[VNandor]

I converted a 192kHz wav which contained frequencies well above 22 kHz to 44.1kHz. Then I made a null test with the 192 and 44.1 wav expecting a perfect null up to 22 kHz but that's not what happened. Is it because of aliasing? Would I get a perfect or near perfect null if I used a low pass filter before converting the 192kHz wav?

 

[RRod]

  I converted a 192kHz wav which contained frequencies well above 22 kHz to 44.1kHz. Then I made a null test with the 192 and 44.1 wav expecting a perfect null up to 22 kHz but that's not what happened. Is it because of aliasing? Would I get a perfect or near perfect null if I used a low pass filter before converting the 192kHz wav?

 
A converter should lowpass before the decimation to the lower rate, but some of them do have options to allow a bit of aliasing. Why not pass a 192k sine sweep through the same converter and see what you get. Also, have you looked up the converter on http://src.infinitewave.ca/ ?

 

[MindsMirror]

A 44.1KHz anti-aliasing filter must have a very steep and narrow transition. You will end up with some compromise between pass-band attenuation, stop-band attenuation, ripple, phase, ringing, aliasing, etc. Unless your resampler used a sinc brick-wall filter, it will still pass some low level signal in the stop-band, which becomes aliasing after it's sampled.
 
There will also be quantization error.

 

[VNandor]

   
A converter should lowpass before the decimation to the lower rate, but some of them do have options to allow a bit of aliasing. Why not pass a 192k sine sweep through the same converter and see what you get. Also, have you looked up the converter on http://src.infinitewave.ca/ ?

Good idea, I tried it and it looks like everything is okay. I got silence after the signal passed ~22kHz.

  A 44.1KHz anti-aliasing filter must have a very steep and narrow transition. You will end up with some compromise between pass-band attenuation, stop-band attenuation, ripple, phase, ringing, aliasing, etc. Unless your resampler used a sinc brick-wall filter, it will still pass some low level signal in the stop-band, which becomes aliasing after it's sampled.

Yeah, I know equalizers are not perfect that's why I assumed I won't get a perfect null. Do you happen to know how big the magnitude of the errors are? Would different software produce different results?

 
There will also be quantization error.

I didn't think about that but it seems obvious now. But again I have no idea how it would look like in a graph. Is it just a random noise across the whole spectrum?

 

[RRod]

 
Good idea, I tried it and it looks like everything is okay. I got silence after the signal passed ~22kHz.
Yeah, I know equalizers are not perfect that's why I assumed I won't get a perfect null. Do you happen to know how big the magnitude of the errors are? Would different software produce different results?
I didn't think about that but it seems obvious now. But again I have no idea how it would look like in a graph. Is it just a random noise across the whole spectrum?

 
Here's what the SoX very-high-quality resampler does with 192->44.1->192 white-ish noise, staying at 32-bit integer the whole way (this is the difference spectrogram):