• Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

How can DAC's have a SOUND SIGNATURE if they measure as transparent?

Status
Not open for further replies.
What differences can be heard but have been routinely ignored by measurements-reviewers?
Just curious- can you cite an example of a difference being actually heard that was not reflected in a standard measurement suite? Just one example?

I haven't dug into what you're claiming for measurements, so can't comment on their validity, apologies. But if you can come up with an actual audibility cite, I'd certainly be delighted to do so and dig in deeper with better tools.
 
Just curious- can you cite an example of a difference being actually heard that was not reflected in a standard measurement suite? Just one example?
Still an open question vis-a-vis proving it with a listening test. I want to test it with ABX for the FiiO KA17 vs. Hiby FC3 once I find a good section of a good song to bring out the difference as much as possible. To be clear - and as I've said multiple times already - I think some audible problems are being reflected in standard measurements, just not zoomed in on, nor discussed in the graph review. It's a different question whether we could use other ways of measuring that could more easily highlight potentially audible problems than what is usually done. My first issue is with the current interpretation of what is "audibly transparent" and what levels of defect are worth zooming in on.

As for the 1 dB scoop in the mids for the Retro Nano, I don't think it needs a blind listening test, I'm just going to assume it's audible and that it's exactly what I heard when the Nano's response seemed muffled and "camel-back" shaped in the bass and mids, after the first 3 songs I played through it. I expect any usual measurements run would show this FR shape as well, if performed using Bluetooth (FR shape is different when running it wired like a dongle, hell it's even different between LDAC and aptX).
 
I think some audible problems are being reflected in standard measurements
Which "audible problems" do you mean, specifically? And how was the audibility determined?
 
I've found it quite weird to see as little focus on frequency response as I've seen in measurements-reviews

Amir posts the frequency response in most of his DAC reviews I believe. When I look at the frequency response of some of these DAC's, all there is to see is a 0,1dB droop towards high frequencies on some occasions (making abstraction of the effect of the reconstruction filter). Nothing that explains the night and day differences people speak of.
 
Which "audible problems" do you mean, specifically? And how was the audibility determined?
FR defects touching or exceeding 0.2 dB away from flat (though for now I'm also wondering about 0.1 dB - how widespread the ability to hear that might be). Not sure how it was determined but that's the number quoted in the Audibility Thresholds thread, sadly without citing any specific research papers.

all there is to see is a 0,1dB droop towards high frequencies on some occasions (making abstraction of the effect of the reconstruction filter)
Wait, why would you ignore the reconstruction filter? Is that not part of the DAC?

Nothing that explains the night and day differences people speak of.
Of course "day and night" is ludicrous, but I'm coming at this from the opposite angle: nothing explains the "audibly transparent" claims that are so often made in this forum, especially in posts advising newcomers what to buy or not buy. That's an extreme claim that requires extremely high quality evidence, which I think has not been provided.
 
FR defects touching or exceeding 0.2 dB away from flat (though for now I'm also wondering about 0.1 dB - how widespread the ability to hear that might be). Not sure how it was determined but that's the number quoted in the Audibility Thresholds thread, sadly without citing any specific research papers.
So have you been able to determine any audibility issues? I'm asking all of this because I just haven't seen frequency response errors in any DAC I've tested and suspect there might be measurement issues rather than audibility. But actual experiments answer objections. :cool:
 
Wait, why would you ignore the reconstruction filter? Is that not part of the DAC?

Because you gave an example of a DAC with a scooped midrange to support your argument. No one questions differences in reconstruction filters and each review has a graph that zoomes in on them, so you can't argue there's to little focus on that aspect.
 
I would assume that if a DAC has frequency response errors in anything other than the filter area, they would be shown in the review. And if not they would still show up in the multitone graph.
 
FR defects touching or exceeding 0.2 dB away from flat (though for now I'm also wondering about 0.1 dB - how widespread the ability to hear that might be). Not sure how it was determined but that's the number quoted in the Audibility Thresholds thread, sadly without citing any specific research papers.

Those DeltaWave results are highly suspect. An RMS null of -19dBFS is way below anything I've seen from hundreds of such measurements of different devices. The worst are about -34dB, and most are below -40dB. I suggest you re-verify your loopback chain and make sure you're testing and comparing the right things with correct settings.
 
It is not that easy to get good or even perfect results from DAC/ADC loops. It is difficult for unqualified persons and it is easy to make mistakes. Starting with SW drivers, over signal ground loops and two different devices on 2 USB ports. Proper control is needed. This is not a plug and play job. And simple soundcards are inadequate for the job. Better not do it, then.
 
FR defects touching or exceeding 0.2 dB away from flat (though for now I'm also wondering about 0.1 dB - how widespread the ability to hear that might be). Not sure how it was determined but that's the number quoted in the Audibility Thresholds thread, sadly without citing any specific research papers.
You can test yourself here. I can tell you that 0.1dB is extremely difficult and noticing a 0.1dB difference in the downward direction is even more difficult. I suggest you start at a level higher than 0.1dB so that you can note that direction change difficulty.
 
It is not that easy to get good or even perfect results from DAC/ADC loops. It is difficult for unqualified persons and it is easy to make mistakes. Starting with SW drivers, over signal ground loops and two different devices on 2 USB ports. Proper control is needed. This is not a plug and play job. And simple soundcards are inadequate for the job. Better not do it, then.

Agreed. It's important not make any conclusions from flawed measurements that produce outlier results.
 
Because you gave an example of a DAC with a scooped midrange to support your argument. No one questions differences in reconstruction filters
That is absolutely false. Multiple people in this forum routinely give newcomers purchase advice based on this unproven notion that "90% of DAC/amps today are audibly transparent", meaning they are talking about real-world performance of complete devices (otherwise it wouldn't be purchase advice), meaning they are systematically ignoring reconstruction filter problems along with everything else.
 
Those DeltaWave results are highly suspect. An RMS null of -19dBFS is way below anything I've seen
Have you tested Bluetooth DAC/amps? Maybe it's normal for those. Also, remember this is between device and music file, not between 2 devices. The fit for these -19 dB ones showed "Very Good" x1 and "Excellent" x2. Don't think there's much I could've done better.

My wired DAC/amp (dongle) tests via the same line-in show around -75 dB TD+N in the static 1 kHz test, and -33 and -44 dB with DeltaWave for the song-vs-file test with the KA17 as you just saw up-thread. I don't think I have too horrible of a setup, it's just way worse through Bluetooth.
 
Have you tested Bluetooth DAC/amps? Maybe it's normal for those. Also, remember this is between device and music file, not between 2 devices. The fit for these -19 dB ones showed "Very Good" x1 and "Excellent" x2. Don't think there's much I could've done better.

My wired DAC/amp (dongle) tests via the same line-in show around -75 dB TD+N in the static 1 kHz test, and -33 and -44 dB with DeltaWave for the song-vs-file test with the KA17 as you just saw up-thread. I don't think I have too horrible of a setup, it's just way worse through Bluetooth.

Bluetooth? No, haven't tested these. But then you're testing Bluetooth codecs, and bluetooth implementation, not just the DAC.

And yes, nearly all loopback measurements I do are between original music file and the recording.

The "Excellent Fit" display refers to drift calculation and not to the loopback result. This is important when two clocks are not synchronized. "Excellent" means the clock drift calculation produced a good fit for observed values. Random timing errors or non-linear clock drift would cause a poor fit.
 
That is absolutely false. Multiple people in this forum routinely give newcomers purchase advice based on this unproven notion that "90% of DAC/amps today are audibly transparent",

Then some are a bit sloppy, or maybe just tired repeating that a DAC should measure transparent and should not be broken by design. A DAC that offers no choice of a capable reconstruction filter is broken, and Amir points that out in his reviews. And your ears need to be fine to notice what’s happening with reconstruction filters (unless these filters are really bad).
 
Bluetooth? No, haven't tested these. But then you're testing Bluetooth codecs, and bluetooth implementation, not just the DAC.
Of course I am, see also my recent replies for Geert. This is about real products in the real world, who cares about "just the DAC"? That would be a purely academic discussion, especially in recent years with manufacturers having started building these more complex "DAC" chips that are really integrated DACs and headphone amps (DAC-amps-on-chip? DAOCs? DAC-amp-systems-on-chip? DASoCs? :) ). I think some people have not woken up to this little complication, which makes it so you can no longer support the same statements you could before about "the DAC chip", since today "the DAC chip" also contains a headphone amp section.

I wanted to see if I could clarify the recording setup issue, so I put the same song through the HiBy FC3:
TIPTS_FC3vsMP3_DSpectrogram.png


Even though it's a much cleaner DAC/amp, I have to admit there may be larger distortions due to the recording gear than I thought, since I'm seeing those weird sub-bass lines again, going through the whole song. I should probably always compare device vs. device so that those errors are subtracted out; device vs. file seems to be messed up too much by the motherboard line-in's defects. :confused:

Also, on deeper thought, the shades going up across the whole spectrum when the song goes quieter are probably just a natural consequence of calculating a delta vs. quieter original content: even if the errors are staying the same in absolute value, they're now higher dB vs. the barely-there content. So those probably mean nothing as well. *sigh*
 
Last edited:
This is about real products in the real world, who cares about "just the DAC"?

Perhaps you should read the title and the OP of this thread?

Even though it's a much cleaner DAC/amp, I have to admit there may be larger distortions due to the recording gear than I thought,

Sure. Which is why it's important to calibrate your measurement equipment to know that you're not skewing the result.

Also, on deeper thought, the shades going up across the whole spectrum when the song goes quieter are probably just a natural consequence of calculating a delta vs. quieter original content
Remember that on a quiet passage, noise or added distortion will be more pronounced/easier to hear, so you may want to know the source and, if significant, try to eliminate it.

Despite all the low frequency errors, HiBy FC3 shows an RMS null of -37dB, or nearly twice better than what you were measuring with your bluetooth DAC.

Again, i suggest that you don't post any (public) conclusions/claims before calibrating your equipment, and validating the results.
 
Pretty sure I can hear 8 bits vs 22 bits....but I'm
What would be the mechanism responsible for that, is there objective proof for that ?

Noise in DACs is below audible levels and when one does hear noise it is usually in the recording. Noise in recordings made with microphones that are not gated will be magnitudes higher than the self noise of DAC circuits.

There is a hearing threshold. Noise can be shaped differently (think white/pink/brown noise) which changes audibility and 'annoyance factor'.
It is quite possible to hear 5dB noise differences in quiet passages especially when the spectrum differs.

I guess that's why studios use 8-bit and not 24-bit .....no, wait :rolleyes:

It also goes against the Nyquist Theorem, but ok.
 
24 bit is just 3x 8 bit :) and has absolutely nothing to do with noise in DACs nor anything I wrote. That is about bit depth and quantization noise.

So ... I will ask again... what would be the 'mechanism' for a DAC reducing performance over time and where is the evidence for this phenomenon ?
Where does '8-bit' come into play here ?

Also you made a remark about 5dB difference in noise so I my answer is about noise and noise spectra and how 5dB difference could well be audible.
I did not say anything about 8bit vs 24bit nor 16 bit vs 24bit nor about quantization noise nor about DS noise shaping nor about dither nor shaping of dither.
It was about noise levels and audibility thresholds and noise spectra, about microphone recordings and noise that comes with it.
 
Last edited:
Status
Not open for further replies.
Back
Top Bottom