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

[signalpath]

But then how to convey the fact that sound quality of properly designed DAC’s has been above human capabilities for decades.

That's not our experience. The very best DAC we've tested (DO-300) exhibits horrific waveform distortions 20dB above the commonly accepted threshold of headphone listening. With certain tests (complex-generated reverberant tails at pronounced levels, etc.), there are clear differences between DACs in A/B blind trials.

 

[Blockader]

That's not our experience. The very best DAC we've tested (DO-300) exhibits horrific waveform distortions 20dB above the commonly accepted threshold of headphone listening. With certain tests (complex-generated reverberant tails at pronounced levels, etc.), there are clear differences between DACs in A/B blind trials.

Can we see the results of these tests?

 

[olieb]

That's not our experience. The very best DAC we've tested (DO-300) exhibits horrific waveform distortions 20dB above the commonly accepted threshold of headphone listening. With certain tests (complex-generated reverberant tails at pronounced levels, etc.), there are clear differences between DACs in A/B blind trials.

Interesting, you seem to know a lot that I do not know and don't understand (I am not an audio engineer). Can you please help me by providing the answers to the following questions?

How is waveform distortion measured? What is the reference for "20dB distortion" (which is a relative quantification) when the waveform is square, sine, triangle or whatever?
What is the commonly accepted threshold of waveform distortion for headphone listening?

And I am sure I and many others are tremendously interested in those "complex-generated reverberant tails at pronounced levels". Can you link to such a test or provide audio samples of those? We could all experience the obvious differences between our DACs. That would be great.

 

[krabapple]

That's not our experience. The very best DAC we've tested (DO-300) exhibits horrific waveform distortions 20dB above the commonly accepted threshold of headphone listening. With certain tests (complex-generated reverberant tails at pronounced levels, etc.), there are clear differences between DACs in A/B blind trials.

i.e. pathological conditions. No one listens to audio that way normally.

 

[signalpath]

Can we see the results of these tests?

Sure. Here's a DO-300 sine wave at -100dBFS. That's quiet, but perceptible, especially with good fitting headphones. The AP2722 analog plot exhibits self-noise at this level, so we developed a special measurement technique. I can share the signal path if anyone's interested. Every other single-path DAC we've tested at -100dBFS looks far worse than this plot. Some are just pure noise, including a >$20,000 DAC, which will remain nameless.

 

[Mnyb]

Sure. Here's a DO-300 sine wave at -100dBFS. That's quiet, but perceptible, especially with good fitting headphones. The AP2722 analog plot exhibits self-noise at this level, so we developed a special measurement technique. I can share the signal path if anyone's interested. Every other single-path DAC we've tested at -100dBFS looks far worse than this plot. Some are just pure noise, including a >$20,000 DAC, which will remain nameless.

Now keep the headphones on and cue up a music track without changing the volume ?

 

[krabapple]

Sure. Here's a DO-300 sine wave at -100dBFS. That's quiet, but perceptible, especially with good fitting headphones. The AP2722 analog plot exhibits self-noise at this level, so we developed a special measurement technique. I can share the signal path if anyone's interested. Every other single-path DAC we've tested at -100dBFS looks far worse than this plot. Some are just pure noise, including a >$20,000 DAC, which will remain nameless.

Surely 'good fitting headphones' aren't the only necessary condition.

 

[signalpath]

i.e. pathological conditions. No one listens to audio that way normally.

In the pro audio world, we listen to reverb tails. Pathological, indeed . We listen to music and ambience until the moment we lose perception. Of course, recording levels this low are only as good as the recording itself (room, mic, adc, pre, technique, program, etc). Some studio friends did a DAC shootout in L.A. a while back, like 25 DACs. Blind. They listened only to complex-pulsed reverb tails, at levels most people (perhaps like yourself) would never consider. Hugely revealing. Very-low-level DAC performance impacts atmospherics and timbre purity. Recording engineers know that imaging and depth is best perceived in the very quiet passages, and DACs with the best very-low-level waveform purity deliver the best spatial reality.

 

[JustJones]

Some are just pure noise, including a >$20,000 DAC, which will remain nameless.

So, only inexpensive DACs get named?

 

[signalpath]

So, only inexpensive DACs get named?

Good performing DACs are always a pleasure to talk about. The DO-300 price-performance is excellent. That said, my original point is that ALL single-path DACs today exhibit poor low-level waveform performance, and such performance is both measurable and perceptible. This was in response to the person who said "sound quality of properly designed DAC’s has been above human capabilities for decades."

 

[JustJones]

What's the $20K DAC price performance?

 

[Mnyb]

This surely shows up in the usual measurements too .

Same DAC that exhibit nice SINAD and good linearity ( there is a graph for the linearity in the DAC tests ) will have regonisable waveforms at low levels ?

I doubt it’s audible listening to music as a normal person does .

Sitting in my living room with speakers and the volume at a reasonable but loud listening level I doubt I hear anything when the signal is -80dB

Recordings/Albums that have meaningful content at -100dB that’s distinguishable from random noise who does those I’m convinced you can contrive a test signal . But can I buy music at this quality level ?

 

[SIY]

That said, my original point is that ALL single-path DACs today exhibit poor low-level waveform performance, and such performance is both measurable and perceptible.

If you're talking about -100dB levels of noise or deliberately incompetent signal production (e.g., non-dithered)... yawn. If not, could you please be specific, beyond showing an unlabeled graph of data taken in an unspecified way?

 

[Blumlein 88]

In the pro audio world, we listen to reverb tails. Pathological, indeed . We listen to music and ambience until the moment we lose perception. Of course, recording levels this low are only as good as the recording itself (room, mic, adc, pre, technique, program, etc). Some studio friends did a DAC shootout in L.A. a while back, like 25 DACs. Blind. They listened only to complex-pulsed reverb tails, at levels most people (perhaps like yourself) would never consider. Hugely revealing. Very-low-level DAC performance impacts atmospherics and timbre purity. Recording engineers know that imaging and depth is best perceived in the very quiet passages, and DACs with the best very-low-level waveform purity deliver the best spatial reality.

What is your measurement technique? Yes, at this signal level there will be wide variety of differences. OTOH, you are mistaken, despite the revered collective wisdom of studio friends that such reverb tails are important for imaging and depth cues. Your ears simply don't work at that level for that. Sorry, that is a fantasy and a myth.

 

[krabapple]

In the pro audio world, we listen to reverb tails.

So do we in the non pro audio world. It's hard to avoid them.

At what playback level do you listen to reverb tails? Can you perceive the difference at the normal playback level, or do you have to crank things up during the reverb tails?

Pathological, indeed . We listen to music and ambience until the moment we lose perception. Of course, recording levels this low are only as good as the recording itself (room, mic, adc, pre, technique, program, etc). Some studio friends did a DAC shootout in L.A. a while back, like 25 DACs. Blind. They listened only to complex-pulsed reverb tails, at levels most people (perhaps like yourself) would never consider.

Uh huh. There's a reason for that.

Hugely revealing. Very-low-level DAC performance impacts atmospherics and timbre purity. Recording engineers know that imaging and depth is best perceived in the very quiet passages, and DACs with the best very-low-level waveform purity deliver the best spatial reality.

Alas, over four-plus decades in this hobby, I've read distressing amounts of mythology and nonsense written by recording engineers (some, not all), in the course of telling us what they 'know'. So that argument hold less water for me.

 

[danadam]

Every other single-path DAC we've tested at -100dBFS looks far worse than this plot.

What does it mean "far worse"?

Here's the output of TANCHJIM Space:

• connected to a laptop on battery power
• playing -100 dBFS sin 1k at 44.1k sampling rate
• connected through 3.5 mm output to RME ADI-2 Pro FS R BE
• recorded at 88.2k sampling rate and +4 dBu ref level
• digitally attenuated by 4.45 dB (that's the difference between the DAC's max output and +4 dBu ref level)

To me it looks similar to the one from DO-300. AFAICT, yes, the noise is about 20 dB below the signal, but that's nothing "horrifc", on the contrary I'd say. The distortions are 40 dB below the signal, and I suspect those 50 and 100 Hz could be from ADC.

(I amplified the signal for the waveform plot but it doesn't change anything)



If I did any blunder in the procedure then I appreciate corrections.

 

[Blumlein 88]

What does it mean "far worse"?

Here's the output of TANCHJIM Space:

• connected to a laptop on battery power
• playing -100 dBFS sin 1k at 44.1k sampling rate
• connected through 3.5 mm output to RME ADI-2 Pro FS R BE
• recorded at 88.2k sampling rate and +4 dBu ref level
• digitally attenuated by 4.45 dB (that's the difference between the DAC's max output and +4 dBu ref level)

To me it looks similar to the one from DO-300. AFAICT, yes, the noise is about 20 dB below the signal, but that's nothing "horrifc", on the contrary I'd say. The distortions are 40 dB below the signal, and I suspect those 50 and 100 Hz could be from ADC.

(I amplified the signal for the waveform plot but it doesn't change anything)
View attachment 363082
View attachment 363072

If I did any blunder in the procedure then I appreciate corrections.

You didn't use the developed special measurement technique which apparently is being kept a secret.

 

[signalpath]

What is your measurement technique?

Our AP2722 injects its own visual distortions starting around 50-60uVrms. To overcome this, we use a low-noise preamp (Millennia HV-3) with -133.2dB EIN (BB/UW) @ 60dB and 50R. The signal path is thus:

AP2722 30uV Digital Sine Wave ==> DUT ==> HV-3 Preamp @ 60dB Gain ==> AP2722 (analog display)

Of course, we're now looking at a 30mV sine wave print (1000X), but it faithfully represents 30uV performance. Anything below 25-30uVrms and the self-noise of the HV-3 preamp becomes visible. Unless the new APx555b has markedly improved self-noise performance, I think this represents the current limit of visual sine wave testing. If anyone knows a method that can reach deeper, please let me know.

EDIT: I just reviewed my notes and see that the DO-300 has a rated headroom of 5.2Vrms (around +17dBu). If I did the math right, this would put our 30uV DO-300 visual waveform test at -105dBFS, not -100dBFS).

 

[Great Sun Jester]

DACs actually DO sound different, so ...

Seriously, there are a couple of reasons they do. The DAC chips, themselves, act differently - for example, I've never been fond of ESS Sabre chips, but like Burr-Brown / TI. In addition, the output stage can make the resulting (i.e., analog) output sound COMPLETELY different, even if the initial digital conversion is identical.

So I believe your premise (as I read it at least, which is: "all DACs sound the same") is incorrect.

I know the thread is old but ESS are clinical sounding and I am also a fan of Burr Browns.

 

[Blumlein 88]

Our AP2722 injects its own visual distortions starting around 50-60uVrms. To overcome this, we use a low-noise preamp (Millennia HV-3) with -133.2dB EIN (BB/UW) @ 60dB and 50R. The signal path is thus:

AP2722 30uV Digital Sine Wave ==> DUT ==> HV-3 Preamp @ 60dB Gain ==> AP2722 (analog display)

Of course, we're now looking at a 30mV sine wave print (1000X), but it faithfully represents 30uV performance. Anything below 25-30uVrms and the self-noise of the HV-3 preamp becomes visible. Unless the new APx555b has markedly improved self-noise performance, I think this represents the current limit of visual sine wave testing. If anyone knows a method that can reach deeper, please let me know.

Yeah that will work for low level signals. Done that before myself, though I don't have a Millenia so not quite as quiet. Used this approach to measure low level linearity of DACs.