[Karl Uppiano]

"Radium" wrote in message
oups.com...
Karl Uppiano wrote:
Digital audio for high fidelity?

Digital audio for any application.

Bzzt! Nope. I will guarantee that your rules will not make any sense for
digital audio for telephony (for example). Or for my application (high
fidelity). Sorry, I would not do that to my FLAC files. My compressed
formats, where unavoidable, consist of WMA, AAC and MP3. I would not import
my files that way.

Why monaural?

Because I want both the L and R channels to sound the same.

Suppose I like stereo?

A sample rate of 44.1 or higher will
give you 20KHz audio bandwidth. That's nice for hi-fi listening, but may
be
more than you need for "You Tube" sound tracks.

Any digital audio requires 44.1 khz or higher in order to sound
pleasant. Aliasing can be a real earsore.

Done right, you can sample at any frequency without aliasing. The sample
rate only affects the bandwidth you can record. While I can understand
wanting full range audio for listening to music, it would be quite
inappropriate, and a big waste of bandiwdth to use 44.1KHz for telephony
(for example).

Okay. What about dither? Does it need to be dithered? I think it needs to
be
dithered at 2/3 LSB (that's my rule).

No need for dither.

Dither eliminates the distortion due to quantization errors present in any
digital system. I feel that there is a need for dither in high quality
applications.

Even if the compressed and uncompressed versions reside in different zip
codes?

Of course. What do zip codes have to do with this?

I was being facetious. There are compressed and uncompressed versions of all
sorts of things all over the world at many different sample rates. They are
not all going to follow your rules. Perhaps I was taking you too literally.

I assume that by this you mean you do not want to reduce the bit rate by
reducing the sample rate, but only by means of bit allocation using a
perceptual coder.

Exactly.

In-phase signals from left and right channels will increase by 6dB when
you
sum them. In order to avoid clipping if left and right channels are full
scale, you would need to reduce the level by 50% You said reduce *by*
77.5%,
so I assume you mean drop the level *to* 22.5%.

You assume correctly.

If it was mono you wanted, you had it at step 4. If you really wanted
(0.725R - 0.275L), you could have done that all in four steps: Reduce the
right channel by 72.5%, reduce the left channel by 27.5%, flip the phase
on
the left channel, and convert to mono. Try that and see if you don't get
the
identical results you got with your 14-step plan.

The audio that was in the center channel [lead vocal, bass,
percussions] are too loud while the audio that was in the periphery
[paino, chours, guitar, synth-pads] aren't loud enough.

I understand what you are trying to do. My point was that you were taking a
very complicated approach to arrive at what you describe as your end result.
I further said you could get the same result in far fewer steps.

People have been mixing down to mono from stereo for 50 years or more.
You
simply add the left and right channels. Listening in stereo in a room
actually does more or less the same thing too (left and right speakers
working in phase (panned to center) will sum 6dB higher in the room,
depending on the frequency, and where you're standing). Record producers
mix
the stereo channels for the proper artistic balance in their professional
opinion. Mixing down to mono should not be a problem.

My technique usually ensures that the sounds that were originally in
the central channel are not significantly louder than the sounds that
were originally in the periphery [and visa versa].

I am not convinced that your technique accomplishes that goal. I won't deny
that it will change the sound. It might even sound better to you in certain
limited cases.

I will say it a different way: There are millions of hours of AM, FM and TV
broadcasts that simply sum L + R for mono receivers. Are you saying that
everyone has got it wrong for lo these many years?