[gregorio]

[1] I don't believe I wrote anything that was incorrect if dither were not in use.
[2] Also despite you acting as if no dither isn't an option, plenty of software will allow you to do things without dithering. It isn't a wise choice to make, but it happens.

1. It's maybe just a misunderstanding, due to the fact that I didn't find it clear that your example and statements excluded dither and a few statements appeared not quite right, for example:
"If you wish to discuss accuracy of the frequency then 44/16 can do so down to 55 picoseconds which is in a around about way the accuracy limits of depicting frequency. However, dither will decrease that number further. 24 bit could depict frequency to a finer level of accuracy." -
Dither will indeed increase that number further, to infinity. Undithered 24bit will give greater accuracy/resolution than undithered 16bit but dithered 16bit has more resolution than undithered 24bit. Dithered 16bit and dithered 24bit have exactly the same (infinite) resolution! lol ... And:
"With enough bit depth you can describe to any level of accuracy desired." - In an undithered system infinite resolution would require infinite bit depth, which is why it's always applied in the quantisation process.

2. It's not an option in the quantisation process. Any recorded/acquired signal will be dithered as all ADCs automatically dither and signals generated in the computer will either not incur quantisation or the processor generating the signal would be expected to internally dither if it does. It's only in the mixing/summing of those dithered signals where a "re-quantisation" process is likely to occur and where dithering that/those re-quantisations is optional. Therefore, I can't think of any commercial digital audio recording scenario in which dither is completely absent, although this doesn't exclude the possibility that at some points in the chain dither hasn't been applied to re-quantisation processes (typically because it's irrelevant) or hasn't been applied correctly.

Even If you have complete perfection for a very short band limited area, you will still need infinite amount of data thus storage. There is no other way.

This is another way, it's called dither!!! That's why dither is a fundamental requirement of digital audio, because without it digital audio is not linear and to make it linear you would need infinite bit depth. That's what the original post explains and why the whole post is effectively based on dither. So, saying that if you eliminate dither the original post is incorrect is completely nuts! Of course it would be incorrect if you exclude dither, which is precisely why the OP does NOT exclude dither!!! Is it just me or are we well past the point of surreal?

G

 

[RRod]

I'm a bit confused by all this. Was there some point in time where people were noticing all kinds of quantization artifacts from their favorite un-dithered CD releases?

 

[StanD]

I'm a bit confused by all this. Was there some point in time where people were noticing all kinds of quantization artifacts from their favorite un-dithered CD releases?

If dithering is done in the ADC (Analog to Digital Converter) then the content of the CD is already dithered.

 

[HAWX]

Dither will indeed increase that number further, to infinity.

I don't agree. But I don't have in depth knowledge about dither to oppose or show a direct proof.

In an undithered system infinite resolution would require infinite bit depth

I have been saying basically the same thing, I agree.

1.
Therefore, I can't think of any commercial digital audio recording scenario in which dither is completely absent, although this doesn't exclude the possibility that at some points in the chain dither hasn't been applied to re-quantisation processes (typically because it's irrelevant) or hasn't been applied correctly.

I can think of non recorded, completely computer generated signals mixed together for elctronic music in computer. However, in real life recordings dither should be applied in the end, which everybody knows.

 

[bigshot]

[Please try to understand my writing fully.

I'd like to, but I don't see anything to indicate that what you are talking about would be remotely audible for human ears listening to a CD of music in their living room. Can you elaborate a little more on the audibility of what you are talking about? Theories are great, but I like listening to music better.

 

[HAWX]

I'd like to, but I don't see anything to indicate that what you are talking about would be remotely audible for human ears listening to a CD of music in their living room. Can you elaborate a little more on the audibility of what you are talking about? Theories are great, but I like listening to music better.

From my original post: "...... (not asking wheather we discern it or need it)?"
Sorry If you are seeking for discenrable diffrence by our hearing. Answer is obviously, no.

 

[RRod]

If dithering is done in the ADC (Analog to Digital Converter) then the content of the CD is already dithered.

Well now we're just into the semantics of whether or not you are deliberately adding the randomness or letting natural phenomena do it. Either way, there was a time when people weren't working in high-bit processing chains that automatically dither, and yet I've never seen a single review from the 80s that said "man, these little crackly sounds at the end of reverb tails really annoy me!"

 

[castleofargh]

@HAWX
we never said we could perfectly encode all sounds without any noise, in fact all of us have been insisting on noise a all lot and you do not seem to care or to get the implications.

also you can encode something in 1 bit and retrieve more than 6db of dynamic from it. delta sigma DACs and DSD work because that is a reality. when you encode in 1bit you create a crippling noise that is added to the signal 6db below. if you stop there then sure it looks bad. but given the right tool and sample rate, you can apply noise shaping to move that crippling noise into another frequency range. and what's left below the first 6db is not nothing. what's left is the original signal in all its glory, down to the point where some other noise is again added to it for whatever reason like the limit of the noise shaping ability or thermal noise or whatever. that's why I've been insisting on looking at things as proper signal+some noise. because the right signal is always there. just not always alone. ^_^

 

[spruce music]

1. It's maybe just a misunderstanding, due to the fact that I didn't find it clear that your example and statements excluded dither and a few statements appeared not quite right, for example:
"If you wish to discuss accuracy of the frequency then 44/16 can do so down to 55 picoseconds which is in a around about way the accuracy limits of depicting frequency. However, dither will decrease that number further. 24 bit could depict frequency to a finer level of accuracy." -
Dither will indeed increase that number further, to infinity. Undithered 24bit will give greater accuracy/resolution than undithered 16bit but dithered 16bit has more resolution than undithered 24bit. Dithered 16bit and dithered 24bit have exactly the same (infinite) resolution! lol ... And:
"With enough bit depth you can describe to any level of accuracy desired." - In an undithered system infinite resolution would require infinite bit depth, which is why it's always applied in the quantisation process.

2. It's not an option in the quantisation process. Any recorded/acquired signal will be dithered as all ADCs automatically dither and signals generated in the computer will either not incur quantisation or the processor generating the signal would be expected to internally dither if it does. It's only in the mixing/summing of those dithered signals where a "re-quantisation" process is likely to occur and where dithering that/those re-quantisations is optional. Therefore, I can't think of any commercial digital audio recording scenario in which dither is completely absent, although this doesn't exclude the possibility that at some points in the chain dither hasn't been applied to re-quantisation processes (typically because it's irrelevant) or hasn't been applied correctly.


This is another way, it's called dither!!! That's why dither is a fundamental requirement of digital audio, because without it digital audio is not linear and to make it linear you would need infinite bit depth. That's what the original post explains and why the whole post is effectively based on dither. So, saying that if you eliminate dither the original post is incorrect is completely nuts! Of course it would be incorrect if you exclude dither, which is precisely why the OP does NOT exclude dither!!! Is it just me or are we well past the point of surreal?

G

Is it just me or are we well past the point of surreal?


Yes you are.

One issue repeatedly when explaining dither to someone who doesn't know is the confusion that infinite resolution would also mean infinitely low noise. I understand what you mean, but what is the reply by Hawx about this? It is exactly what I see over and over, they don't understand what you mean by infinite resolution. So you stated it and the other fellow didn't understand it. So you have conveyed pretty much no useful information to the person asking a question.

Sometimes being too ticky picky to the Tee correct is an impediment to getting the point across. I get that this idea should have long ago died about resolution of digital audio, but I also see that simply stating it is so doesn't get the job done helping people understand it. Yes I was sloppy in my explanation, but it was in a way that to someone new to it they wouldn't even see the difference. Yes it would have been better to be more careful.

To reiterate, get someone to understand how digital sampling works without dither because it is simpler. They can see the relation between sampling resolution and timing and noise. Then next show how the dither works and they usually have the light bulb go off and see how you have finessed the resolution issue of limited bit depth and then they understand why it works that way. Then with this idea in their mind in a way that is rationally understandable they are no longer snookered by some of the snake oil issues related to this situation. Simply saying this is how it works trust me and don't question it doesn't work well with lots of people.

You have the same problem with the ADC statements you made to someone who doesn't know how it works. What you wrote will explain exactly nothing to them. While it is all true and correct you haven't provided anything to someone who doesn't already know.

 

[StanD]

Well now we're just into the semantics of whether or not you are deliberately adding the randomness or letting natural phenomena do it. Either way, there was a time when people weren't working in high-bit processing chains that automatically dither, and yet I've never seen a single review from the 80s that said "man, these little crackly sounds at the end of reverb tails really annoy me!"

Some folks like to split hairs or have overactive imaginations regarding audio tech, I think they are called ........... You can fill in the blanks.

 

[gregorio]

[1] I don't agree. But I don't have in depth knowledge about dither to oppose or show a direct proof.
[2] I can think of non recorded, completely computer generated signals mixed together for elctronic music in computer.
[3] However, in real life recordings dither should be applied in the end, which everybody knows.

1. By saying you don't agree you ARE "opposing"! Did you not understand from the OP what dither does?
2. No, it would have to be even more specific than that, for 3 reasons: Firstly, the vast majority of electronic music is not purely electronically generated, it's at least partly based on samples (which have been recorded) and therefore had dither applied. Secondly, even if it is purely synthesized, it would need to be synthesised at a different bit depth, in order to incur a quantisation process and therefore any quantisation error and Lastly, if the soft-synth does incur a quantisation process it would have to do so without applying dither internally. While one could in theory create a piece of music which fulfilled all these conditions, I'm not sure in practise if there have ever been any commercial releases which do.

Well now we're just into the semantics of whether or not you are deliberately adding the randomness or letting natural phenomena do it. Either way, there was a time when people weren't working in high-bit processing chains that automatically dither, and yet I've never seen a single review from the 80s that said "man, these little crackly sounds at the end of reverb tails really annoy me!"

It appears that some here still don't fully appreciate how integral dither is to digital audio, how integral it's always been and that it's ALWAYS been automatically applied. Some of the very earliest ADC stages generated dither with an analogue dither unit/stage (although they couldn't produce very accurate dither) but since the early '80's accurate digital TDPF dither was automatically applied by all ADC stages. So, the reason you've never seen a single review from the 80's about reverb tails cutting in and out is because there have never been any digital recordings without dither!

I think the reason for some of the confusion is that the situation changed at the start of the 90's, with the introduction of bit depths greater than 16bit, initially 20bit. Dither was still integral and automatically applied during the quantisation process (as it had always been) but working at 20bit introduced the new concept of a "re-quantisation" process, an additional quantisation process required to get from 20bit to CD (16bit). The application of dither for this re-quantisation process was a choice, initially: Truncation (no dither), Triangular PDF or Rectangular PDF. Then, during the early/mid 90's we saw the introduction of a new re-quantisation dithering choice, noise-shaped dither (Sony's Super Bit Mapping and Apogee UV22, for example). I suspect that this choice of re-quantisation dither is what is leading some to believe that dither is some sort of optional bolt-on extra just added at the end of the mastering process, rather than realising that it's integral to sampling theory (digital audio) and that all quantisers are dithering quantisers! (Though not necessarily all re-quantisers).

To reiterate, get someone to understand how digital sampling works without dither because it is simpler. They can see the relation between sampling resolution and timing and noise. Then next show how the dither works and they usually have the light bulb go off and see how you have finessed the resolution issue of limited bit depth and then they understand why it works that way.

Your response just seems to be extending the surreality even further! I did explain how digital audio works without dither, I did go on to explain how dither works and I did explain that we end up with a perfect measurement (infinite resolution) plus some noise, that the quantisation errors have been converted to noise, and I did go into considerable detail about the effects of that noise. My original post is pretty much ENTIRELY based on exactly what you're now suggesting 8 years later!?

G

 

[HAWX]

also you can encode something in 1 bit and retrieve more than 6db of dynamic from it. delta sigma DACs and DSD work because that is a reality. when you encode in 1bit you create a crippling noise that is added to the signal 6db below. if you stop there then sure it looks bad. but given the right tool and sample rate, you can apply noise shaping to move that crippling noise into another frequency range. and what's left below the first 6db is not nothing. what's left is the original signal in all its glory, down to the point where some other noise is again added to it for whatever reason like the limit of the noise shaping ability or thermal noise or whatever. that's why I've been insisting on looking at things as proper signal+some noise. because the right signal is always there. just not always alone. ^_^

Yeah, you got my example. But I think I might have mistake. Practical limit can be 1/4 bit although It doesn't make sense to me, but still gives an indication about what I mean. I need to dig that further. But I was not talking about the dithered and noise shaped case. By the way does DSD files work exactly on the same logic as PCM? Like say in 11.2MHz can you encode up to 5.6 MHz?


Does anybody has access to "
[SIZE=5][B]Resolution Below the Least Significant Bit in Digital Systems with Dither" [/B]from audio engineering society or somewhere else? Thanks.[/SIZE]

[QUOTE="gregorio, post: 13473223, member: 69811"]1. By saying you don't agree you ARE "opposing"! Did you not understand from the OP what dither does?
2. No, it would have to be even more specific than that, for 3 reasons: Firstly, the vast majority of electronic music is not purely electronically generated, it's at least partly based on samples (which have been recorded) and therefore had dither applied. Secondly, even if it is purely synthesized, it would need to be synthesised at a different bit depth, in order to incur a quantisation process and therefore any quantisation error and Lastly, if the soft-synth does incur a quantisation process it would have to do so without applying dither internally. While one could in theory create a piece of music which fulfilled all these conditions, I'm not sure in practise if there have ever been any commercial releases which do.
[/QUOTE]

1) My English is not top notch. I'm opposing but I don't have any real proof to show. But I will dig futher when I have time.
2) I already stated "I can think of." and gave a broad example. I didn't said there is a recording which have never seen dithering process either, which you examined the conditions.

 

[sonitus mirus]

Does anybody has access to "
[SIZE=5][B]Resolution Below the Least Significant Bit in Digital Systems with Dither" [/B]from audio engineering society or somewhere else? Thanks.[/SIZE]



1) My English is not top notch. I'm opposing but I don't have any real proof to show. But I will dig futher when I have time.
2) I already stated "I can think of." and gave a broad example. I didn't said there is a recording which have never seen dithering process either, which you examined the conditions.

http://www.drewdaniels.com/dither.pdf

 

[HAWX]

http://www.drewdaniels.com/dither.pdf

WoW Wasn't expecting that fast. Thank you!

 

[Yuri Korzunov]

also you can encode something in 1 bit and retrieve more than 6db of dynamic from it. delta sigma DACs and DSD work because that is a reality. when you encode in 1bit you create a crippling noise that is added to the signal 6db below. if you stop there then sure it looks bad. but given the right tool and sample rate, you can apply noise shaping to move that crippling noise into another frequency range. and what's left below the first 6db is not nothing. what's left is the original signal in all its glory, down to the point where some other noise is again added to it for whatever reason like the limit of the noise shaping ability or thermal noise or whatever. that's why I've been insisting on looking at things as proper signal+some noise. because the right signal is always there. just not always alone. ^_^

Low bit number and noise shaping for signal-noise-ratio increasing is band reserve matter.