DAC types and their sonic signature

[solderdude]

You mean that stepped look of a waveform from a HF range when produced by non-OS DACs?

Funny, Fourier analysis says the perfect base tone frequency sinus wave is present there in its pure sinus form. Harmonics of the base tone which are obviously as well present there make the resulting waveform look stepped.

That is ONLY true when the sample-and-hold stairsteps, coming out of an R2R DAC chip, have passed through a very steep post filter.
That filter is not present in external NOS DAC's.
And yes, the slow filter at 60kHz, or there about, are not fitting the criteria of a proper reconstruction filter.

The 'steps' are NOT the harmonics at all, they are quantization errors that need to be removed.
That is what the reconstruction filter is for.
The steps have the same relative amplitude (during a sample width) that a sample point (which is much smaller in time) had during the encoding process. There WAS no 'step' in the sampling process. Merely a point. As a DAC cannot do a 'point' it makes a 'stripe'.
This means ONLY at the beginning of that 'stripe' the sample value is correct. 99.99% behind that point is the incorrect value (considering sine waves are the input).

When oversampling is done the quantization errors are smaller in amplitude and higher in frequency.
Because of this a less steep and less complex analog post filter can be used.

 

[Calexico]

With 192khz high res you don't need oversampling at all with multibit dacs.
You can filter slowly from 20khzto 96khz with analog lpf. Then you don't make any process to the original master file.
Oversampling was added for 44.1khz because it was too difficult to have an analog filter that start at 20khz
and filter all after 22khz.
For high res there is no need for it.

With dsd if you have a dsd root file decoded by a dsd dac you don't do any process at all. You just need an analog lpf also.
Then both methodologies are no.

In modern dacs reconstruction filters are made because of the multibit delta sigma toplogy. Not because of theory.
And delta sigma multibit is chosen because of price and good marketing measurements.

If high res and multibit were the standard we wouldn't need any digital overdampling and filter and that would be better for sound

 

[solderdude]

With 192khz high res you don't need oversampling at all with multibit dacs.
You can filter slowly from 20khzto 96khz with analog lpf. Then you don't make any process to the original master file.
Oversampling was added for 44.1khz because it was too difficult to have an analog filter that start at 20khz
and filter all after 22khz.
For high res there is no need for it.

It doesn't work that way... You have a weird picture of how digital works...
192kHz files also will be reproduced using a sharp filter at around 90kHz.
You cannot select a 'slow' filter at 20kHz on a 192kHz file.

In modern dacs reconstruction filters are made because of the multibit delta sigma toplogy. Not because of theory.

What a bunch of nonsense. ... waiting for @SIY to show up telling you to get a refund.

And delta sigma multibit is chosen because of price and good marketing measurements.

If high res and multibit were the standard we wouldn't need any digital overdampling and filter and that would be better for sound

??

 

[Calexico]

You don't need oversampling with high res.
An analog lpf is enough.
I know modern dacs do it. That's because they are not multibit.
Then they have no choice other than oversampling whatever frequency (except for dsd).
With 192khz pcm an analog filter is enough with a true multibit dac to filter before nyquist and remove the image.

 

[Calexico]

Then to be sure you understood:
- multibit nos dac
- analog lpf that filter from 20khz (but a slow one is enough as it must be strong only at 96khz)
- high res file of 192khz
Would be ideal.
To work with 44.1khz it would need x4 oversampling from computer.

 

[solderdude]

You don't need oversampling with high res.
An analog lpf is enough.
I know modern dacs do it. That's because they are not multibit.
Then they have no choice other than oversampling whatever frequency (except for dsd).
With 192khz pcm an analog filter is enough with a true multibit dac to filter before nyquist and remove the image.

You don't 'need' oversampling with 44.1 kHz either.
With hires you need an equally sharp reconstruction filter, just at a higher frequency.

Your lack of understandig filters and DAC topologies is .. how to say it politely .... below par.

 

[Blumlein 88]

@Calexico with respect to you, it is clear you have very little understanding about how digital sampling and reconstruction work. And little understanding of the ideas behind them. It isn't clear what you are thinking in regards to what makes a good DAC, and that is because you are filled with misconceptions about them. Why do you hold such strong convictions about what would be good?

Further you complain about oversampling delta sigma DACs, then what you describe for a good way to do it is actually not far from what those kind of DACs do only your parameters aren't well chosen.

You say you don't need oversampling for hires, then describe sampling at high rates and using a slow filter for 20 khz bandwidth. That is oversampling. Then you tack on that for 44.1 khz the computer will need to oversample and feed it to your imagined 192 khz multi-bit DAC. It is a mish-mash of poorly conceived ideas about what is ideal. You clearly don't understand how it works.

Delta sigma DACs give you performance rather close to a theoretically perfect NOS DAC. Far better than can actually be achieved with an actual NOS DAC running at the native sampling rate. What do you think you are gaining.

 

[solderdude]

Then to be sure you understood:
- multibit nos dac
- analog lpf that filter from 20khz (but a slow one is enough as it must be strong only at 96khz)
- high res file of 192khz
Would be ideal.
To work with 44.1khz it would need x4 oversampling from computer.

Just to be sure you understood.

Filterless Multibit NOS DAC:
- plays 192kHz files with lots of crap above Nyquist at very high levels.
- plays 44kHz files with lots of crap above Nyquist at very high levels.

Well designed multibit NOS DAC:
- plays 192kHz files without crap above Nyquist.
- plays 44.1kHz files without crap above Nyquist.

Well designed multibit OS DAC:
- plays 192kHz files without crap above Nyquist.
- plays 44.1kHz files without crap above Nyquist.

DS DAC with 'fast filter':
- plays 192kHz files without crap above Nyquist.
- plays 44.1kHz files without crap above Nyquist.

DS DAC with 'slow filter':
- plays 192kHz files with some crap above Nyquist.
- plays 44.1kHz files with some crap above Nyquist.

DS DAC in NOS mode (yes, it exists):
- plays 192kHz files with lots of crap above Nyquist at very high levels.
- plays 44kHz files with lots of crap above Nyquist at very high levels.

This we can validate with measurements... there is no mystery nor magic.. it is science and electronics.


You know what would be ideal ? If you understood how DA conversion really worked.

 

[zalive]

That is ONLY true when the sample-and-hold stairsteps, coming out of an R2R DAC chip, have passed through a very steep post filter.
That filter is not present in external NOS DAC's.
And yes, the slow filter at 60kHz, or there about, are not fitting the criteria of a proper reconstruction filter.

The 'steps' are NOT the harmonics at all, they are quantization errors that need to be removed.
That is what the reconstruction filter is for.
The steps have the same relative amplitude (during a sample width) that a sample point (which is much smaller in time) had during the encoding process. There WAS no 'step' in the sampling process. Merely a point. As a DAC cannot do a 'point' it makes a 'stripe'.
This means ONLY at the beginning of that 'stripe' the sample value is correct. 99.99% behind that point is the incorrect value (considering sine waves are the input).

When oversampling is done the quantization errors are smaller in amplitude and higher in frequency.
Because of this a less steep and less complex analog post filter can be used.

If you have a stable waveform when sampling one frequency then Fourier analysis says you can get any waveform by solely applying its harmonics. So is it a stable waveform? It should be stable because of equal time units applied in sampling process. And it shows on oscilloscope as a stable stepped waveform, right? But even that is not important. When you have a base frequency, it can only be a sine wave. Its harmonics show on a measurement waveform but there will still be a sine wave beneath. Filtering removes those higher frequencies but it doesn't change the ground sine wave of a base frequency. It's there, no matter whether you remove harmonics (by filtering) or not. And with non-OS, you don't need any filter. I use currently one such filterless NOS AD1865 DAC and it sounds great.

Filterless Multibit NOS DAC:
- plays 192kHz files with lots of crap above Nyquist at very high levels.
- plays 44kHz files with lots of crap above Nyquist at very high levels.

That's true, there's some crap there beyond audible range. Possibly a bit in audible range too, but you can't discern anything bothering.
Local amateur audio designer talked against it, he said it's stupid to send this non-filtered load to the speakers. Now he uses one, he loved the sound and he doesn't care no more for the crap

 

[Blumlein 88]

If you have a stable waveform when sampling of one frequency then Fourier analysis says you can get any waveform by solely applying its harmonics. So is it a stable waveform? It should be stable because of equal time units applied in sampling process. And it shows on oscilloscope as a stable stepped waveform, right? But even that is not important. When you have a base frequency, it can only be a sine wave. Its harmonic show on a measurement waveform but there will still be a sine wave beneath. Filtering removes those higher frequencies but it does no change to the ground sine wave of a base frequency. It's there no matter whether you remove harmonics or not. And with non-OS, you don't need any filter. I use currently one such filterless NOS AD1865 DAC and it sounds great.



That's true, there's some crap there beyond audible range. Possibly a bit in audible range too, but you can't discern anything bothering.
Local amateur audio designer talked against it, he said it's stupid to send this non-filtered load to the speakers. Now he uses one, he loved the sound and he doesn't care no more for the crap

Filterless DACs are broken by design. They might work anyway if you basically have the crap high enough in frequency your ears, and your speakers effectively form filters. Now what is the advantage of not having a proper filter? And just saying it sounds better isn't a good enough answer unless you have data to show how the reproduced waveform is better in some sense.

 

[Thomas savage]

Member @Calexico is banned with immediate effect upon discovery of the fact he's a previously banned individual who's signed up under another name .. formally @Bluespower .

 

[zalive]

Filterless DACs are broken by design. They might work anyway if you basically have the crap high enough in frequency your ears, and your speakers effectively form filters. Now what is the advantage of not having a proper filter? And just saying it sounds better isn't a good enough answer unless you have data to show how the reproduced waveform is better in some sense.

Simple digital filters introduce audible ringing. Why it doesn't show in those simple tests - I don't know. My current knowledge is certainly not enough to even try to explain. However, mind that some of best sigma-delta DAC designs paid lot of attention to filtering, and a simple brick wall filter was not the choice. It seems some of world's best audio designers at one point become aware that hard digital filtering such as brick wall filter is, is not the best idea for the sound, and made a lot of effort to build a better custom filter. At the same time some expensive non-OS DACs like Audio Note's avoid filters. Again...why? One must really ask himself, do those guys know/understand something that I don't?

 

[zalive]

No single person came with sound impressions on link sample I put. It was really simple and quick, to listen to the sound blindly and comment how do you like it. Just your preferrence.

 

[Blumlein 88]

Simple digital filters introduce audible ringing. Why it doesn't show in those simple tests - I don't know. My current knowledge is certainly not enough to even try to explain. However, mind that some of best sigma-delta DAC designs paid lot of attention to filtering, and a simple brick wall filter was not the choice. It seems some of world's best audio designers at one point become aware that hard digital filtering such as brick wall filter is, is not the best idea for the sound, and made a lot of effort to build a better custom filter. At the same time some expensive non-OS DACs like Audio Note's avoid filters. Again...why? One must really ask himself, do those guys know/understand something that I don't?

Sigma Delta DACs use considerable oversampling with analog and digital filtering.

Is the ringing you are referring to that resulting from bandwidth limiting? Did you know if you use analog filtering it also rings in this situation?

Audio Note? Why would I care what they do. I get the whole tube thing, but who knows what their motivation is other than a story of not being like other people. They make a very expensive amp that some good headphone amps could out-muscle on power.

 

[Blumlein 88]

No single person came with sound impressions on link sample I put. It was really simple and quick, to listen to the sound blindly and comment how do you like it. Just your preferrence.

I think I missed the link, can you tell me which post it was in?

 

[jsrtheta]

I think Bose get a bad rap in some ways. I am not advocating their products but they read the way the market would go with respect to wireless active speakers and user (and fashion) friendly audio very early on and have been very successful serving their identified target markets. Weirdly I find those I know who buy Bose products are invariably very happy with them and keep them for a long time and aren't constantly fiddling and buying new gear like many who have a much greater knowledge of audio. And they do have a good research heritage.
The Bose products I do rate highly are their noise cancelling headphones, their noise cancelling know how is superb and if people want NC headphones then I would just point them to Bose or the top Sony NC models.

I've noticed a strong correlation between people who own Bose and people who drive Volvos.

 

[jsrtheta]

Archimago has a lot on digital filters and ringing, here: http://archimago.blogspot.com/search?q=ringing

 

[JohnYang1997]

Member @Calexico is banned with immediate effect upon discovery of the fact he's a previously banned individual who's signed up under another name .. formally @Bluespower .

What a relief. Thank you boss.

 

[zalive]

I think I missed the link, can you tell me which post it was in?

Blind test:
Listen to this without looking the content (it's a yt link so obviously looking is not desireable if you want to take it blind).
Blind Test Sample

Then describe the sound and rate it. Afterwards freely look what's the system (but pls don't spoil with comments which would reveal details about the system).

This, page 8...I'm interested in sound impressions through blind listening. I myself didn't listen blind to it, however I'm curious as well will blind listening impressions be still similar to mine.

 

[Jim777]

I've noticed a strong correlation between people who own Bose and people who drive Volvos.

Hahahaha....................... P. S. Sony's noise canceling headphones are probably better than Bose's.