Mike - Thanks for writing with the suggestion. After a bit of work, I agree that TPDF seems to do a very good job. Here's what I did to determine that.
I installed the free Voxengo Span VST plugin into JRiver and put it at the end of the DSP chain. Then I generated a 1khz test tone in JRiver and played it back in repeat mode. I also enabled the "bitdepth simulator" in JRiver, and played the 1khz file at 32, 24, and 16 bits and I even went down to 4 to see how messy it would get without dithering. Then I set the bitdepth simulator to 18 bits, checked "dither" (TPDF) and the graph was very clean, so this is what I'm going to use.
It took a while to get this figured out but it does seem that using the bitdepth simulator in JRiver is a good solution.
Dave
I installed the free Voxengo Span VST plugin into JRiver and put it at the end of the DSP chain. Then I generated a 1khz test tone in JRiver and played it back in repeat mode. I also enabled the "bitdepth simulator" in JRiver, and played the 1khz file at 32, 24, and 16 bits and I even went down to 4 to see how messy it would get without dithering. Then I set the bitdepth simulator to 18 bits, checked "dither" (TPDF) and the graph was very clean, so this is what I'm going to use.
It took a while to get this figured out but it does seem that using the bitdepth simulator in JRiver is a good solution.
Dave