The lovely, perfectly shaped sine wave shown here is an illusion...
The point of
that entire post is to show that we will often run into a DAC's audioband noise floor before we reach the limits of its ability to decode linearly at the lowest dBFS levels. So if we want to observe linearity below the noise level across the audioband bandwidth, we have to use a bandpass filter.
...Nobody knows and can show how that signal looks like, no matter what DAC is used, because it is covered too deeply in noise...
Actually in those two figures (copied below, Fig.1 and Fig.2), we
do know what the 1 kHz stimulus looks like to the analyzer, first with the noise from the audioband present...
Fig.1 Schiit Yggdrasil 2 output of 1 kHz sine at -130 dBFS (from its balanced analog output), with no bandpass filtering.
...and without all the noise from the audioband present (by using a bandpass filter, in this case elliptic filters on either side of the 1 kHz stimlus)...
Fig.2 Schiit Yggdrasil 2 output of 1 kHz sine at -130 dBFS (from its balanced analog output), with 1 kHz elliptic high-pass filter and 1 kHz elliptic low-pass filter applied.
While it does take bandpass filtering to see the waveform properly, it is coming from the DAC's outputs, and thus it does have to be decoding it at that dBFS level for this to show.
...If you apply a steep and narrow bandpass filter to a signal - whatever signal - it will always look like a sine wave having the frequency of the bandpass center. Feed it triangle, sawtooth, square - doesn't matter...
The signal has to be there in the first place. If at a given dBFS level the DAC is not outputting the waveform then any amount of filtering isn't going to unveil (or make clearer) something that doesn't exist.
...Feed it triangle, sawtooth, square - doesn't matter. The output of the DAC might have steps, stairs, other irregularities (from theory it definitely will have those) - but there is no way to show it as waveform...
I think you're referring to the sinc interpolation that happens by default in APx for time domain measurements. It can be switched off, but is on by default. Here's the DAC outputting a 1 kHz signal at 0 dBFS (sample rate 44.1 kHz, bandwidth-limited by high-pass elliptic filter at 1 kHz and low-pass elliptic filter at 1 kHz), with the interpolation option visible (bottom-left):
Fig.3 Schiit Yggdrasil 2 outputting a 1 kHz signal at 0 dBFS (sample rate 44.1 kHz, bandwidth-limited by high-pass elliptic filter at 1 kHz and low-pass elliptic filter a 1 kHz), with interpolation switched off (and the interpolation option showing bottom-left)
And if you want to see the waveform without interpolation at -130 dBFS:
Fig.4 Schiit Yggdrasil 2 outputting a 1 kHz signal at -130 dBFS (sample rate 44.1 kHz, bandwidth-limited by high-pass elliptic filter at 1 kHz and low-pass elliptic filter a 1 kHz), with interpolation switched off
(At this low level, the sine wave is not particularly stable, but does maintain its sinusoidal shape.)
...The only way to get a grasp of its pureness is doing a very deep THD analysis via FFT. But at -130 dBFS even that usually fails, as the noise floor in the FFT will not go lower than around -160 dBFS. So anything better than -30 dB THD is covered in noise again.
That's what I did in the post you quoted:
Nested FFT measurements. Beneath Fig.2 and Fig.3 in that post, you'll see links to view those same plots with the Y-axis zoomed out.
If I'm misinterpreting any of your comments, let me know.