But is there REALLY? And not just for NOS, I've long wondering what would happen if you did not have an output filter at all since it causes problems. Yeah in theory all this aliasing of images blah blah blah. However that supposes there is significant ultrasonic content coming into the A/D in the first place, which I seriously question how much of that there is in the recording space, and further question how much of what there is actually makes it through the microphones. I've never seen any serious investigation of this. Maybe that's a good thread topic; some members might be set up to measure that?
No, it doesn't suppose there is significant ultrasonic content into the A/D to start with. I'm not sure what you're saying, because you acknowledge images. The images are result of PCM. Of course, the DAC is not putting out impulses, at least ideal ones, so there will always be some filtering (if implicit). But if you sampled a 1 kHz sine at 48 kHz, the first image is at 47 kHz. The point being that you don't need to record "significant ultrasonic content" to get significant ultrasonic content if you leave off the filter.
I probably don't fully understand what a NOS DAC might be—from what I've read it seems a big of a non-specific description. At face value, it would be one without an oversampling filter, and to me that implies an R-2R ladder DAC. But as
@solderdude points out, people take it to mean no reconstruction filter. But since they aren't using (and can't use) ideal impulses, whatever they are doing results in convolution that is in effect, some type of filter. @solderdue mentions sample and hold (zero order hold), which is the typical shortcut, and in fact that's a weak filter. A problem with that is that if you're not running a higher sample rate, you have frequency droop in the audio band nearing the Nyquist frequency. That's usually compensated for by the filter. But any way you want to look at it, exactly what a "filterless" DAC outputs depends on...how it outputs its signal. That opposed to, for a DAC with a "reconstruction" filter, which would also make up for any design choice that precede it int he output path.
I don't know, but to me this sort of thing is an extension of "we were so happy with analog, then digital came along and ruined music" sickness. Oh, for the days of hiss, pops and hum. Now we (some) have to worry about what their filters looks like on paper. Before digital, I'd never heard people concede they couldn't hear something, yet were concerned it could be making them feel uneasy or something.