New DAC Chipset from AKM

[FeddyLost]

It looks like they had to rise frequencies so much that it will be troublesome in single chip. Once I have read some interview with DAC designer, who said that all "proper" delta-sigma processing would require a lot of mathematic calculating power, so all that we currently have is a mediocre substitute of ideal mathematic model.
It has increased the oversampling rate of digital processing to 256x from the previous 8x
If it plays sine waves like older designs, but handle impulses better (due to shorter ringing) it might be somehow beneficial in real life.
Downsides of current delta-sigma is disputable, but there must be some reason why some competent people prefer multibit designs.

 

[mansr]

It has increased the oversampling rate of digital processing to 256x from the previous 8x

That's the digital interpolation preceding the delta-sigma stage.

Downsides of current delta-sigma is disputable, but there must be some reason why competent people prefer multibit designs.

Modern delta-sigma designs are multibit.

 

[boXem]

It looks like they had to rise frequencies so much that it will be troublesome in single chip. Once I have read some interview with DAC designer, who said that all "proper" delta-sigma processing would require a lot of mathematic calculating power, so all that we currently have is a mediocre substitute of ideal mathematic model.
It has increased the oversampling rate of digital processing to 256x from the previous 8x
If it plays sine waves like older designs, but handle impulses better (due to shorter ringing) it might be somehow beneficial in real life.
Downsides of current delta-sigma is disputable, but there must be some reason why competent people prefer multibit designs.

Sorry to write this this way, but it smells like BS.
One of the niceties of delta sigma is that it doesn't require a lot of mathematical power.
Current delta sigma dacs are indeed multibit (I have 6 bits in mind for ESS).
Are you sure you are not confusing delta sigma with DSD?
@mansr may prove me wrong.

 

[Killingbeans]

If it plays sine waves like older designs, but handle impulses better (due to shorter ringing) it might be somehow beneficial in real life.

http://archimago.blogspot.com/2018/01/audiophile-myth-260-detestable-digital.html

Downsides of current delta-sigma is disputable, but there must be some reason why competent people prefer multibit designs.

Are you referring to R2R designs? And what area of competence are we talking about?

 

[FeddyLost]

Modern delta-sigma designs are multibit.

I meant R2R.

Are you referring to R2R designs? And what area of competence are we talking about?

Some of sound engineers prefer R2R designs.
Don't know why exactly.

As for myself, I'm a very untrained and undemanding (in general) listener so I just listen through DS converters and don't even had a chance to hear some decent R2R. Holo May is relatively new and expensive, and MSB/Rockna/ etc is beyond my budget.

Anyway, I just don't like to call any design exchanges useless, unless I definitely know exact reason for solutions applied. Digital sound is very complicated thing, as I see. Maybe there are some real improvements in resulting sound quality.

 

[Killingbeans]

Some of sound engineers prefer R2R designs.
Don't know why exactly.

I can only speculate.

I suspect it has nothing to do with sound quality. So far all I've seen is people claiming that R2R does something "special" that we haven't found a way of measuring yet. I'm not buying that. The most rational explanation would be that the idea of R2R simply speaks to the subjective in a pleasing way.

As for myself, I'm a very untrained and undemanding (in general) listener so I just listen through DS converters

Me too. I'm just a guy with a modest amount of technical know-how trying to make sense of a hobby that fascinates me. To be honest, I have very little first hand experience with high-end gear. My main reason for lurking around is to make my decisions in the future as rooted in empirical evidence as possible, without going to extremes.

and don't even had a chance to hear some decent R2R.

Me neither. In fact, I don't think I've ever purposely listened to any R2R DAC at all. So feel free to call me a bigot

Anyway, I just don't like to call any design exchanges useless, unless I definitely know exact reason for solutions applied.

The real question is whether the solution is for a problem that was never an issue to being with.

Maybe there are some real improvements in resulting sound quality.

Call me a pessimist, but I doubt it. Measurable improvements, sure. But audible?

 

[FeddyLost]

AFAIK, fundamentally, R2R or fully parallel converter, is able to play absolutely uncorrelated samples. Total random, 24 bits - 0 bits - 24 bits ...
DS is correlated by design (delta) so it have to work much faster to get same differential between samples if it have lesser internal bit depth (currently 6 bits usually). And it requires error feedback loops a lot of math to reduce existing correlations like ILA, DWA, DEM and hell knows what else...

And if you read datasheet for AK4191EQ, you'll find a lot of options and 64-bit DF, so it could be even power consumption issue (like low-nm outboard math with a lot of transistors and ordinary audio chip).

I'm not a DAC designer and can't even imagine all possible reasons that could be behind such solution.

Measurable improvements, sure. But audible?

Audible for ordinary guy from the street - barely. But there must be some reason for chipmaker to make something new in big series. It's not hand soldered premium hiend stuff, it's just a chipset which should be put in some equipment, and AKM expect good sales I think.
So, let's wait for something completed on this chipset and see.
Or hear.

 

[Killingbeans]

DS is correlated by design (delta) so it have to work much faster to get same differential between samples if it have lesser internal bit depth (currently 6 bits usually).

I wouldn't worry about it. That "much faster" speed is still laughably slow compared to what technology can offer these days.

But there must be some reason for chipmaker to make something new in big series.

Yes, because new things sell better and make the competition look bad

 

[mansr]

AFAIK, fundamentally, R2R or fully parallel converter, is able to play absolutely uncorrelated samples. Total random, 24 bits - 0 bits - 24 bits ...
DS is correlated by design (delta) so it have to work much faster to get same differential between samples if it have lesser internal bit depth (currently 6 bits usually). And it requires error feedback loops a lot of math to reduce existing correlations like ILA, DWA, DEM and hell knows what else...

The design spec of any DAC includes bandwidth, dynamic range, and distortion. Outside the specified bandwidth, dynamic range and distortion figures need not be maintained. In fact, the conversion stage itself, regardless of operational principle, always creates high-frequency junk that has to be removed with an analogue filter. For audio applications, it turns out that sigma-delta designs achieve much better performance at much lower cost compared to R-2R or similar. A sigma-delta DAC can reproduce an arbitrary 48 kHz sample sequence _better_ than an R-2R DAC. Oversampling being part of how it achieves this is neither here nor there.