CAKES' 32-bit MIX ENGINE?

[MELVIN MCKEE]

What all does the 32-bit mix engine do behind the scenes? Is Cakewalk a
32-bit application now? How much DSP is done by this mix engine? What
processing is not done in 32-bit and at what bit rate is it done?

I have already purchased V.7, but I would like to have this detailed
info.

[Mike Garrett]

MELVIN MCKEE wrote in message <357B4AA7...@deltech.net>...

CWPA has been a Win32 application since 5.0. Now, in 7.0, the audio path
internally used for mixing is 32 bit normalized float (right guys?) This is
a new high precision format supported by most DirectX plugins. It actually
offers >24 bit resolution in most cases. One neat thing about the format is
that you can overdrive samples up to 2x while maintaining 24 bit precision.

Mike

[Catena]

1) Cake is a 32 bit app. It uses some important portions of 16 bit code,
because Win95 drivers and many kernel modules are still 16 bit and better
performance can be got using 16 bit code.
2) The mix engine now works in 32 bit floating point sample format, and also
the DirectShow graph is done in this format. In previous versions, it worked
in 16 bit integers thorugh DirectShow and in 32 bit integers for the mix
accumulation.
3) File and I/O sample format is still 16 bit integers.
4 ) <<How much DSP is done by this mix engine?>> What??? If you call DSP to
do a 16 int to 32 bit fp conversion after reading the samples from file,
doing a multiplication and finally a sum, what's voice recognition?
5) <<What processing is not done in 32-bit and at what bit rate is it
done?>> Bit rate? about ten times your brains throughput, I believe. :-)
J.M.Catena
ad...@sesa.es

1) Cake is a 32 bit app. It uses some important portions of 16 bit code,
because Win95 drivers and many kernel modules are still 16 bit and better
performance can be got using 16 bit code.
2) The mix engine now works in 32 bit floating point sample format, and also
the DirectShow graph is done in this format. In previous versions, it worked
in 16 bit integers thorugh DirectShow and in 32 bit integers for the mix
accumulation.
3) File and I/O sample format is still 16 bit integers.
4 ) <<How much DSP is done by this mix engine?>> What??? If you call DSP to
do a 16 int to 32 bit fp conversion after reading the samples from file,
doing a multiplication and finally a sum, what's voice recognition?
5) <<What processing is not done in 32-bit and at what bit rate is it
done?>> Bit rate? about ten times your brains throughput, I believe. :-)

J.M.Catena
ad...@sesa.es

[Catena]

>offers >24 bit resolution in most cases. One neat thing about the format
is
>that you can overdrive samples up to 2x while maintaining 24 bit precision.

You read it in the Sonic Foundry plugins SDK documentation, eh?
That's wrong. You can overdrive or 'underdrive' as much as you want
maintaining MORE than 24 bits resolution. The 32 bit floats have a 24 bit
mantissa and a 8 bit exponent. That means that you have at least 24 bit
resolution at ANY level. In fact, 24 bits the minimum resolution near the
peaks of any given waveform at any level, but it 'grows' (as the exponent
changes) relative to those peaks as the signal is smaller, as the minimum
'step' can be even much smaller than using 32 bit integers.

J.M.Catena
ad...@sesa.es

[MELVIN MCKEE]

Catena wrote:

> .
> 4 ) <<How much DSP is done by this mix engine?>> What??? If you call DSP to
> do a 16 int to 32 bit fp conversion after reading the samples from file,
> doing a multiplication and finally a sum, what's voice recognition?
>

I admitt that I am probably missing something here, but are you saying that
during offline editing with delay, reverb, ect. that the mix engine is not
involved?

[Catena]

>are you saying that
>during offline editing with delay, reverb, ect. that the mix engine is not
>involved?

I didn't say that, but it's right. An OFFLINE editing is totally separate
from the mix engine, obviously.
Where do you want to go today? <g>

J.M.Catena
ad...@sesa.es