Which is a better THD+N Ratio Vs Freq graph?

[Fly2High]

I am looking at two DACs- The SMSL SU-9 and the soon to be available Topping D30 Pro.

This is the graph from the SMSL SU-9:

This is the one for the Topping D30 Pro

Which is better/preferred?

The SMSL SU-9 has lower Ratio but it is not flat.
The Topping D30 Pro is much higher but relatively very flat in comparison.

How would each manifest audibly? If both are below human threshold, let's pretend then their curves are move higher up the graph to be audible. What would I hear? How would they sound differently?

thank you.

Frank

 

[AnalogSteph]

How would each manifest audibly?

These graphs will tell you nothing about that. You would need to have a look at the actual FFT spectra (ideally, for every single frequency) in order to determine how much is actually contained in the audible range. (Anything ultrasonic, unless at substantial levels, can generally be ignored.) Note that the measurement bandwidth in this particular test is 90 kHz, including plenty of ultrasonics. While this means it can pick up several 20 kHz harmonics, it also makes it prone to all kinds of artifacts, including imaging due to imperfect reconstruction lowpass filters.

In case of the D30 Pro, it turns out most of the IMD+N is actually ultrasonic noise that the noise shaper has shoveled up there (an approach sort of analogous to dealing with snow on the roads in winter time):

Meanwhile, things in the audible range are completely uneventful, as the other (20 kHz band-limited) tests show.

I would actually like to see these spectra for some ESS-based DACs, including ones afflicted with the infamous "ESS hump".

These sorts of things should mainly be interesting you if you need a DAC for measurement purposes. or for running an FM stereo modulator or SDR transmitter or something, where what's going on up there is actually relevant.