range of a symphony orchestra, they're talking about how loud it is from a close perspective. Put your ear right next to a trumpet or tympani and you'll hear some pretty loud sound. But in practice, the louder instruments are arranged further back and the recording is made from a little distance to get a more listenable perspective. The actual dynamics of an orchestral recording is in the same range as any other dynamic music- 35-40dB max. It isn't comfortable to listen to music with wildly exaggerated dynamics.

I have one surround sound orchestral recording that has the stupidest dynamics I've ever heard. It's a violin concerto, and they arrayed the band all around the conductor in a circle and miked the conductor's perspective. Someone had the bright idea of placing all the instruments at the same distance from the mic. They put the tympani right next to the violin soloist, so every once in a while there's a massive kettle drum hit that blasts you out of your chair. Unlistenable.

Technique counts for something! lol

 

[RRod]

"
He may be right. My classical CDs are all original albums released in the mid to late 1980s, yet they all have rock DR values: DR11-13.

I *know* that a typical symphony sitting features much greater actual DR swings than what was put on my CDs. *Something* was done to them!

And here's a 1979 classical disc (Soundstream recording) with DR15. And here's a 1986 recording with a track with DR20, but of course DR is not really a dynamic range measure since you can add material and have the DR go down…

 

[pinnahertz]

And here's a 1979 classical disc (Soundstream recording) with DR15. And here's a 1986 recording with a track with DR20, but of course DR is not really a dynamic range measure since you can add material and have the DR go down…

The DR meter is a tool for measuring a very specific case of dynamic range, short term dynamic range that is strongly affected by loudness war processing. But it does not measure total dynamic range, that's not it's function..

 

[castleofargh]

The DR meter is a tool for measuring a very specific case of dynamic range, short term dynamic range that is strongly affected by loudness war processing. But it does not measure total dynamic range, that's not it's function..

you're telling it to the wrong guy ^_^.

 

[pinnahertz]

you're telling it to the wrong guy ^_^.

I know, but I'm also aware that he's not the only one here.