Serious Question: How can DAC's have a SOUND SIGNATURE if they measure as transparent? Are that many confused?

[Vinman]

I would love for you to do a blind listening test with this DAC and maybe a good Topping or Schiit DAC, and maybe another blind listening test with a "very mediocre" DAC like on of the Audio-GD products.

It is my contention - after some half-baked blind A/B tests of my own- that people can't readily hear the differences between DACs. I did some A/B listening tests myself and also with a professional musician using a Topping D50s DAC and the very same Audio-gd R2R11 DAC that I sent to Amir to test ( https://www.audiosciencereview.com/...-measurements-of-audio-gd-r2r11-dac-amp.5779/ ) and we could not tell the difference at a level greater than chance using Red book CDs with Quad ESL-57 speakers or Sennheiser HD-800 'phones. The Audio-gd R2R11 has pretty abysmal measurements so you'd think the difference would be plainly audible. Using music, it seemed it was not.

We did not try pure tones, which may have shown audible differences - but DACs are used to listen to music, not tones, so we used music.

My blind testing between the on-board DACs in my preamp, NodeX, and a Weiss DAC 204 yielded 100% correct identification of the Weiss. The two built in DACs yielded about a 70% correct ID - but the differences became easier to ID when I learned that the three chips involved (ESS 9018, AKM 4499, ESS 9028) showed the widest sound differences with big classical pieces. tried this with four people and then had someone test me. You can state that DACs don’t make a difference, but for those who have heard differences, we know better. The winner in my system was the costly Weiss using the oldest chip design (the 9018).

 

[solderdude]

care to describe the comparison and the controls that were used as well as the statistics.

 

[Blumlein 88]

care to describe the comparison and the controls that were used as well as the statistics.

Well of course not.

 

[solderdude]

The 3.75V output voltage on RCA (default) and double that on XL,R of course, probably didn't help at all picking the Weiss as the 'best sounding one'.
Nah ... it must have been the 9018 DAC chip that did it as 5 people could tell the Weiss apart.

 

[Geert]

Could also be a 'broken by design' case, if it resembles Weiss-dac205-dac-review.

"Your guess is as good as mine as to why we have such high levels of intermodulation distortion."

"The worst jitter I have ever measured in a performant DAC."

But still "a winner".

 

[abm0]

Very late to the party, but I think I've recently found my answer to this question: a lot of DACs and DAC-amps that are reviewed and discussed online based on measurements do really have a "sound signature" and it really does reflect in the measurements, so they do not actually measure as transparent, but #1. people don't pay attention to the FR measurement (or ask for it when it's missing, or zoom in on the 32-tone test to see if the spikes differ from eachother even by 1 pixel, which may be enough at the scale the 32-tone test is usually presented) and/or #2. people don't know just how tiny of a spectral tilt can actually be heard by humans (see audibility thresholds topic: the limit is 0.1 dB per octave).

 

[solderdude]

0.1dB roll-off at 20kHz for instance isn't directly audible but measurable.
That 0.1dB per octave is a tilt of -1dB at 20kHz and DACs usually don't do that. They only roll-off near 1/2 fs and is only 'detectable' on direct switching and that roll-off cannot be called a 'sound signature'.

 

[abm0]

0.1dB roll-off at 20kHz for instance isn't directly audible but measurable.
That 0.1dB is only 'detectable' on direct switching and a lot of training as that 0.1dB in reality is closer to 0.2dB.
It also cannot be called a 'sound signature'.

It's not always at 20 kHz. Look at the recent and highly gushed over FiiO KA17: it has a V-shaped cutout at 6-9kHz somewhere, rises back up, then rolls off again toward 20kHz, both times it's 0.4 dB below neutral, clearly audible, clearly a sound signature. Or look back at the favorably reviewed Hidizs S8: Amir didn't show the FR plot but you can almost see the information on the 32-tone plot: treble is veiled from 6 to 10k, then rises again. Compare with my Hiby FC3 response: flat flat flat all the way to 20 kHz, real neutrality.

Not all DACs sound the same. Manufacturers are creating measurable "signatures" to draw customers to their side and keep selling new units every year. Competition and continuous sales wouldn't work nearly as well if it was all placebo.

 

[solderdude]

Why would any DAC 'dip' at 6-9kHz ?
Only when some EQ would be used (which it has) you could do that but that is EQ and not the frequency response of the DAC.

KA17_parameters

KA17_parameters

www.fiio.com

The vertical scale on the plots are 1dB/div. Where is the 'sound signature' ?

S8 ... where is the dip ? Where is the 'veil' and where can you 'almost' see that ?

ALL DACs measure differently. NONE are the same that does not mean they also sound different.

 

[abm0]

Why would any DAC 'dip' at 6-9kHz ?

For example to push some sounds back and create the illusion of "soundstage depth". Like I told you: to differentiate vs. other products and give customers a reason to keep buying buying buying.

S8 ... where is the dip ? Where is the 'veil' and where can you 'almost' see that ?

Zoom in.

ALL DACs measure differently. NONE are the same that does not mean they also sound different.

What is this, ignoratio elechi? Are you pretending not to understand the exact measurement I told you tends to be different and the amount by which it tends to be different and that research says that amount is audible?

 

[solderdude]

WHERE is the audible threshold reached in the posted FR plots of the KA17 ?
Please zoom in in show the audible 'dip' in the S8 that is considered a sound signature ?
Also the consensus here is not ALL DACs sound the same but when they don't is is very measurable and breaches audible thresholds.

 

[aagstn]

Not all DACs are the same. Manufacturers are creating measurable "signatures" to draw customers to their side and keep selling new units every year. Competition and continuous sales wouldn't work nearly as well if it was all placebo.

It is mostly placebo and has been common in audio for all kinds of cables, dacs, amps and so on for decades. Some manufacturers like PSAudio put audible noise in their DACs to make them sound different but even this is subtle. If it was so true and obvious that dacs, cables and so on sounded different the influencers would be lining up to do a scientific blind test to prove it and shut the doubters up. Instead it is crickets because there is no way they kill the golden goose by testing this. Measurements aren't perfect. There are flaws in dacs that sometimes don't get discovered with just measurements. Audio leaking out RCA jacks, pops and distortion that shouldn't be there when changing bitrates and other faults. Subjective reviews are flawed in almost every way. A YouTube guy telling me something has more color, depth, or other meaningless words when I don't have their listening room or their ears is useless. At least a measurement of noise and distortion is a starting point that will lead me in the right direction to find a good component and not get ripped off with overpriced junk.

 

[abm0]

WHERE is the audible threshold reached in the posted FR plots of the KA17 ?

Are you seriously posting a manufacturer's graphs as an argument in the ASR forum? You don't even know how much smoothing they applied when they made that graph. You need independent measurements. I heard it as veiled, then I measured it through my soundcard and I found why I was hearing it as veiled: https://forum.hifiguides.com/t/fiio-ka-17-opinions-and-usage/43721/5

And no, I will not teach you how to zoom in, you're perfectly capable of that at your level of experience, I don't know what game you're trying to play here but I'm not playing. You're just trying to hang on to #2 in my original answer: you don't want to acknowledge hard data that would force you to believe there are audible differences already shown with measurements, in the response of highly popular modern DACs.

 

[MaxwellsEq]

Or look back at the favorably reviewed Hidizs S8: Amir didn't show the FR plot but you can almost see the information on the 32-tone plot: treble is veiled from 6 to 10k, then rises again. Compare with my Hiby FC3 response: flat flat flat all the way to 20 kHz, real neutrality.

I've zoomed in on the picture. Are you identifying that the 2 multitone frequency peaks at between 5 and 10 are a pixel or so below the line?

 

[abm0]

I've zoomed in on the picture. Are you identifying that the 2 multitone frequency peaks at between 5 and 10 are a pixel or so below the line?

I'm seeing 3, but yes.

 

[solderdude]

Are you seriously posting a manufacturer's graphs as an argument in the ASR forum? You don't even know how much smoothing they applied when they made that graph. You need independent measurements. I heard it as veiled, then I measured it through my soundcard and I found why I was hearing it as veiled: https://forum.hifiguides.com/t/fiio-ka-17-opinions-and-usage/43721/5

And no, I will not teach you how to zoom in, you're perfectly capable of that at your level of experience, I don't know what game you're trying to play here but I'm not playing. You're just trying to hang on to #2 in my original answer: you don't want to acknowledge hard data that would force you to believe there are audible differences already shown with measurements, in the response of highly popular modern DACs.

Who says your measurement is accurate ?
How linear is your ADC ?
Why would an AP measurement be unreliable ?
Why would a 0.4dB dip be there anyway ?
Ever seen that in any DAC ?
How much smoothing would need to be applied to an AP measurement for such a broad dip to be 'smoothed out' ?
Have you measured it in a different way than using Pauls software ?
Is it possible something on your end is wrong ?
Do you have independent measurements other than your attempt that show the same ?
What was the load you used ?

 

[MaxwellsEq]

I'm seeing 3, but yes.

Interesting. I've never seen multitone used in this way. It's not really a test of frequency response, more a noise and distortion test.

 

[GXAlan]

For the benefit of the thread, this is what @abm0 measured. He has a control measurement of a different DAC showing a more normal response.

Intentional or not, that’s a sound signature for sure that might be difficult to identify all the time in blind testing but could likely be identified some of the time.

 

[abm0]

Interesting. I've never seen multitone used in this way. It's not really a test of frequency response, more a noise and distortion test.

pkane's multitone tool does it that way, and it makes sense to me as more similar to music than a tone sweep. And if it makes sense to plot an FR based on a multitone stimulus with hundreds of tones, I'm thinking we should be able to spot some things even in the spike heights if we look close enough, and even with a lower number of tones if the affected band is wide enough.

 

[abm0]

Other one: Questyle M15, highly praised, glowing reviews across the board. Does it have a signature? Yes. Very close to inaudible, but not close enough, depending on listener health.

https://cdn.l7audiolab.com/wp-content/uploads/2022/07/Relative-Level-1.jpg