[linux-audio-user] Re: Sample Rates / Sample Sizes

[Mark Knecht]

Steve Harris wrote:
> On Thu, Jun 24, 2004 at 09:52:54 -0700, Mark Knecht wrote:
> 
<SNIP>
>>   Thanks for the clarifications though. It does raise an interesting 
>>difference between something like Ardour, I guess, and Pro Tools where I 
>>can insert a dithering plugin pretty much anyplace in the signal chain 
>>if I choose to. Maybe the RTAS standard makes some decision to go to int 
>>based on where the output is going? I don't know.
> 
> 
> I suspect that the difference is that in RTAS the conversion is specified
> (well, as theres only one host, it /is/ specified ;).
> 
> - Steve
> 

Yes, that's correct. Although there are some strange paths I hadn't 
quite considered.

1) The hardware and Pro Tools is native 24-bit, so I think everything 
that goes to the hardware is actually 24-bits deep.

2) If I insert a dither then presumably what's sent to the hardware is 
24-bits deep with the bottom 8 bits zeroed out and the upper 16-bits 
containing the audio + dither. you hear 16-bits over 24-bit hardware. 
Still OK.

3) Since bouncing is non-realtime, when bouncing with a dither inserted 
it likely bounces first a 24-bit file ala #2, then converts it to 16-bit 
by stripping out data. There is an option for conversion on the fly vs. 
afterward. Still make sense, I think.

4) If I insert a dither in the middle of a session on a single track 
then it's unlikely that the output is converted to int unless Pro 
Tools/RTAS is design to convert it right back to float so that the track 
can mix with the rest of the session. I don't have a clue what really 
happens in this case. I've never had a reason to use this, but it does 
raise the question of how to best import a 24-bit track into a 16-bit 
session. I've have to do this once or twice. Possibly I should have 
dithered the track. I'd guess the normal conversions just strip the 
bottom 8-bits?

    Anyway, interesting, I suppose, to some of us. Sorry for consuming 
bandwidth for those that are not.

Cheers,
Mark