Ok - so I did an audition this AM between native 16/44.1 and my 'super sauced' converted redbook, this time with the Gungnir. With the Squeezbox Server its easy. I can play one file then the other in quick succession, back and forth. Same file, native, then upsampled.
One thing I can say for sure. The Gungnir handles 16/44.1 very well. With my Transporter DAC and the MSB DAC, there was a CLEAR lack of digititis with the 'massaged' files. Clearly 'smoother', less irritating, more 'musical', 'better' more natural highs etc.
With the Gungnir the task of finding a 'difference' was much, much harder. When I thought I noticed one, then on the next repeat try I heard something else that canceled that previous observation. There is NO WAY I could tell them apart in a blind test. At least now, with only about 60 min. of trying. Possibly with hours of exhaustive testing I could find a difference that I could pick out.
I'm sticking with the upsampled files only because I 'think' I liked them better. I think they were ever so slightly more 'musical'. Probably purely psychological. I even took the 16/44.1 and applied the same filters, dither, aliasing, and smart limiter to make for a more even comparison. I was hoping the 16/44.1 would sound better. There was no change.
The take away is that I dont completely agree with the above statement bashing upsampling techniques. I agree it shouldnt be done in the DAC. That doesnt make it all 'bad'.
Summary of my upsample 'formula' - Windows 7, Foobar with Sox SRC, output - WAV, 24bit, dither (always), resampler - quality (best), passband 95%, phase response minimum (0%), allow aliasing. Followed by advanced limiter (senses digital clipping).