Upsampling and filtering

Just a quick question:

When I'm converting from 44.1k at 32-bits to 48k for 24-bit/48k release, how concerned do I need to be with pre- and post-filtering?

I always put the quality slider to 999 and then don't use filtering.  This was effective at a scary-good level in CEP2.1 and before, which I still use for sample rate conversions.  Audition still pretty much rocks in this department.

Generate some tones that sweep up (and over, if starting from a higher rate) around the Nyquist rate, convert, and compare the files in spectral view with a high dB number.

When I'm converting from 44.1k at 32-bits to 48k for 24-bit/48k release, how concerned do I need to be with pre- and post-filtering?

Well, since the idea is to get rid of potential aliasing, then not at all in the direction you're going in! 

As for the quality slider, Syntrillium said that it didn't make a lot of difference at any setting over 250 - but the argument made about this applies more to downsampling, especially when converting to 16-bit. Personally I don't think that you could measure a lot of difference at settings between 250 and 999, but it certainly takes longer to do the sums when it's set higher. With modern processors, I doubt whether this is much of an issue though.

When I'm converting from 44.1k at 32-bits to 48k for 24-bit/48k release, how concerned do I need to be with pre- and post-filtering?

Well, since the idea is to get rid of potential aliasing, then not at all in the direction you're going in! 

That was my assumption. The help file is, however, somewhat ambiguous on this point, merely stating "When upsampling, results from lower values sound almost identical to those from higher values." (My italics)

What niggles me is not really understanding what they're getting at here.

Downsampling sounds better than upsampling?  Upsampling takes more of a hit than downsampling?  Backsampling is better than forthsampling?

Maybe they weren't sure either.