AK unveils new 'AK4498' DAC with dedicated D-S chip

[Veri]

"AK4191EQ and AK4498EQ split the traditional DAC's digital filtering & delta sigma modulation and analog parts to reduce the interference between the digital and analog parts. The general idea is similar to removing the current-voltage conversion with the AK4499EQ. The more scattered the better, moderately reduce the degree of integration to reduce signal interference. As for the indicator, it is inconvenient to directly compare with the AK4499EQ because it is split into two parts.
While conventional DACs use a single chip to convert signals from digital to analog, new products split the conversion process between the AK4191 delta-sigma modulator and the AK4498 DAC. This innovative approach has resulted in an improvement in sound quality that is easily perceptible while listening to music."

Not much of a fan of the marketing blurb

Some official AKM info here:
https://www.akm.com/global/en/products/audio/audio-dac/ak4498eq/
https://www.akm.com/global/en/products/audio/audio-dac/ak4191eq/

I read about it at this diyaudio thread:
https://www.diyaudio.com/forums/digital-line-level/350733-ak4191eq.html

These new chips supposedly are not targeted for manufacturing until Q2 2021

 

[mansr]

So we're back where it all started with the SAA7030 and TDA1540.

 

[boXem]

Looking at the few numbers given, I don't see the interest compared to the AK4497. Better filter maybe?
And Akm continues with the bad habit (since AK4493) of not publishing their specs...

 

[mansr]

Looking at the few numbers given, I don't see the interest compared to the AK4497. Better filter maybe?

It appears that they're now doing proper interpolation up to 256x rather than 8x followed by zero-order hold.

And Akm continues with the bad habit (since AK4493) of not publishing their specs...

AKM have gone full audiophool.

 

[Veri]

Looking at the few numbers given, I don't see the interest compared to the AK4497. Better filter maybe?

I suppose so. After all their rather slow attenuating filters I suppose this time they'll do it "right"

 

[boXem]

It appears that they're now doing proper interpolation up to 256x rather than 8x followed by zero-order hold.


AKM have gone full audiophool.

These two sentences look like fitting well together. I had in mind that "proper interpolation" has no interest for audio above ~400 kHz.

 

[majingotan]

It appears that they're now doing proper interpolation up to 256x rather than 8x followed by zero-order hold.

I smell some burning Chord in this! 256x is similar to 256Fs from that Chord stuff

 

[mansr]

These two sentences look like fitting well together. I had in mind that "proper interpolation" has no interest for audio above ~400 kHz.

256x interpolation ought to be easy enough with modern IC processes, and it is technically preferable to zero-order hold. Unless targeting very low power or low cost, I see no reason to not do it in a new design. Most audio DAC chips on the market seem to be based on fairly old designs even if they've received the occasional tweak.

 

[Yviena]

So they are basically doing what people with AKM dacs, and HQplayer + DSDD modulator bypass mode has been doing for years by upsampling to 256Fs using interpolation?

 

[mansr]

So they are basically doing what people with AKM dacs, and HQplayer + DSDD modulator bypass mode has been doing for years by upsampling to 256Fs using interpolation?

Similar but better: the output is multi-level. Two-level (1-bit) DSD has a raft of issues that go away when the number of levels is increased.

 

[Yviena]

Similar but better: the output is multi-level. Two-level (1-bit) DSD has a raft of issues that go away when the number of levels is increased.

Interesting I wonder if the more complex/resource intensive modulators in HQp would offset the issues with 1-bit DSM something like quality instead of quantity?

Looking at the HQP manual it seems that it has some kind of pseudo multi-bit modulator:

AMSDM7 512+fs Special adaptive seventh order “pseudo-multi-bit” modulator optimized for rates above >= 20.48 MHz.

 

[mansr]

Interesting I wonder if the more complex/resource intensive modulators in HQp would offset the issues with 1-bit DSM something like quality instead of quantity?

Looking at the HQP manual it seems that it has some kind of pseudo multi-bit modulator:

AMSDM7 512+fs Special adaptive seventh order “pseudo-multi-bit” modulator optimized for rates above >= 20.48 MHz.

He can call it whatever he wants, the output is still 1-bit.

 

[FrantzM]

Will we be able to hear the differences?
We seem to be(are?) at a point where on the audible, the plateau has been reached. It is a market economy, manufacturers have to come up with new blurbs to sell their ware.

 

[mansr]

Will we be able to hear the differences?

There will be a much louder "ka-ching" sound.

 

[Gradius]

And Akm continues with the bad habit (since AK4493) of not publishing their specs...

Don't believe on specs too much. Over the decades I saw a lot manufactures lies on numbers published there.

 

[andreasmaaan]

Don't believe on specs too much. Over the decades I saw a lot manufactures lies on numbers published there.

Including DAC chip manufacturers? I've seen it from finished DAC manufacturers, but was not aware it also happened at the chip level.

 

[boXem]

Don't believe on specs too much. Over the decades I saw a lot manufactures lies on numbers published there.

Where I am coming from, one single erroneous parameter on a spec can have huge financial consequences. So component specs don't lie. It's not a question of believing, it's a question of having proper work baselines for product development. How do you develop a product if you start questioning each single component parameter?

 

[Gradius]

Including DAC chip manufacturers? I've seen it from finished DAC manufacturers, but was not aware it also happened at the chip level.

At chip level I don't know. That one only someone responsible on some IC could tell us.

 

[Gradius]

Where I am coming from, one single erroneous parameter on a spec can have huge financial consequences. So component specs don't lie. It's not a question of believing, it's a question of having proper work baselines for product development. How do you develop a product if you start questioning each single component parameter?

Check all those measures available here. You will see only the best ones come close to the specs. But they never match.

 

[boXem]

Check all those measures available here. You will see only the best ones come close to the specs. But they never match.

A component and it's implementation are two different animals.
One also needs to distinguish the notions of typical, min, max. Only min and max are commitments, typical is more of an information.