If I record at 16 bits and mixdown to 24 or 32 bit have I improved anything or made it worse?
I use Band-in-a-Box to create background tracks and it will only produce 16 bit files.
Would it help to convert to 24 bit before I record vocals or lead tracks?
What about mastering? I could mixdown at 32bit and apply mastering changes, them dither to 16 bit for the CD?
Am I accomplishing anything?
Any ideas or comments?
Bax
Do a null test at 32 and 16. I think I did this before and found going to 32 bit gave a little better highs in cymbals and other high frequencies. I would mix down @32 bit take the stereo track and mix it down again to 16@ 48,000hz for CD quality.
EDIT: I think If I remember correctly when I mixed down a project 24BIT@96,000hz to 16 bit@48,000hz and did a null test, I found some high frequencies not being dithered in the 16 bit stereo track, When I mixed down to a stereo track 32 bit and then mixed it down again to the same 16bit@48,000 using the same dither, those frequencies in the null test didn’t show up and I notice a slight improvements in some highs in my song. Hope that helps you Bax.
Cheers
PACO 
Thank you so much for the research and responding to my question.
I hope you will indulge me with some additional answers…
I have been unable to find any information on going to a higher bit rate.
In the User manual Flavio does suggest going to 32 bit if additional processing is wanted, but mothering on 24 bit.
So, if I understand correctly, mixing to 24 bit has minimal effect and may even not dither correctly when starting from 16 bit - is that your finding?
I realize that 48000 is a direct transition, but why go to that first instead of just going directly to 16 x 41100 if I’m going for a CD?
Does 48000 work belter for MP3?
Thanks,
Bax
48,000hz converted to 320k mp3 yes is better. The thinking is the higher bitrate means a better sample for the MP3. You can put 48,000hz on a CD, at least I can with nero burning rom although the end result is a 44,000hz cd. Again the thinking is the higher bitrate means a better sample. Can’t put anything higher on the CD won’t burn. I can’t create a project higher than24bit@ 96,000hz due to a limitation on my line 6 UX2 and my NI Audio, both capped at 96,000hz@24 bit. “So, if I understand correctly, mixing to 24 bit has minimal effect and may even not dither correctly when starting from 16 bit - is that your finding?” Yes. A null test showed some high frequencies in the test. Where there should be nothing seen. This is the purpose of the null test to see if the mix down it exactly the same as the original “frequency wise”. When I rendered the mix, a 24bit,96,000hz project down to 32 bit floating point stereo wave file and then took that file and rendered it again down to 16bit@48,000hz stereo wave file both the 32 bit and the 16 bit renders, the null tests came back clean. It’s was only rendering directly from 24@96000hz to 16 bit@48,000 do the null tests reveal issues. Hope that explains it a bit further Bax 
EDIT: There was some lengthy discussion about this a long time ago and I was told back then, before I knew what a null test was, that “I could hear there was a slight improvement” rendering it to 32 bit floating point, then down to 16 bit 48,00hz and I was told abruptly that it was impossible from a rendering perspective. The person claiming it was impossible said I should do a null test and that’s when I proved him wrong. And my ear was, in fact, correct, I have very good hearing that I could hear those very high frequencies. The null test proved it, both to the guy claiming it was impossible for the renders to be different and to my own ears.
Thank you. That information will be a big help to me. My hearing suffers from the “old guy†sendrom where high frequencies are not heard well at all…
I had looked everywhere I could and the answer just was not there.
Bax 
You welcome Bax. Anytime.