[RRod]

Screenshot of original 96kHz and various resampling algorithms. From top: original 96kHz, converted to 44.1kHz with: dBPoweramp (via new dBPoweramp engine option in foobar2000 1.3.9), Audacity 2.0.6 Best quality, foobar2000 1.3.9 PPHS Ultra, foobar2000 1.3.9 PPHS.

Note that some differences between the original and the converted are inherent to the difference in sample rate and how the spectrogram is computed different at the different sample rates. For example, the same 32768 sample Blackman window is used for the original 96kHz as well as the 44.1kHz conversions, which resulted in a wider effective window width view in the conversions, causing the start / end transient artifacts to appear wider. There are differing shapes of rolloffs, which aren't really visible in this view. The dBpowerAMP algorithm has crazy bandwidth, practically not rolling off until 22000Hz...

I will have to update to foobar2000 1.3.9 for the dBPoweramp engine option it sounds like, although the harmonics were probably inaudible using PPHS Ultra to begin with (will have to go home and listen some more)
 

 
Thanks for running those, Joe. Any way I could bug you for some pretty pics of the impulse responses?

 

[Joe Bloggs]

Just out of interest, what settings are you using for the spectrograms in Audacity there? (the numbers in Edit->Preferences->Spectrograms) I'd like to repolicate them and compare to Audition. There's certainly a lot of alising and associated harmonics in the PPHS versions.

I've come home and left the spectrograms at work, but I can reproduce the results as follows
Sweep: 0-22050Hz 96000Hz 0.8 magnitude 1 min. linear chirp
Output: 32bit float
Window: 32768 Blackman
Display:
0-22050Hz, 0dB gain, range: about 190dB (!)

Those harmonics are WAY down in magnitude compared to the signal.