@DonR, @tmtohm,
@voodooless,
@Geert
"Impulse response can be converted to frequency and phase, and vice versa. They are two sides of the same coin."
Two devices having linear frequency response from 20hz to 20khz can show different time responses in the way of their impulse/square wave throughout that bandwidth, so having only the frequency-magnitude chart is going to be insufficient.
"They can’t reproduce a square wave offset between samples very well"
That is an issue of quantization/sample rate limitations.
"They don’t!"
A delta-sigma's innate use of oversampling affects the impulse response and puts it at a disadvantage in comparison to a filter-less non-oversampling dac's impulse response. A dac may have low sample rate music or oversampled music played into it and may have a reconstruction lpf (all affecting the impulse response), but these are not innate qualities of say a resistor ladder architecture, hence my use of a filter-less NOS kind for emphasis.
"(or high frequency content for that matter)."
A filter-less dac is going to have an abundance of high frequency content. It is supposedly as a result of limited bandwidth of upstream devices that this would result in intermodulation, but if intermodulation is understood as distortion in the form of frequencies that weren't in the original signal, then all that would really be the cause is the modulation frequency (ex. 44.1khz). I suspect that when intermodulation tests are taken on a dac that it is from 20hz to 20khz, so the program does not recognize/is not informed of the higher-than-20khz sample rate.
Hearing 44.1khz or 48khz music played through a filter-less NOS, people have both good and bad to say about it; it is realistic, but also 'hazy' sounding. This is because the amplitude changes that are recreating audible tones are at a rate that is less than half of our time sensitivity, yet 44,100 times per sec is still a high rate generally speaking (hence 'hazy'). This is why non-oversampled 96khz (near the aforementioned 100khz limit) music played through the dac would overcome the hazy sound and the need for a reconstruction lpf and oversampling (read delta-sigma aswell).
"The fact that they have high distortion means they cannot reproduce the origin signal very well. How can something that is worse at reproduction the original signal have better timing performance?"
A device can have high distortion in some areas and low distortion in others, i.e. something can be worse at reproducing other aspects of a signal and better at the timing aspect.
"That’s not a DAC."
That's just my point, because the idea that you can't necessarily infer from positive multitone and THD+N vs frequency that the impulse/square wave will also show positive applies to not just dacs but amps and speakers, meaning more measurements have to be shown to draw conclusions about sound quality.
"But that doesn't mean that an audio reproduction device needs to be able to reproduce 100kHz sounds - nor does it mean that there are digital devices that aren't "fast enough" to accurately reproduce amplitude changes in recorded music. If that were true, then digital sampling theory would be invalid and plenty of gear and systems - way beyond just hi-fi equipment - simply would not work."
First, given frequency is a continuum, I'd be surprised if most audio electronics are incapable of playing 100khz, instead of just playing it at a very low level. Second, of course a device is not broken if it can't play linearly up to 100khz or more, it is to say that there is room for improvement.
"Phase distortion above the audio band does not matter: https://www.audiosciencereview.com/...se-distortion-shift-matter-in-audio-no.24026/"
"You seem to be saying that we can detect changes in amplitude with a speed that exceeds the speed of the wavelength of a 20kHz sound."
"Reference?"
At 13:05 in the video by amirm that is linked he says that time differential is a main factor in localization, and we can ask: what amount of time?
My previous post stating a 100khz-equivalent time sensitivity was gotten by a quick search while making the post, the context of which is even the same as in amirm's video:
"One final fascinating phenomenon is the interaction of the two ears in interpreting sound. A big advantage to having two ears is the ability to accurately localize sound. There are two cues that the brain uses in doing this. The first is, as we mentioned earlier, the fact that the time of arrival of sounds in the two ears is slightly different. The closer ear receives the sound slightly earlier. The brain is sensitive to differences in time of arrival of as small as 10 microseconds, and can use this to pinpoint the location of the sound. The second cue is the fact that sounds arrive at slightly different amplitudes in the two ears. This is because the head causes an acoustic shadow which attenuates a sound coming from the opposite direction. By comparing the amplitude of the sounds in the two ears, the location of the source can be identified.
But if this is true, then how do we localize indoors? Sounds bounce numerous times off walls, ceilings, floors and other objects which would totally confuse the brain. It turns out that there is a precedence effect by which the brain only pays attention to the first wavefront that reaches it. All the subsequent echoes are ignored for the purpose of localization."
(From:
https://web.mit.edu/2.972/www/reports/ear/ear.html)
What is interesting is that I wasn't even going to bother making a search to check and was going to put the same number from memory because I had read on this subject prior and from other sources. They are consistent with each other. I will find the other website I came across, it is somewhere in my phone's files.