On Sun, Sep 22, 2019 at 7:09 PM Ralf Mardorf <ralf.mardorf(a)alice-dsl.net> wrote: I'm not sure, but I think the problem is related to ASIO or the Jack->ASIO bridge, not jack directly. As I understood it, Cakewalk recorded in 24 bits? And you get audio from jack through the Jack->ASIO bridge?

Am Mittwoch, 25. September 2019 18:16 CEST, lowkeyoutlaw <quietizkept(a)gmail.com> schrieb: I think you're getting something conpletely wrong here. Your interface will send integer values to your computer - this is where the bit depth really matters since a higher bit depth allows for more sloppy/careless input gain control (i.e. you can record at a relatively low input gain and still get "enough bits" after normalizing the signal). Your DAW (like any DAW currently on the market) internally stores/processes audio data with 32bit _floating point_ values (which are totally different from integers). This prevents both clipping and rounding errors during sound processing. The 'overkill' argument you most likely encountered was most likely talking about the bit-depth of your output. Anything above 16 or maybe 24 bit _is_ overkill, unless you produce for mebers of the Elephantidae species ... :-) What realtime conversion? Your DAW is converting integer samples to float, no matter what you do. HTH, RalfD

On Wed, Sep 25, 2019 at 09:16:54AM -0700, lowkeyoutlaw wrote: That doesn't mean you get 32 valid bits. Most probably you get 24 and 8 zeros, and if you really get 32 the lower 10 or so are completely useless. You are confusing the format (a 32 bit integer) and the actual content (most likely 24 bits). There is no debate about this, except maybe among the ignorant. It is perfectly possible to make a 32 bit converter clip. It will do so at exactly the same gain settings as a 24 or 16 bit interface. would at the same gain setting also be ruined when using 32 bits. Just leave enough headroom by using conservative levels. This will NOT affect your S/N ratio which is entirely determined by the analog circuits if you convert to 24 (or even 16) bits. You can do exactly the same with 24 bits. There is no such CPU hit. Advice: stop believing everything you read on the web. Ciao, -- FA

On Wed, September 25, 2019 11:16 am, lowkeyoutlaw wrote: That may not be a direct comparison, many programs which record in 32 bit actually record 32 bit floating point, which at the point of conversion is the same as 24 bit integer. Could possibly be 32 bit integer, but that would be rare. The online documentation for Cakewalk still has this note: "The choices for the recording bit depth setting are 16, 24, 32, and 64. At present, no devices can record at 32-bit or 64-bit resolution, so you don’t need to choose these options yet." What interface do you have which delivers audio as 32 bits? Which driver interfaces are supported? I found a copy of the ASIO 2.2 specification, that only supports up to 24 bit sample values, although those can be stored in 32 bit words (only 24 bits of sample data, just longer word length for easier handling on most archictectures), or 32 bit float. Most DAW software will convert 24 bit integer values coming in and save as a 32 bit float file (which is just equivalent to 24 bit values, plus an exponent so that large gain ranges can be handled). I would be surprised if any hardware natively provided 32 bit float values, but I guess it is possible. I also saw a reference to an ASIO driver which did provide 32 bit integer data, but I have reason to doubt that is the case. -- Chris Caudle