[Jack-Devel] Jack Problems

[Fons Adriaensen]

On Tue, Mar 26, 2019 at 08:02:29PM -0400, Tim wrote:

> 2) Refer to Fons' comments, if I may paraphrase, signal levels
>     are best adjusted at the source if possible at an A/D converter,
>     not somewhere down the digital line. Adjusting down the line
>     introduces some digital error.

That is NOT what I said, there is nothing wrong with a gain
control in the digital domain. My argument is that having a
gain control in *every connection* makes it very difficult
to ensure that any non-trival system will have a sane level
structure. 

> To that second point I would like to mention something I've been...
>  musing about:
> 
> It would cool if Jack optionally delivered the pure unadulterated
>  integer values from the various converters and drivers.

> Be this awesome super router, AND do integer values just like ALSA.
> Yes, I know there are problems, formats etc. but surely we could
>  come up with a system. Although, if you ask for an unavailable
>  format it means Jack would have to convert, but not expensive is it?
> 
> I am under the impression that one of the benefits of using say
>  Ardour with ALSA instead of Jack is that it might allow
>  working with the pure integer wave recordings for example.

There is no benefit. The FP format Jack uses is at least as
precise as 24-bit integer. And consider that even if you have
a 24-bit (or more) converter, at most 20 or 21 bits are actually
valid, the rest is noise. There is nothing 'pure' about the 
integer values. You can easily test this: use two channels
to convert the same analog input signal, and compare
the digitial values. You may be surprised.

Ciao,

-- 
FA