[rescue] Hello?

[Joshua D Boyd]

On Fri, Apr 05, 2002 at 11:32:40PM -0500, Sridhar the POWERful wrote:
> > If the public saw broadcast quality, most of them wouldn't be able to tell
> > the difference.  Just like most people insist that MP3s are just as good.
> > Frankly, with current prices, I think it is getting to be near time for
> > a computer juke box.  Just slap a pair of 60 gig drives in a machine, then
> > leave everything uncompressed.
> 
> With MP3's it depends on the bitrate.  At 320, I have trouble *measuring*
> the difference.  If you feel the need, you can go up to 384.  A 60GB drive
> wouldn't be enough for some of us.  There are people like me, who have
> several hundred GB of MP3 at 320.

With 128 I can put my finger on the differences.  For most songs at 256, I
can't place my fingers on the difference, but I get sick of music a lot 
faster.  I've never tried higher.  I wonder if it is possible to get lossless
compression from the MP3 format.  I also haven't tried Ogg yet.

120gigs wouldn't hold all my audio CDs if they were actually full.  A rough
sampling indicates that they usually only use 400 to 500 megs though.  But,
I'd be pretty close to having 120 gigs completely full.  Still, 180gigs only
costs about $300-400.  I don't know what people do to fill a hundred gigs with
MP3s.  Have they actually listened to all those songs?  By rough estimate, 
that is 71 days worth of non stop music at 128k, or 24 days worth of non stop
music at 384k.  Being generous, that is 768 CDs, if not more.

> > > Granted, there's a lot of cool manipulations that one can do with digital
> > > video, but his point is well put.  The public just doesn't know any better.
> >
> > Of course there are lots of cool manipulations.  The problem isn't the digital
> > (well, it is, but that is a very small part), it is the compression.
> 
> You can lose some quality with digitization, but as long as you use a fast
> enough sampling rate, and big enough samples, you're fine.

Right, but most people don't use big enough or properly scaled samples for
video.  10bit linear is commonly used in video.  It can make nice pictures, 
but it would cause less headaches if it were 10bit log or 16bit linear, or
better yet, it just degraded gracefully.  Ideally, we would do source 
recording in analog, then the digitizing would be a two pass project to
use 16bit log and set the highest value to the highest value present, and the
lowest value to the lowest present, so the full dynamic range would be 
preserved, even if detail is lost on scenes with very high dynamic range.

On a related note, I make a point of writing all my software to use doubles.
This way it is pretty much ready to process stuff that comes from 8bit,
10bit, 16bit, log or linear, without too much detail loss, and then output
in any format.  Maybe I'm delusional about how effective this is...

-- 
Joshua D. Boyd