"Chord Electronics FPGA DAC Technology Explained" - What went wrong?

[Geert]

He also said I am an engineer who wants to learn. This part, I did not get.

Simply that I'm after true understanding of how things work. Say someone proofs high res audio makes a substantial audible difference, great, more opportunity to learn and to see how we can make people benefit from it.

 

[IAtaman]

Like Keith said earlier, there's not much reason for people manufacturing expensive audio stuff to get involved. And here we are on ASR.

Well @Dan Clark is a big proponent of Chord products. In one of his posts here you guys have all missed, he claimed, and I quote, "Chord Mojo 2, Hugo 2, and TT2's we use renders highs with an extremely natural timbre that most DACs we've heard don't". I proposed a blind test wager, he kindly declined. He said he guesses (or he and his team, I am not sure as he keeps using "we") that it might be the filter. Maybe he can enlighten us, or even convince his buddy Rob to volunteer to a test?

 

[fpitas]

Well @Dan Clark is a big proponent of Chord products. In one of his posts here you guys have all missed, he claimed, and I quote, "Chord Mojo 2, Hugo 2, and TT2's we use renders highs with an extremely natural timbre that most DACs we've heard don't". I proposed a blind test wager, he kindly declined. He said he guesses (or he and his team, I am not sure as he keeps using "we") that it might be the filter. Maybe he can enlighten us, or even convince his buddy Rob to volunteer to a test?

Well, right. He has nothing to gain by proving his expensive stuff is nothing special.

 

[Mart68]

@Mart68

Boston Audio Society - ABX Testing article

Keith

That's the one!

 

[Purité Audio]

I really like that article, Linn made such a fuss of wristwatches, single speaker demonstrations analogue superiority, all complete b*****ks.
What’s worse people believed it.
Keith

 

[Multicore]

Yes but then there are still issues with being unable to control people gaming the test. And with a null result typical audiophile criticisms such as 'Deaf', 'Equipment not resolving enough', 'Not experienced listeners' and so forth.

Sure. I was thinking of those with good-faith curiosity. I'm not sure there's value in producing test results or doing demos for those who aren't open to learning something.

 

[Mart68]

Well, right. He has nothing to gain by proving his expensive stuff is nothing special.

Yes. Almost everyone making a living in the industry will always couch their comments to the 'Maybe there are mysteries' side of things.

It's just much easier to knock up a product and sell it on the back story than it is to make something that's a genuinely good value proposition and sell it based on that.

 

[Mart68]

I really like that article, Linn made such a fuss of wristwatches, single speaker demonstrations analogue superiority, all complete b*****ks.
What’s worse people believed it.
Keith

People still do.

 

[fpitas]

It's just much easier to knock up a product and sell it on the back story than it is to make something that's a genuinely good value proposition and sell it based on that.

The latter takes expensive and laborious engineering. The former simply a glib tongue and a lack of integrity.

 

[antcollinet]

Well, right. He has nothing to gain by proving his expensive stuff is nothing special.

But if he genuinely believes it is special, he has a lot to gain by proving it does : imagine being able to tell all those heathen unbelievers over at ASR that they don't know as much as they think they do.

The fact these people refuse to try makes me seriously doubt their integrity.

 

[fpitas]

The fact these people refuse to try makes me seriously doubt their integrity.

See, I had very little trouble doubting their integrity to start with

 

[voodooless]

There's a poor track record too

I’d call that an excellent track record

But if he genuinely believes it is special, he has a lot to gain by proving it does : imagine being able to tell all those heathen unbelievers over at ASR that they don't know as much as they think they do.

As long as we’re a minority, they have little to fear. But the fact that we’re making an impression probably has them scared more than we know.

The fact these people refuse to try makes me seriously doubt their integrity.

Well, it’s really simple: assure they are a fraud by default, unless they prove otherwise.

 

[amirm]

Is it that hard? Can't we take an audio file and run it through 1-million tap reconstruction filter to produce a second one, make them both available for download and then compare them by listening on our own gear? Software can blind and randomize the switching between the two files.

To be clear, I'm not interested personally but some people might like to do something like that to see if they can tell a difference.

I provided such files for M-Scalar in the review. No one ran an ABX test showing they can tell them apart. Hardly anyone guessed without it either.

 

[amirm]

This part I don’t get. If a null test is saying 250db difference in audible range, what kind of engineering curiosity would drive you to look for a listening test? Do you think the test is missing something?

There is a reason to do it and that is to find extraneous factors that lead to differences. Let me give an actual example.

Many years ago in another forum the topic of demagnetizing CDs and LPs came up. There were folks that swore by the latter using a specific DeMag device from Japan. A friend and high-end manufacturer volunteered to capture the output of a virgin LP (never played before), first as is and then after the de-mag process. I listen to the two captures and there was clear difference with the latter sounding more open. Objective analysis using an audio editing program showed clear differences as well. Quite surprising for those of us who thought there is nothing to demagnetize in an LP.

Fortunately, I thought of another possibility and I asked my friend to capture a virgin (never played) LP twice without any de-mag process. Well, wouldn't you know it: the same difference was there! In other words, the sound of LP improved on second playback compared to first. That would explain why people would hear a difference. But of course association with the de-mag was faulty.

Now, this kind of thing happening is rare so likely all such tests will fail but it doesn't mean we shouldn't try it.

As to not believing it, yes, we would be incredulous but as long as we can replicate the test, all is well. We could do that and do our own analysis of any audible difference. Or show that the test was wrong.

 

[amirm]

Can you help me understand how taps are relevant to DAC's? I understand how they are relevant for DSP, but not DAC's. I get from this thread that more taps = higher oversampling, but what is the exact calculation, and how much oversampling is considered "adequate", and how much is "overkill"?

(Again, not challenging or arguing, I only want to know the answer!)

A few useful things:

1. DACs like Chord take a multi-bit PCM sample and convert it to 1 bit. To do that you have to keep upping the sample rate until the variation can be described using 1 bit changing. This translates into Megahertz region. This also applies to delta-sigma DAC chips although they usually use more than 1 bit these days.

2. The trick then is how to find the intermediate samples. You can perform math to find those. Or replicate the samples and then filter them. The latter is what is more typically done. If you don't filter, you will have created extraneous high frequency information which you don't want. Here is a demonstration using an image:

Notice that the interpolated (upsampled and filtered) image is the softer version of pixel replicated on. In other words, the process of interpolation is really a low pass filter.

How you do the filtering is up to the implementation. Using standard FIR filters, you pick N samples which becomes the number of taps or the order of the filter.

Note that if you had a straight DAC that could convert all 24 bits to analog as is, you would not need interpolation and the notion of taps. But building such accurate DACs is hard and hence the advent of higher frequency but lower bit depth DACs.

 

[DonR]

I really like that article, Linn made such a fuss of wristwatches, single speaker demonstrations analogue superiority, all complete b*****ks.
What’s worse people believed it.
Keith

I remember a salesman in a high-end store in the late 80s trying the wristwatch argument on me. I was eventually politely asked to leave the store.

 

[voodooless]

Note that if you had a straight DAC that could convert all 24 bits to analog as is, you would not need interpolation and the notion of taps

Well, you’ll still need interpolation. The sample value is only valid for one point in time, not for the whole sample. The only thing you did is shift the burden of the interpolation to the analog part of the DAC: you’ll need to construct an analog low-pass with a very steep filter and deep stop and. That’s going to be hard, and the phase error will be significant.

 

[wwenze]

Considering that scientists routinely test others' claims themselves especially if it is a big claim e.g. the recent superconductor claims which turned out to be either false or practically pointless btw, all it takes is for the claimant to document their experiment findings better (apart from "omg I can hear it") and various 3rd parties can see if that experiment has something not found in thousands of other already-concluded experiments and hence verify the effect of that extra thing. Discoveries take time to come, but only require a relatively short time to verify. And looking at how pretty much the entire part of the audio industry with actual functionality have achieved a lot of things that require way more time, the burden - a very small one in comparison - is on the claimant.

Perhaps a better question would be, if the claimants believe what they claim is real and their experiment methods are correct, why do they invariably believe that other people performing a more rigorous experiment would not be able to detect the difference? Minds too closed?

 

[fpitas]

the burden - a very small one in comparison - is on the claimant

Of course. That's true everywhere in science and industry. When people show up and want us to prove their bizarre preconceptions, I just assume we're being trolled.

 

[wwenze]

Well, you’ll still need interpolation. The sample value is only valid for one point in time, not for the whole sample. The only thing you did is shift the burden of the interpolation to the analog part of the DAC: you’ll need to construct an analog low-pass with a very steep filter and deep stop and. That’s going to be hard, and the phase error will be significant.

Captured output of a TDA1543 running at either 44.1 or 48kHz I forgot