Thanks for your reply which I especially appreciate due to your tremendous electro-technical knowledge ("jemand, der richtig Ahnung hat"). In order not to overdo the flattery, I'd nonetheless challenge some parts of this response regarding its precision.
KaiS wrote:You're aware that these "true peaks" caused by intersample peaks do not exist in the digital domain.
No worries, I am. In the original PCM data, by definition, we only have discrete, quantized sample values from a finite set.
They only occur after DA-conversion, in the analog signal.
However, intersample peaks are not an inherently "analog-only" property. It is a property of the reconstructed waveform, and that waveform can be reconstructed numerically as well as physically.
In other words: if one accepts that intersample peaks may be "introduced" (for a better word than "exist") eventually in the analog domain, one should consequently also consider the introduction of them in the digital domain during calculations such as resamplings as evidently, one can obtain samples exceeding 0 dBFS during digital oversampling/resampling from an original PCM stream whose individual samples all remain below 0 dBFS, without any D/A conversion taking place. Which is how a " 'true' peak meter" gets mathematically implemented. Otherwise one would need a DAC/ADC-loopback chain and appropriate level calibration which probably is rarely performed in practise.
The same principle applies to images or video by the way, where individual nominal pixel values can exceed the range of the original samples as a result of scaling, interpolation or other processing operations.
It's possible to simulate them by some amount, but as even the reconstruction filter's ('DA-filter's) shape has some influence on those, it can only be an estimate.
Due to the former argument, I wouldn't call it a "simulation" (rather "estimation") because from a mathematical perspective, it is as "real" as the generated analog output, but I agree that the calculated peak display will not exactly reflect the actual result. But I'm sure we concur that it would provide substantially more information than a simple 1:1 mapping of the values as we have now. In that sense, it feels odd to me to reject the idea right away to improve the precision of peak display just because it will never entirely match the "real" result.
ADI-2 BTW has an analog headroom of 2.5 dB to deal with ISPs, so their detrimental effect does not occur on the device itself.
I know, so the display would rather be "nice/cool" to have as stated in my original post, not an ultimate necessity.
My best guess is, as ISPs are omnipresent these days (masters commonly are "smashed to the wall") converter chip manufacturers have taken measures to handle them.
Therefore I think they are of minor interest now.
Sure, one can also easily argue that most productions nowadays are so lousy that any intersample clipping which may occur, won't make any significant difference anymore - if it does in general, which is also under quite some debate.
From an engineering perspective, it is always good to have additional headroom or jitter suppression - audible or not.