Let's develop an ASR inter-sample test procedure for DACs!

[krabapple]

If a tree falls in the woods but no one hears it, is that bad?

 

[DVDdoug]

If I recorded an acoustic guitar, and I set my gain so that the captured signal is very close to 0dBFS (very high SNR), to the point that it often just barely clips the ADC for fractions of a second (almost inaudible), would you say that's a good recording? Would it be acceptable if almost every recording were like that?

That's an "artistic" question. Many commercial recordings are intentionally clipped (during mastering) and/or electric guitar is often intentionally distorted.

"Almost inaudible" is likely totally inaudible if the listener can't A/B.

and I set my gain so that the captured signal is very close to 0dBFS (very high SNR),

FIY - In most cases turning down the recording level knob and leaving headroom turns-down the signal and noise so it doesn't hurt the SNR. A strong acoustic into the mic is usually a good idea, and a more sensitive mic can better overcome preamp noise..

 

[tccalvin]

Many commercial recordings are intentionally clipped (during mastering) and/or electric guitar is often intentionally distorted.

Yes, this is true. Although we should specify that clipping in an ADC or a DAC is not really the same as the "graceful" harmonic distortion that you can find in analog gear (or digitally emulated analog gear). Even in music production, specific devices and/or plugins are used to get specific musical flavors of harmonic saturation. This process is more about shaping the wave with harmonics and filtering. DAC/ADC clipping means running out of data and causing the oversampling filter to behave in an unintended manner. I guess it's still possible to use this kind of clipping "artistically" but hardly anyone does.

 

[krabapple]

The 'artistic' choice is to lower the dynamic range, which some equate with impact. It could also just be a commercial choice (e.g., the artist would prefer a broader DR, but knows the public likes louder)

 

[Mnyb]

Thomas Lund has written several papers about this . Here is one of them .

https://service-tcgroup.tcelectronic.com/media/Level_paper_AES109%281%29.pdf

We should separate the issue of hot masters intentionally somewhat clipped vs actuall ISO’s that’s actually an error with to high level.
Similar but different is it not ??

Clipping gives a traditional distortion increase, ISO behavior can be different in different products software filters and use scenarios .
It can be benign to actuall register wrap around and overflows.
A fancy DAW ( or equivalent process) that does everything in floating point math vs some circuit with integer math only ?

So even for loudness war masters i would not want 0dBfs digital volume.

For peace of mind one can introduce some negative digital gain somewhere especially if you use EQ