GoldenSounds passes apparently ABX test for DACs (NOT Really)

[Ken Tajalli]

Ahem.

you're late to the party.
It has already been explained.
But Ahem ... is cute.

 

[amirm]

If anything good comes out of this, it'll be Chinese brands seeing it as a new challenge and optimising their filters, hopefully after addressing the hundred other more important things people have been asking for like EQ and DSP.

Those DACs use the filters in the DAC chip itself. Doing better requires disabling that and using external filtering if that is allowed. That would add expense. Before that happens, it needs to be shown that these sharper filters make things sound better. ABX testing only shows differentiation. Not better fidelity as the person taking the test doesn't know which one is which.

 

[Ken Tajalli]

What do you mean by "believe"? Did you do the Foobar ABX?

No, no Foobar, no need.
Believe! because it may well turn out that say sample A and C were one and the same.
I don't know ....

 

[MacClintock]

No, no Foobar, no need.
Believe! because it may well turn out that say sample A and C were one and the same.
I don't know ....

Spoiler: My choice

Spoiler is my choice, let's keep it hidden till we know the results.

haha, saying that one "believes" to hear differences and "would" pass the test but not doing it is like saying how much you would make everything better if you were the politician with power. It has now value as it is just talk. So your choice is "A - C - B but C and B being close, A clear winner". Let´s see.

 

[oleg87]

No, no Foobar, no need.
Believe! because it may well turn out that say sample A and C were one and the same.
I don't know ....

Well, if you feel there is "no need" (I have not tried it myself, will give it a go) you may find it instructive trying to repeat the same feat in a blind ABX.
Since I have no idea which DACs were used I also don't know if we should expect them to be audibly transparent.

 

[Ken Tajalli]

haha, saying that one "believes" to hear differences and "would" pass the test but not doing it is like saying how much you would make everything better if you were the politician with power. It has now value as it is just talk.

What are you on about? pass the test? what test?
The link I supplied is for a different challenge, run by somebody else.

 

[virtua]

Those DACs use the filters in the DAC chip itself. Doing better requires disabling that and using external filtering if that is allowed. That would add expense. Before that happens, it needs to be shown that these sharper filters make things sound better. ABX testing only shows differentiation. Not better fidelity as the person taking the test doesn't know which one is which.

I know it is largely a technical pursuit. But wouldn't a filter that perfectly cuts everything off as sharp as possible right before nyquist be the best filter implementation? I don't really consider it a 'subjectively sounds better' pursuit - the overwhelming majority of people won't even be able to hear any difference in a regular listening situation. However, if we look at a DAC going from 110 to 120+ SINAD - nobody will hear any difference, yet we rate those DACs as better because we understand objectively that their implementation is better. If ABX testing shows differentiation (or even if it didn't), and we understand the sharper filter to be a better implementation, why shouldn't manufacturers try to improve their filters to be closer to the ideal state, as they have with DACs and SINAD?

I understand it is cost prohibitive, and the costs will be passed onto consumers - but there has been a market for high end DACs with top tier technical performance for quite some time. Even if it's only theoretically better and not audibly better (in the case of transparency), it hasn't stopped people from purchasing in the past.

 

[Ken Tajalli]

Well, if you feel there is "no need" (I have not tried it myself, will give it a go) you may find it instructive trying to repeat the same feat in a blind ABX.
Since I have no idea which DACs were used I also don't know if we should expect them to be audibly transparent.

This is just a preference test. I think the only thing that can be proven may be that a large portion preferred a $100 DAC, to a £5K DAC!
Or possibly, most couldn't decide.
Transparency does not come into it.
I am going to ask my daughter to change the names of samples without telling me, and tomorrow I take the test again, to see if I pick the same ones.

Just found that there is a thread about this already:

Archimago has a new DAC listening test

I see Archimago has a new listening test using 3 mystery dac’s, the difference in cost between them is 100x so it’ll be interesting to see if if there is much, if any difference when I listen. (I’m betting I can’t hear any difference at all). Perhaps this fact might encourage you to listen for...

www.audiosciencereview.com

 

[MacClintock]

I know it is largely a technical pursuit. But wouldn't a filter that perfectly cuts everything off as sharp as possible right before nyquist be the best filter implementation? I don't really consider it a 'subjectively sounds better' pursuit - the overwhelming majority of people won't even be able to hear any difference in a regular listening situation. However, if we look at a DAC going from 110 to 120+ SINAD - nobody will hear any difference, yet we rate those DACs as better because we understand objectively that their implementation is better. If ABX testing shows differentiation (or even if it didn't), and we understand the sharper filter to be a better implementation, why shouldn't manufacturers try to improve their filters to be closer to the ideal state, as they have with DACs and SINAD?

I understand it is cost prohibitive, and the costs will be passed onto consumers - but there has been a market for high end DACs with top tier technical performance for quite some time. Even if it's only theoretically better and not audibly better (in the case of transparency), it hasn't stopped people from purchasing in the past.

Well, first of all, these extremely sharp filters of the brickwall type are difficult and expensive to implement, one reason why Chord DACs are expensive. Secondly, they introduce ringing and pre-echo. This might not be an audible issue, but is just a mathematical fact and could be considered as a disadvantage.

 

[Sokel]

Well, first of all, these extremely sharp filters of the brickwall type are difficult and expensive to implement, one reason why Chord DACs are expensive. Secondly, they introduce ringing and pre-echo. This might not be an audible issue, but is just a mathematical fact and could be considered as a disadvantage.

There is some middle ground here,like the ancient Bricasti Amir measured and found it really nice.
Still at the higher price range but not as exotic as the Chord ones.
(text over the chart is by Amir)

 

[oleg87]

I know it is largely a technical pursuit. But wouldn't a filter that perfectly cuts everything off as sharp as possible right before nyquist be the best filter implementation? I don't really consider it a 'subjectively sounds better' pursuit - the overwhelming majority of people won't even be able to hear any difference in a regular listening situation. However, if we look at a DAC going from 110 to 120+ SINAD - nobody will hear any difference, yet we rate those DACs as better because we understand objectively that their implementation is better. If ABX testing shows differentiation (or even if it didn't), and we understand the sharper filter to be a better implementation, why shouldn't manufacturers try to improve their filters to be closer to the ideal state, as they have with DACs and SINAD?

I understand it is cost prohibitive, and the costs will be passed onto consumers - but there has been a market for high end DACs with top tier technical performance for quite some time. Even if it's only theoretically better and not audibly better (in the case of transparency), it hasn't stopped people from purchasing in the past.

Engineering is a world of trade-offs. The Chord DACs for example have a very appreciable amount of latency, which makes them unappealing for uses like gaming.

 

[Ken Tajalli]

Engineering is a world of trade-offs. The Chord DACs for example have a very appreciable amount of latency, which makes them unappealing for uses like gaming.

Don't need a Chord DAC. Cameron used software on his computer to make a 2 million tap filter. That's one million more than mScaler.
Not expensive either (these days).
But should the question be, regardless of mathematical superiority, do they make a difference?
@pkane has shown that in digital domain, a 2 million tap filter is the same as a standard filter, save for a slight bandwidth extension beyond 19kHz.

 

[Blockader]

Well, first of all, these extremely sharp filters of the brickwall type are difficult and expensive to implement, one reason why Chord DACs are expensive. Secondly, they introduce ringing and pre-echo. This might not be an audible issue, but is just a mathematical fact and could be considered as a disadvantage.

They are not really hard to implement, they just need high computation power. Chord dacs don't even have high computation power compared to 1500$ RTX 4090. FPGA's make the implementation easier and simpler.(no memory wall = less power consumption and less hardware) Chord creates the impression that implementing these filters is extremely challenging as a justification for their high prices.

HQPlayer costs 300$ and it offers sharper filters with more taps and thousands of times better attenuation than Chord filters. I forgot the program's name but there's also a DSP software that can apply a sinc filter with 1 billion taps. That's something only CPU's and especially GPU's can handle. One cheap software + your computers GPU is thousands times better than Chord filters and easily 10 times cheaper too. One more note: That doesn't mean any of this is audible.

 

[Sokel]

They are not really hard to implement, they just need high computation power. Chord dacs don't even have high computation power compared to 1500$ RTX 4090. FPGA's make the implementation easier and simpler.(no memory wall = less power consumption and less hardware) Chord creates the impression that implementing these filters is extremely challenging as a justification for their high prices.

HQPlayer costs 300$ and it offers sharper filters with more taps and with thousands of times better attenuation than Chord filters. I forgot the program's name but there's also a DSP software that can apply a sinc filter with 1 billion taps. That's something only CPU's and especially GPU's can handle. One cheap software + your computers GPU is thousands times better than Chord filters and easily 10 times cheaper too. One more note: That doesn't mean any of this is audible.

One can justify Chord as portable,that was their big hit after all.
But yes,for rack applications it makes no sense.

 

[Ken Tajalli]

One can justify Chord as portable,that was their big hit after all.
But yes,for rack applications it makes no sense.

TT2 and Dave are surely desktop.
Mojo2 and Hugo2 are portable, and very very good. I got one.
But the filter inside, had nothing to do as to why I bought one!
The over all performance of a portable device is not dictated by just one criteria, the filter.

 

[Sokel]

TT2 and Dave are surely desktop.
Mojo2 and Hugo2 are portable, and very very good. I got one.
But the filter inside, had nothing to do as to why I bought one!
The over all performance of a portable device is not dictated by just one criteria, the filter.

Agreed,the filter fuzz was strangely pointed out after the portables got their success,probably by the people trying to justify their performance.
And of course the marketing team could not unsee this as they would do with anything that could differentiate them from the competition.

 

[Ken Tajalli]

Agreed,the filter fuzz was strangely pointed out after the portables got their success,probably by the people trying to justify their performance.
And of course the marketing team could not unsee this as they would do with anything that could differentiate them from the competition.

Chord DACs are different from ground up.
Filter is just a part of its inner working.

 

[amirm]

But wouldn't a filter that perfectly cuts everything off as sharp as possible right before nyquist be the best filter implementation?

It would be. I used to note this on every review but got tired of it and don't mention it anymore. I call it "typical lazy filter."

 

[MacClintock]

They are not really hard to implement, they just need high computation power. Chord dacs don't even have high computation power compared to 1500$ RTX 4090. FPGA's make the implementation easier and simpler.(no memory wall = less power consumption and less hardware) Chord creates the impression that implementing these filters is extremely challenging as a justification for their high prices.

HQPlayer costs 300$ and it offers sharper filters with more taps and thousands of times better attenuation than Chord filters. I forgot the program's name but there's also a DSP software that can apply a sinc filter with 1 billion taps. That's something only CPU's and especially GPU's can handle. One cheap software + your computers GPU is thousands times better than Chord filters and easily 10 times cheaper too. One more note: That doesn't mean any of this is audible.

Yes, in software these filters are not a big problem, but in a stand-alone DAC it is different and that's why Chord uses FPGA's. And not everybody likes to always have a laptop or Desktop around. I am not saying by any means that Chord DACs are worth it, not at all. The difference is only barely audible for a tiny fraction of people with special material.

 

[JSmith]

GS hate

No it's not about hate, like or dislike. It's about respect really... one doesn't gain respect by disrespecting others and expecting to be shown respect in return or taken seriously afterwards.

this thread is not about getting personal, it is about the results and conditions of the ABX test.

Absolutely and what was pointed out were facts, not personal attacks. Plus it seems the "test" has been well and truly covered over these 28 pages of kerfuffle, so maybe there isn't much else to discuss here?


JSmith