Jack Carroll wrote:
On Sun, Apr 09, 2006 at 02:12:28AM -0700, James Richard Tyrer wrote:
Justin Findlay wrote:
On 4/3/06, Timothy Miller <[EMAIL PROTECTED]> wrote:
I know basically nothing about audio technology. But given what
little I do know, here are some things that I think would be
relatively easy to do with OGD1 and, what shall we call it, OAC1:
- 60+ audio channels (pick your combination of in and out)
- (A specialized card could handle lots more channels)
- 24-bit precision per channel
I'd say more than 24, perhaps 32 bits per channel.
What are you going to do with all those bits? 16 bits out is all you
need (96 dB dynamic range).
After mix-down. If you want to end up with 16 bits after mixing a
couple dozen channels and doing all kinds of scaling and filtering, you need
to start out with several bit more resolution than that.
You need an accumulator register that is larger than 16 bits but you
don't need more than 16 bits of data. Typical DSP hardware uses a
accumulator register that is 2 words wide and may have guard bits on the
LSB end and overflow bits on the MSB end. The guard bits have no
function if you use single precision mixer coefficients. So, you
multiply your 16 bit data by a 16 bit positive mixer coefficient and get
a 32 bit result which you add to the accumulator. After you go through
all the streams to be mixed, you output the 16 MSBs of the accumulator
or you might want to check for overflow and implement an AGC function to
eliminate clipping. So, you can see that you won't loose any
significant bits unless you mix at least 16 channels -- this is not
really the loss of significant bits but rather truncation error. To
improve this, you need to use double precision mixer coefficients and
rounding.
First point to consider is that you don't need any more digital range on
the input than your analog input dynamic range.
We don't know what that is, though.
Sorry, that wasn't clear I mean the dynamic range of the analog input
buffer.
--
JRT
_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)
