[Yuri Korzunov]

how about 24bit vs 32bit? (32 bit 384khz)

Both bit and sample rate is potential noise matter.

If any parameter increasing is applied, need check result. There are many variables for univocal answer.

 

[gregorio]

Both bit and sample rate is potential noise matter.

That sounds suspiciously like one of the most common tricks used in audiophile marketing BS!

While your statement is literally true, does that "potential" actually apply or is it completely out of context? It definitely does apply in certain areas of the digital audio chain, for example, pro ADCs have for decades absolutely relied on manipulating/exchanging bit depth and sample rate for noise (and filtering) reasons. But, does your statement apply to distribution file formats? Answer: Absolutely not! It's not even vaguely close to being applicable to distribution file formats!

G

 

[Yuri Korzunov]

But, does your statement apply to distribution file formats?

What purpose of distibution?

 

[gregorio]

What purpose of distibution?

I'm assuming the poster was referring to distribution to consumers. However, I can't think of any circumstances where distributing audio files at 32bit would make any difference, even in the case of distribution between audio professionals for further processing.

G

 

[Yuri Korzunov]

I can't think of any circumstances where distributing audio files at 32bit would make any difference, even in the case of distribution between audio professionals for further processing.

Which parameters difference you mean?

 

[gregorio]

Which parameters difference you mean?

Any parameters!

 

[Yuri Korzunov]

I don't know such parameter.

 

[castleofargh]

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

sure, I wasn't trying to entertain the idea that noise shaping was magically creating bits at no cost. even alchemists didn't believe in making stuff out of nothing.
DSD turns sample rate into more dynamic in a given frequency range, MQA uses some bit values to store more samples. but tricks only go so far.
my point was simply to say that a 1 bit encoding didn't necessarily condemned the final signal to have only 1 bit of dynamic.

 

[Yuri Korzunov]

sure, I wasn't trying to entertain the idea that noise shaping was magically creating bits at no cost. even alchemists didn't believe in making stuff out of nothing.
DSD turns sample rate into more dynamic in a given frequency range, MQA uses some bit values to store more samples. but tricks only go so far.
my point was simply to say that a 1 bit encoding didn't necessarily condemned the final signal to have only 1 bit of dynamic.

Noise shaping is system with limited stability. Because it have feedback.
When the system work about overload limit, it can come to unstable state: silence or contant sine(s). For back to normal state need reset noise shaper.

If noise shape have sloping transient from lowest to maximal level noise (for dithering, as example), we work far stability limit. But 1 bit on 2.8 MHz is hard in implementation for providing noise level close to 24 bit PCM.

 

[gregorio]

But 1 bit on 2.8 MHz is hard in implementation for providing noise level close to 24 bit PCM.

Again, what you say is true but also again, what difference does it make? To put it in terms of the quote; Why do you need a consumer distribution file format to provide "noise level close to 24 bit PCM"?

Yes, 1 bit DSD has problems, limitations and difficulties in implementation but AGAIN, just stating this is typical of marketing BS, because it ignores context, the fact that other aspects of DSD implementation are far easier/cheaper. And yes, there is potentially some instability but in practice that was overcome (beyond audibility) over 20 odd years ago. So in effect this is just like a typical audiophile marketing "red herring", a real problem that we/our product solves, which is simply a lie because either the problem, though real, does not actually affect digital audio or it's a problem which does affect digital audio but was already solved years/decades ago and should already be built into ANY competent product!

G

 

[Yuri Korzunov]

Again, what you say is true but also again, what difference does it make? To put it in terms of the quote; Why do you need a consumer distribution file format to provide "noise level close to 24 bit PCM"?

Yes, 1 bit DSD has problems, limitations and difficulties in implementation but AGAIN, just stating this is typical of marketing BS, because it ignores context, the fact that other aspects of DSD implementation are far easier/cheaper. And yes, there is potentially some instability but in practice that was overcome (beyond audibility) over 20 odd years ago. So in effect this is just like a typical audiophile marketing "red herring", a real problem that we/our product solves, which is simply a lie because either the problem, though real, does not actually affect digital audio or it's a problem which does affect digital audio but was already solved years/decades ago and should already be built into ANY competent product!

What is marketing definition?

 

[Arpiben]

Dealing with noise shaping, the audio layman I am, always wondered about the added value of high order Sigma Delta Modulators in consumer DACs.
For sure,in theory, a 5 bit 17th order shaper with OSR=2048 may increase digital dynamic range around 50bits.
In a mixing or heavy DSP environment I may find some advantages. In a DAC perspective, I failed to understand the reasons of such range.
Marketing BS probably?

 

[Yuri Korzunov]

Dealing with noise shaping, the audio layman I am, always wondered about the added value of high order Sigma Delta Modulators in consumer DACs.
For sure,in theory, a 5 bit 17th order shaper with OSR=2048 may increase digital dynamic range around 50bits.
In a mixing or heavy DSP environment I may find some advantages. In a DAC perspective, I failed to understand the reasons of such range.
Marketing BS probably?

Order is not target there.
Order of noise shaper filter need for achieving steeper amplitude response of noise shaper filter.
It can increase signal/noise ratio in wider band of low frequencies.
But it can reduce stability of sigma-delta modulator to overload.

Need carefully increase steepness of the filter.
Easier way is increasing of sample rate.
It allow to decrease steepness of the filter to increase signal/noise ratio in wider band and keep the stability.

 

[Arpiben]

Order is not target there.
Order of noise shaper filter need for achieving steeper amplitude response of noise shaper filter.
It can increase signal/noise ratio in wider band of low frequencies.
But it can reduce stability of sigma-delta modulator to overload.

Need carefully increase steepness of the filter.
Easier way is increasing of sample rate.
It allow to decrease steepness of the filter to increase signal/noise ratio in wider band and keep the stability.

Relaxing filtering, reducing costs and avoiding instability all right,fine. But do we really need for example,discrete shapers running at 104MHz with high orders in a DAC ? That was/is my wonder.
Some pretend that by increasing noise shaping characteristics you increase depth perception....

 

[Yuri Korzunov]

Relaxing filtering, reducing costs and avoiding instability all right,fine.

17th order is too much, in my opinion for 104 MHz.
I'd refer to -170 ... -200 dB level, because it is give ability of "transparent for user" work with audio stuff.
Therefore enough 7...10th order approximately.
Need also account that precision used for calculations. If there 24 bit is used, we can decrease filter order for providing noise level comparable with quantization noise.
Here I can't call certain figures (need model the case). May be 5th order is enought.
Reducing precision and order may be reasonable for DAC, because after the DAC no other processings and DAC is limited by electronic component's noise level.