DAC audible differences are finally proven with ABX?

[Leporello]

(this thread's topic name is bad)

Agreed, but we live in attention economy.

 

[voodooless]

There is mention that in the version the author used the relays for switching between A and B sound different, and that the manufacturer fixed this for later versions.

So you can get a perfect score by not at all listening to the DUTs

 

[AndyBell]

So you can get a perfect score by not at all listening to the DUTs

Mathematician Matt Parker once wrote ‘it’s amazing what you can prove if you’re prepared to ignore enough data…’

 

[WillBrink]

You make a good point here (given I understood you correctly) - The DACs are either transparent or you can tell them apart. Right? If you've managed to tell them apart, they are obviously not transparent or not sufficiently transparent.

No one ever said it's never the case you can here the difference. The point was - if you can hear it (and you shouldn't), it's not working properly. Other point being that the difference maybe coming from an output stage, some gain or something and not converting zeros and ones into an analogue signal.

That's the part I always think of. I assume various manufacturers offering a "house sound" of sorts will manipulate the output stage.

 

[RandomEar]

That's the part I always think of. I assume various manufacturers offering a "house sound" of sorts will manipulate the output stage.

There's no proof of that even existing, so... Why assume something like that? The output matters. The measurements capture the output. Why should we care about output stages, capacitors, op amps or whatever else if the resulting performance is essentially identical? It's a repeating discussion even here on ASR which I do not understand. I have seen no logical argument which supports it. If the end result at the output RCA/XLR terminals is identical, why focus on the parts inside the little black box? The output is the only thing which matters for the actual sound, because that is what we listen to. Everything else is only relevant to other qualities like longevity.

 

[killdozzer]

Well, I brought it up so I guess it's on me. But I think you've missed my point. I wanted to say IF you hear it still doesn't have to be down to the converter. You're using the example where, as you say: 'If the end result at the output RCA/XLR terminals is identical'. But if it's identical, then it doesn't fall under what I was trying to point out.

 

[CDMC]

Just my two cents based on a quick read of the article:

1) I wonder if the difference can be attributed to different reconstruction filter settings. It would be interesting to see him compare two identical dacs with different filter settings to see if he can identify differences.
2) It would be great if he could have measured all his units to see if any measurement differences could identify potential sound differences (a big ask as he likely doesn't have the equipment to do all the measurements).
3) I think he did a pretty good job of balancing his biases and conclusions. No hyperbole. We all have them and he was doing this as a solo project with virtually no budget., It is far more rigorous than most stuff we see.

 

[Asher]

I'm a bit skeptical. I'd like to see the results of someone feeding the same DAC into both A and B (without telling the author, of course) to see whether or not he can perceive a difference. Or the same DAC in both inputs, but with the level of one attenuated within the level-matching threshold of the other tests.

 

[Somafunk]

I'm just glad DACs are as good as they are now.
I can thus expend more time, effort and money on recorded and live music.

Yeah, at no point in my life have I cared about my dac “house sound”, I bought my current dac only because it was cheap and had the possibility of displaying vu meters (which I soon tired of)

 

[WillBrink]

There's no proof of that even existing, so... Why assume something like that? The output matters. The measurements capture the output. Why should we care about output stages, capacitors, op amps or whatever else if the resulting performance is essentially identical? It's a repeating discussion even here on ASR which I do not understand. I have seen no logical argument which supports it. If the end result at the output RCA/XLR terminals is identical, why focus on the parts inside the little black box? The output is the only thing which matters for the actual sound, because that is what we listen to. Everything else is only relevant to other qualities like longevity.

The point is, if someone can reliably differentiate between to units/boxes containing a DAC chip, I assume it's not due to the chip but something done at the output stages as we know there's been dead transparent DAC chips for decades now. I have not followed the objective testing of it via Amir and co as that's about as interesting to me as watching paint dry. If say the testing being discussed in the OP is a legit example of audible differences, is that due to the chips used or something else? I assume it's due to the something else part of it in the audio signal at this time. I also don't put it past some manufacturer to create a "house sound" by altering output stage, but I will take your word for it that it has yet to be demonstrated via objective testing.

 

[gzost]

Finished skimming the Dagogo artcile part 3, and the author there could reliably identify differences between interconnect cables. So I continue to assume that there is an issue with the comparator box for at least some test conditions.

 

[RandomEar]

The point is, if someone can reliably differentiate between to units/boxes containing a DAC chip, I assume it's not due to the chip but something done at the output stages as we know there's been dead transparent DAC chips for decades now. I have not followed the objective testing of it via Amir and co as that's about as interesting to me as watching paint dry. If say the testing being discussed in the OP is a legit example of audible differences, is that due to the chips used or something else? I assume it's due to the something else part of it in the audio signal at this time. I also don't put it past some manufacturer to create a "house sound" by altering output stage, but I will take your word for it that it has yet to be demonstrated via objective testing.

If someone could differentiate them, there would have to be a measurable and audibly significant difference. In that case, why attribute that difference to "output stages" or any other specific part, for that matter? It could be a mix of a dozen small differences. Why assume the chips or their settings would not be the cause? Sorry, but I do not follow that reasoning. If a difference exists, it can be in any or multiple part of the chain inside the little black box. There is also no reason why "house sound" should be specifically due to output stages. Manufacturers can simply implement their own reconstruction filters - better DAC chips provide at least one slot for custom filter designs. Put in a super slow roll-off filter and BAM: "Mellow and relaxed house sound"

EDIT: Back to the topic of this thread, though. A couple of very good points have been made, including audible differences between the relay clicks, other tests indicating possible differences between the A and B channels of the line level signal path. In addition, there is the question of different delays I mentioned.

 

[popej]

There were tested 7 DACs in different pairs. And there were obvious difference in each pair (obvious as no doubt in recognition). Apart of DACs in each test, difference in sound can be attributed to ABX switcher and DDC with 2 outputs. The logic suggest, that it was rather one of components, that were present in each test.
Small note: Mutec MC 1.2 doesn't provide 2 identical S/PDIF outputs. There are optical, RCA and BNC outputs. Not sure, which were used.

 

[Redacted]

AFAIK, all DACs require a fair boost to get to line-level or to headphone drive current. It's not impossible that poorly deigned op-amp (etc) circuits may affect sound, but it's not rocket science to get some clean analogue gain....

 

[kemmler3D]

This all strongly suggests that for comparing line-level signal sources the comparator itself produces some audible artefacts between the A and B paths, and that these are sufficient to allow clearly distinguishing the two.

Seems like a good possibility to run down.

I applaud the author for doing the work and at the very least giving the community something concrete to discuss and investigate.

As others have mentioned the results seem too good to be true, so as ever the next step is to attempt to replicate the results.

 

[WillBrink]

If someone could differentiate them, there would have to be a measurable and audibly significant difference. In that case, why attribute that difference to "output stages" or any other specific part, for that matter? It could be a mix of a dozen small differences. Why assume the chips or their settings would not be the cause? Sorry, but I do not follow that reasoning. If a difference exists, it can be in any or multiple part of the chain inside the little black box. There is also no reason why "house sound" should be specifically due to output stages. Manufacturers can simply implement their own reconstruction filters - better DAC chips provide at least one slot for custom filter designs. Put in a super slow roll-off filter and BAM: "Mellow and relaxed house sound"

Because according the powers that be, chips have been audibly transparent for decades now. As you point out, there are various places along the signal they could play with that to offer something different, yes, I assumed at the output stage. First, we'd have to establish some can actually differentiate them under controlled conditions (and that's being debated here now) and then we'd have to figure out why/how.

 

[amirm]

The author certainly did more than we expect from any non-industry person to perform. I have looked at buying the Van Alstine comparator and even contacted the company a few years ago. Alas, they never responded to me.

I think the next good step would be to measure the comparator. I wonder if the author would be willing to send it. Alternatively Erin had one.

As noted, there should have been a control with two identical sources to rule out comparator having issues. That would have required having another helper so logistics would not have been as easy.

 

[pma]

I think the next good step would be to measure the comparator.

Yes. And it also would be good to have test files recorded by a good ADC like Cosmos and analyze it. Otherwise, it is difficult to make a conclusion based on descriptions only.

 

[JaMmy]

First impression: no mention of any settings on those DACs. I mean those are complex processors and not one-way headphone amps. Factory defaults out of the box? Customized boxes? Which filters have been used?

Second thought: people never make stuff like that up. That's a thing of the art world. No need for any proof it all did actually happen.

Unfortunately science history is littered with examples of sometimes quite prestigious people who made up or finagled results. But science overcomes this through replication. A one off study has little weight until another lab follows the same protocols and gets the same or similar results. Usually faked results are too perfect. Error is almost impossible to eradicate in any experimental set up.

 

[Sokel]

Based on statistics alone, I am extremely sceptical of these results. I have done a couple of "easier" ABX tests, for example between a signal where there was a small gradual drop of up to 0.5 dB in the response above 6 kHz. Even with these comparatively major differences, I did not score 10/10 in each test, repeatedly again and again. It is hard to distinguish such tiny differences and your ears occasionally will fail you. So 54 out of 56: I am extremely, extremely sceptical of that result, based on the numbers alone.

That being said, there's two major things not adressed in the article:

• Choice of reconstruction filters: The X Sabre Pro for example uses a slow roll-off filter by default, most other DACs in the test use a linear fast roll-off by default. Yes, the material was supposedly 96 kHz or upsampled to that. But I would nevertheless expect such essential settings to be listed and - if he was looking for a fair comparison - I would expect the author to select similar or if possible identical filters on all DACs.
• Delay: This is a big one. According to what I have seen in the article, the DACs were not time-aligned (-> directly fed by the same S/PDIF source). Due to the likely different filters in play, the effective delay between the DAC outputs will vary between around 150 µs up to roughly 900 µs, maybe more. Since humans can distinguish delays larger than around 7 to 18 µs depending on which trigger is investigated, this could potentially allow anybody to ABX these DACs under the given conditions. Simply because there's a time difference of multiple hundred microseconds between many pairs of them, or maybe all pairs - we can't know because as far as I could see, the reconstruction filters were not listed for any comparison and delays were not measured.

Soooo... This delay thing needs to be clarified. Otherwise, this looks like a false positive to me.

About filters, in order to eliminate their effect as he says, all material was 24/96 (the one that wasn't was upsampled with XLD.

In general, it seems that he made his homework about the upcoming debates, gear is above suspicion, (Cosmos, etc)
What I haven't understand though, is the way of reporting the results.

Do they come straight from the machine or we have to take his word from them?
A traditional ABX test with foobar for example will generate a checksum* anyone can check, as we have done here many times.
So, where to check?

*Edit: like this one, see the signature: