High resolution audio. The science, or lack of...?

[shadders]

As has been pointed out, analogue does not & cannot have infinite resolution.

Hi,

Can you explain why this is not the case ?. Thanks.

Regards,

Shadders.

All real world analogue systems have noise superimposed on the wanted signal. Suppose the wanted signal changes by a very small amount, but this small amount is less than the noise amplitude - it will be impossible to tell whether the signal has changed because of an intended change, or just because of noise. So the 'resolution' of an analogue system is a function of the noise inherent in the system. Hence why SNR (signal to noise ratio) has meaning for an analogue system but can also be derived from the resoultion of a digital system.

Hi,

Thanks.

I understand that small signals will be lost in the noise - but they will still be processed by the amplifier. An amplifier may not be able to resolve the signal for you to hear, but it does still process the signal.

So, an amplifier can have infinite resolution, but obviously from a practical aspect, that resolution is impacted by system noise - not useable. Perhaps resolution is not the correct term for analogue electronics, and we should use S/N.

Regards,

Shadders.

 

[jcbrum]

I was asking regarding whether you were stating dither is noise shaping. Thanks

Dither and noise shaping are different things, but since dither is an essential process in this topical context, and dither produces noise, then it is usual to apply dither and noise shaping together in the same process application, so as to produce an optimised result for human ears.

JC

 

[busb]

Hi,

This is not correct . . // . . Dither does NOT nullify quantisation error . . // . . . Regards,

Shadders. .

Ah well, . . . It seems we cannot agree, and probably should leave it there, for the time being. I fear to go on with such a technical argument may well be tedious for many readers who might not wish to study the theory in such detail.

It's a good job that Claude Shannon was such a brilliant mathematician, and working with Harry Nyquist's theorem, laid out all the rules for digital audio so precisely, so that we don't have to.

JC

Hi,

I introduced the dither argument since :

1. Dither operates on the LSB of a 16bit word for CD and hence is at the bottom end of the maximum Signal to Noise ratio, that is -96dB to -90dB.

2. fr0g has stated and others agreed that you cannot hear signals at the -90dB level.

If fr0g et al are correct, then why is dither applied to 16bit audio samples ?

Dither is applied to remove the artefacts at the -96dB to -90dB level since they can be heard. (laymans terms used)

fr0g's analysis is that you cannot hear -90dB. This is in error based on fr0gs statement that the -90dB value is an absolute signal level in his plot.

It is not, it is a difference value, which equates to 0.003% difference.

I could construct a waveform - frequency sweep that will be within 0.003% of the 16bit/44.1kHz spectrum fr0g produced. The two waveforms, fr0gs audio track, my frequency sweep, differ by only 0.003%, and hence have the same sprectral signature. Obviously they are completely different in the time domain. For fr0g to state they are not different - then his test must be true for all cases.

They are not true for all cases - hence fr0g's conclusion is not correct.

This is a case of misunderstanding the process results to obtain the wrong conclusion.

Regards,

Shadders

?? Dither is used to remove otherwise audible low level distortion (well above -90 to -96dB) at the expense of wideband but inaudible noise. This noise can be fairly broadband or shoved up towards the HF. It's amazing what randomising the LSBs can achieve. It took me ages to work out what the hell dither was for! It's also important to dither the conversion from 24 to 16 bit appropriately to avoid truncation artefacts. Any effective (downward) conversion of bit rate will be inexact. Applied dither may well reside with the LSB but much analogue noise will impinge on higher bits during initial conversion that will effectively do the same thing as pseudo random/Gausian/triangular dither does.

Hi,

Agree that if the signal is not full scale deflection, that effectively the number of bits used is less than 16bits, which then reduces the S/N ratio, and hence the dither acting upon the LSB, will in effect be higher with regards to the noise floor it operates in.

I do not see how it impinges upon the higher bits, if only the LSB is modulated, and the DAC is monotonic.

If the signal is not full scale then the -90dB to -96dB is not the range of the dither noise.

Regards,

Shadders.

The noise introduced by dither may reside in the LSBs but it's going to be superimposed on ALL output. If it wasn't, it wouldn't do anything apart from raising the noisefloor. I don't understand your use of the term "monotonic" here.

 

[jcbrum]

Hi,

Thanks.

I understand that small signals will be lost in the noise - but they will still be processed by the amplifier. An amplifier may not be able to resolve the signal for you to hear, but it does still process the signal.

So, an amplifier can have infinite resolution, but obviously from a practical aspect, that resolution is impacted by system noise - not useable. Perhaps resolution is not the correct term for analogue electronics, and we should use S/N.

Regards,

Shadders.

It's an interesting point.

Mathematically, noise and distortion are the same thing, however they arise. So it's reasonable to use the terms 'resolution' or 'SNR' pretty much equally, as long as it's understood what is being referred to.

However in human hearing terms, and common language (neither methematical nor scientific) they might convey different interpretations.

Most people probably think of 'noise' as something like 'hiss', and 'distortion' as some 'roughening' of the signal. I don't think many people actualy know what 'resolution' is, but it's probably considered as something to do with being able to discern small differences.

Actually, it's all just 'signal' with some 'noise' as well.

JC

 

[shadders]

I was asking regarding whether you were stating dither is noise shaping. Thanks

Dither and noise shaping are different things, but since dither is an essential process in this topical context, and dither produces noise, then it is usual to apply dither and noise shaping together in the same process application, so as to produce an optimised result for human ears.

JC

Hi

Apologies - i misread a post as yours - hence the dither and noise shaping question must have perplexed you.

I think that nosie shaping is used in Sigma-Delta DAC's to push the noise into the higher bandwidths.

Regards,

Shadders.

 

[shadders]

Hi,

This is not correct . . // . . Dither does NOT nullify quantisation error . . // . . . Regards,

Shadders. .

Ah well, . . . It seems we cannot agree, and probably should leave it there, for the time being. I fear to go on with such a technical argument may well be tedious for many readers who might not wish to study the theory in such detail.

It's a good job that Claude Shannon was such a brilliant mathematician, and working with Harry Nyquist's theorem, laid out all the rules for digital audio so precisely, so that we don't have to.

JC

Hi,

I introduced the dither argument since :

1. Dither operates on the LSB of a 16bit word for CD and hence is at the bottom end of the maximum Signal to Noise ratio, that is -96dB to -90dB.

2. fr0g has stated and others agreed that you cannot hear signals at the -90dB level.

If fr0g et al are correct, then why is dither applied to 16bit audio samples ?

Dither is applied to remove the artefacts at the -96dB to -90dB level since they can be heard. (laymans terms used)

fr0g's analysis is that you cannot hear -90dB. This is in error based on fr0gs statement that the -90dB value is an absolute signal level in his plot.

It is not, it is a difference value, which equates to 0.003% difference.

I could construct a waveform - frequency sweep that will be within 0.003% of the 16bit/44.1kHz spectrum fr0g produced. The two waveforms, fr0gs audio track, my frequency sweep, differ by only 0.003%, and hence have the same sprectral signature. Obviously they are completely different in the time domain. For fr0g to state they are not different - then his test must be true for all cases.

They are not true for all cases - hence fr0g's conclusion is not correct.

This is a case of misunderstanding the process results to obtain the wrong conclusion.

Regards,

Shadders

?? Dither is used to remove otherwise audible low level distortion (well above -90 to -96dB) at the expense of wideband but inaudible noise. This noise can be fairly broadband or shoved up towards the HF. It's amazing what randomising the LSBs can achieve. It took me ages to work out what the hell dither was for! It's also important to dither the conversion from 24 to 16 bit appropriately to avoid truncation artefacts. Any effective (downward) conversion of bit rate will be inexact. Applied dither may well reside with the LSB but much analogue noise will impinge on higher bits during initial conversion that will effectively do the same thing as pseudo random/Gausian/triangular dither does.

Hi,

Agree that if the signal is not full scale deflection, that effectively the number of bits used is less than 16bits, which then reduces the S/N ratio, and hence the dither acting upon the LSB, will in effect be higher with regards to the noise floor it operates in.

I do not see how it impinges upon the higher bits, if only the LSB is modulated, and the DAC is monotonic.

If the signal is not full scale then the -90dB to -96dB is not the range of the dither noise.

Regards,

Shadders.

The noise introduced by dither may reside in the LSBs but it's going to be superimposed on ALL output. If it wasn't, it wouldn't do anything apart from raising the noisefloor. I don't understand your use of the term "monotonic" here.

Hi,

Can you explain how dither affects the higher bits ?. Thanks.

A monotonic function is where an input to a system produces a single output and no other possible output, such as a graph y = x.

If it was y = x ^ 2 - then y could be basedd on -x or x.

Regards,

Shadders.

 

[jcbrum]

Can you explain how dither affects the higher bits ?. Thanks.

It randomises their values with respect to the time domain.

A monotonic function is where an input to a system produces a single output and no other possible output, such as a graph y = x.

If it was y = x ^ 2 - then y could be basedd on -x or x.

Regards,

Shadders.

In your terms, digital systems are not naturally monotonic, unless forced into that condition, usually by an analogue condition.

JC

p.s. I've gotta go shopping, see you later

 

[busb]

As has been pointed out, analogue does not & cannot have infinite resolution.

Hi,

Can you explain why this is not the case ?. Thanks.

Regards,

Shadders.

All real world analogue systems have noise superimposed on the wanted signal. Suppose the wanted signal changes by a very small amount, but this small amount is less than the noise amplitude - it will be impossible to tell whether the signal has changed because of an intended change, or just because of noise. So the 'resolution' of an analogue system is a function of the noise inherent in the system. Hence why SNR (signal to noise ratio) has meaning for an analogue system but can also be derived from the resoultion of a digital system.

Hi,

Thanks.

I understand that small signals will be lost in the noise - but they will still be processed by the amplifier. An amplifier may not be able to resolve the signal for you to hear, but it does still process the signal.

So, an amplifier can have infinite resolution, but obviously from a practical aspect, that resolution is impacted by system noise - not useable. Perhaps resolution is not the correct term for analogue electronics, and we should use S/N.

Regards,

Shadders.

Absolutely! Noise is the work of the devil. Take a signal booster for your TV. It's purpose is to boost a weak signal. The problem is that the weak signal has a lot of noise that will also be boosted! Any amplification needs to be carried out before local losses degrade the S/N ratio so a signal booster should be built into the aerial before it degrades further down the coax cable. A geeky aside would be to downconvert that RF signal at the masthead to IF first but I digress from audio..

 

[busb]

Hi,

This is not correct . . // . . Dither does NOT nullify quantisation error . . // . . . Regards,

Shadders. .

Ah well, . . . It seems we cannot agree, and probably should leave it there, for the time being. I fear to go on with such a technical argument may well be tedious for many readers who might not wish to study the theory in such detail.

It's a good job that Claude Shannon was such a brilliant mathematician, and working with Harry Nyquist's theorem, laid out all the rules for digital audio so precisely, so that we don't have to.

JC

Hi,

I introduced the dither argument since :

1. Dither operates on the LSB of a 16bit word for CD and hence is at the bottom end of the maximum Signal to Noise ratio, that is -96dB to -90dB.

2. fr0g has stated and others agreed that you cannot hear signals at the -90dB level.

If fr0g et al are correct, then why is dither applied to 16bit audio samples ?

Dither is applied to remove the artefacts at the -96dB to -90dB level since they can be heard. (laymans terms used)

fr0g's analysis is that you cannot hear -90dB. This is in error based on fr0gs statement that the -90dB value is an absolute signal level in his plot.

It is not, it is a difference value, which equates to 0.003% difference.

I could construct a waveform - frequency sweep that will be within 0.003% of the 16bit/44.1kHz spectrum fr0g produced. The two waveforms, fr0gs audio track, my frequency sweep, differ by only 0.003%, and hence have the same sprectral signature. Obviously they are completely different in the time domain. For fr0g to state they are not different - then his test must be true for all cases.

They are not true for all cases - hence fr0g's conclusion is not correct.

This is a case of misunderstanding the process results to obtain the wrong conclusion.

Regards,

Shadders

?? Dither is used to remove otherwise audible low level distortion (well above -90 to -96dB) at the expense of wideband but inaudible noise. This noise can be fairly broadband or shoved up towards the HF. It's amazing what randomising the LSBs can achieve. It took me ages to work out what the hell dither was for! It's also important to dither the conversion from 24 to 16 bit appropriately to avoid truncation artefacts. Any effective (downward) conversion of bit rate will be inexact. Applied dither may well reside with the LSB but much analogue noise will impinge on higher bits during initial conversion that will effectively do the same thing as pseudo random/Gausian/triangular dither does.

Hi,

Agree that if the signal is not full scale deflection, that effectively the number of bits used is less than 16bits, which then reduces the S/N ratio, and hence the dither acting upon the LSB, will in effect be higher with regards to the noise floor it operates in.

I do not see how it impinges upon the higher bits, if only the LSB is modulated, and the DAC is monotonic.

If the signal is not full scale then the -90dB to -96dB is not the range of the dither noise.

Regards,

Shadders.

The noise introduced by dither may reside in the LSBs but it's going to be superimposed on ALL output. If it wasn't, it wouldn't do anything apart from raising the noisefloor. I don't understand your use of the term "monotonic" here.

Hi,

Can you explain how dither affects the higher bits ?. Thanks.

A monotonic function is where an input to a system produces a single output and no other possible output, such as a graph y = x.

If it was y = x ^ 2 - then y could be basedd on -x or x.

Regards,

Shadders.

A monotonic function - got it. Non aliasing in other words

 

[Anonymous]

[EDITED BY MODS - and user banned for multiple registration]

 

[altruistic.lemon]

Hey, how're the AVIs going, mate? Still happy?

 

[fr0g]

I agree with this guy:

http://people.xiph.org/~xiphmont/demo/neil-young.html

Yes, excellent and informative article.

"Thus, 16 bit audio can go considerably deeper than 96dB. With use of shaped dither, which moves quantization noise energy into frequencies where it's harder to hear, the effective dynamic range of 16 bit audio reaches 120dB in practice [13], more than fifteen times deeper than the 96dB claim.

120dB is greater than the difference between a mosquito somewhere in the same room and a jackhammer a foot away.... or the difference between a deserted 'soundproof' room and a sound loud enough to cause hearing damage in seconds.

16 bits is enough to store all we can hear, and will be enough forever."




"It's true enough that a properly encoded Ogg file (or MP3, or AAC file) will be indistinguishable from the original at a moderate bitrate."

Absolutely.

 

[jcbrum]

Shadders, Your enquiry as to exactly how quantisation errors, and applied dither, affect the LSB is dealt with in this paper . . .

Optimal Dither and Noise Shaping

Your liking for mathematical analysis will be fully satisfied by this paper, but it's rather long and complex for the average casual reader.

JC

 

[shadders]

Shadders, Your enquiry as to exactly how quantisation errors, and applied dither, affect the LSB is dealt with in this paper . . .

Optimal Dither and Noise Shaping

Your liking for mathematical analysis will be fully satisfied by this paper, but it's rather long and complex for the average casual reader.

JC

Hi,

Thanks for this - much appreciated - i will read later tomorrow.

Regards,

Shadders.

 

[shadders]

Hi,

Yes, excellent and informative article.

"Thus, 16 bit audio can go considerably deeper than 96dB. With use of shaped dither, which moves quantization noise energy into frequencies where it's harder to hear, the effective dynamic range of 16 bit audio reaches 120dB in practice [13], more than fifteen times deeper than the 96dB claim.

I read the paper very briefly - the reference [13] is not to a paper - there is not proof that 16bits can extend to 120dB for audio. This reference is for pure tone testing only.

120dB is greater than the difference between a mosquito somewhere in the same room and a jackhammer a foot away.... or the difference between a deserted 'soundproof' room and a sound loud enough to cause hearing damage in seconds.

16 bits is enough to store all we can hear, and will be enough forever."

He has based this claim on the above previous regarding the single tone test, which does not apply to audio music files. He has no proof that 16bits for audio files can extend to -120dB. If he has - can you provide this ?. Thanks.

"It's true enough that a properly encoded Ogg file (or MP3, or AAC file) will be indistinguishable from the original at a moderate bitrate."

Absolutely.

This is not relevant to the discussion where your analysis has been used in error. No data given in his statement - generalisations are not fact.

Regards,

Shadders.

 

[CnoEvil]

It is great that this interesting and knowledgeable discussion has mostly been conducted with politeness and respect....even if some of it leaves me scratching my head.

What has become clear, is that this topic is far from straight forward, and simplistic answers are probably wrong.

I'm delighted that Frog used my suggested track, because if there is an audible difference (between Linn's own 16 & 24 bit versions), this is as good an example as any to use.

On that note, I am retreating back to the sidelines.,,,,as you were.

 

[jcbrum]

On that note, I am retreating back to the sidelines.,,,,as you were.

I know how you feel, - me too

JC

 

[manicm]

other engineering types should not be allowed to design anything, nothin', nothin' Geeks are even worst than them arty farts!

So engineers should not be allowed to design speakers. Ok then who will? You? I love rants like this. They're amusing.

 

[manicm]

If it's inaudible, it doesn't matter whether it's present or absent.

Frog couldn't hear it, and I don't think I could, and I don't think anyone could.

That's it as far as I'm concerned.

Regards, JC

So much for not trusting one's ears then.

 

[manicm]

As soon as you introduce analogue to digital conversion you will introduce some kind of noise. It's only the other way around where the signal in the audible spectrum can be perfectly recreated and tested this way.

With their new Exakt system Linn would disagree with you.

I doubt it.

Wanna bet mate? Linn have done demos with customers connecting Linn Sondeks to the Exakt - many have been really surprised and impressed - even compared to the conventional analogue Sondek based systems. Yes the Exakt contains an ADC for analogue sources.

 

[jcbrum]

If it's inaudible, it doesn't matter whether it's present or absent.

Frog couldn't hear it, and I don't think I could, and I don't think anyone could.

That's it as far as I'm concerned.

Regards, JC

So much for not trusting one's ears then.

Nothing wrong with trusting your ears . . .

It's your brains which are the suspect item.



JC

 

[busb]

other engineering types should not be allowed to design anything, nothin', nothin' Geeks are even worst than them arty farts!

So engineers should not be allowed to design speakers. Ok then who will? You? I love rants like this. They're amusing.

Rants maybe amusing, even mine maybe amusing but please try the other end

 

[busb]

On that note, I am retreating back to the sidelines.,,,,as you were.

I know how you feel, - me too

JC

Yup - Fourier analysis is bad enough but Hilbert transforms are a wee bit cloudy (to me)! I'm grateful that many, many people do understand this stuff.

 

[fr0g]

Hi,

Yes, excellent and informative article.

"Thus, 16 bit audio can go considerably deeper than 96dB. With use of shaped dither, which moves quantization noise energy into frequencies where it's harder to hear, the effective dynamic range of 16 bit audio reaches 120dB in practice [13], more than fifteen times deeper than the 96dB claim.

I read the paper very briefly - the reference [13] is not to a paper - there is not proof that 16bits can extend to 120dB for audio. This reference is for pure tone testing only.

120dB is greater than the difference between a mosquito somewhere in the same room and a jackhammer a foot away.... or the difference between a deserted 'soundproof' room and a sound loud enough to cause hearing damage in seconds.

16 bits is enough to store all we can hear, and will be enough forever."

He has based this claim on the above previous regarding the single tone test, which does not apply to audio music files. He has no proof that 16bits for audio files can extend to -120dB. If he has - can you provide this ?. Thanks.

"It's true enough that a properly encoded Ogg file (or MP3, or AAC file) will be indistinguishable from the original at a moderate bitrate."

Absolutely.

This is not relevant to the discussion where your analysis has been used in error. No data given in his statement - generalisations are not fact.

Regards,

Shadders.

You are definitely over analysing now Shadders. I know it isn't relevant, I am simply expressing my agreement to the article

 

[manicm]

If it's inaudible, it doesn't matter whether it's present or absent.

Frog couldn't hear it, and I don't think I could, and I don't think anyone could.

That's it as far as I'm concerned.

Regards, JC

So much for not trusting one's ears then.

Nothing wrong with trusting your ears . . .

It's your brains which are the suspect item.



JC

Ah, no.