Upsampling Imaging - Anyone Know How to

[lokkerman]

I've attached a spectrogram from my audio analyser of choice - Sonic Visualiser www.sonicvisualiser.org. It's a simple Redbook analysis of Luther Vandross's classic "Stop to Love" - you can see that it is a CD as it has no more energy after 22K.

Now I have a conundrum. If I upsample this using SoX, to 192KHz, using Xrecode 3  I get the following results:

Looking at the BIN frequency in the right hand top corner the visual states a frequency of 25312.5 - 25875Hz - this is a distinct mirror of the top range of the original, just below the bin frequency, we can also see shadows vertically up the image.

Now my conundrum. I use AIMP in upsampling mode. This uses a very good upsampler, see http://src.infinitewave.ca/  for comparisons.

I have it set to output everything to WASAPI analogue out in my ASUS Xonar D2 at 32bit/192 KHz. And it appears to work (see above)  there appear to be peaks around 50KHz that were never there before and the imagining is the echo above at around 75KHz. Question is am I correct in my understanding of this? For the sake of completeness below it was played through the Reel to Reel monitor circuit, before recording using Audacity to record at 32/192 wav.

Finally my tests using a Reel to Reel; (Ferrograph Logic 7 with new record head and recapped) these appear to have naturalised the new peaks and yes it is content out of analogue above 38KHz. Can anyone please help explain that I am correct in my conjectures? And BTW the R2R sounds incredibly better. tape used semi-virgin BASF LP35LH which is what the Logic 7 was designed to use as the standard tape.

[lokkerman]

Sorry what I was trying to get to understand (but the post appeared to have chopped it out) was:

Am I right in understanding the visuals of upsampling artefacts known as imaging.

And that it appears that R2R somehow adds an analogue edge to upsampled sound.