How to prevent noise that exceeds Nyquist frequency of input signal when upsampling

[Reffi]

I am not an expert in audio signal processing, and have no experience in creating resampling software.
I hesitate to explain things that you probably know more about than I do, but let me check to clarify where the problem is occurring.

http://bfin.sakura.ne.jp/oldsite/2191/p ... se02.shtml
https://bfin-sakura-ne-jp.translate.goo ... r_pto=wapp
I couldn't find any English literature (mainly because of my English skills).
The above site is in Japanese, so please forgive me if the machine-translated text is difficult to read.

I am concerned that the low-pass filter (Interpolation Filter) is not working properly during VB-CABLE's upsampling process.
It is because the frequency band and shape of the noise occurring in this phenomenon seem to be similar to
the result of only zero-interpolating the sample values ​​without applying a low-pass filter.
(A ghost spectrum that is a mirror image at the boundary of the input signal's Nyquist frequency (half the frequency of the input signal) occurs.
In principle, it will always occur. Therefore, it is generally necessary to cut it out with a low-pass filter as post-processing.)
Does VB-CABLE apply a low-pass filter in the upsampling post-processing?
(Or is there no problem with the signal output by VB-CABLE, and something is wrong with the software that receives it?)

[Vincent Burel]

yes, i've 2 or 3 notions on the subject.

The first point is to be able to reproduce your measure, because as i already said, the VB-CABLE SRC is not expected to generate more than -70 dBfs noise (for a zero dBFs source).

So i've noted your question, i will come back to you when i will find the time to work on this old subject .