Compressing video and audio to save disk space - some insights

Lostgallifreyan wrote in
:

people
SERIOUSLY underestimate how good low bit DEPTH can sound. Currently, all
the emphasis seems to be on lowering the RATE.


That sounds a tad silly, but the end result IS lower bitrates. What I'm
getting at is that people are looking at reducing sample rates and cutoff
freqencies when using codecs. The simple fact that lower level sound
compresses better is usually overlooked. Most people want to max out the
power of rock music and many other sounds these days, and it mocks the whole
idea of increasing dynamic range. So why not deliberately reduce the range to
the point where further reduction audibly impairs the sound? For modern
highly compressed rock that makes a lot of sense because the bits allocated
for dynamic range would be far better allocated to frequency range instead.
Same applies to anything that was originally on cassette tape or old vinyl,
where allocating 16 bits (CD quality) makes no sense because they were
limited to around 48 dB range and easily handled with 8 bits per sample.

In message ,
Lostgallifreyan writes:
[]
gains more than it loses. I hate when a big file sounds bad, it might as well
be small.
Agreed!

I used to think differently but after slowly growing a collection over ten
years, nothing beats doing it this way now. I stick with lossless methods for
disk-based copies of my CD's though. And anything coded with high bitrate MP3
or MPC tends to stay as I got it or made it.

I too tend to keep it - though if I notice something _very_ bad, I may
redo it (though more likely to abandon it or look for a better recording
elsewhere).
--
J. P. Gilliver. UMRA: 1960/1985 MB++G.5AL-IS-P--Ch++(p)Ar@T0H+Sh0!:`)DNAf

.... the older [studio] has a fixed position camera at five foot ten. I'm five
foot four-and-three-quarters, so I have to stand on a box, or I'd never reach
the Hebrides. Helen Young (BBC TV weather presenter), 10/2000.

In message , Bill in Co
writes:
Lostgallifreyan wrote:
[]
Even if a bad-sounding 128 or 192 source exists I still don't mind doing
this
at risk of slight further loss, as 80 kbps for joint stereo and 48 for
mono
gains more than it loses. I hate when a big file sounds bad, it might as
well
be small.

The resulting files would still be a bit too big for me to store on my
portable player at those bitrates (which I'm trying to cram as much as I can
on). :-)

Out of curiosity, why are you trying to do that? For a portable player,
unless it's _very_ small in memory, even with modern cells, you're not
going to get anything like enough play time out of the battery to get
anywhere near playing it all, even at the top bit rates anyone uses.

But this is again missing the point I was trying to make. Believe it or
not, you can go all the way down to 32 kbps with *Joint Stereo* (2 channels)
using WMA (but NOT mp3), and still get very good results! And that is
really nice for old radio broadcasts (which have music), if you are trying
to fit a whole slew of these on a portable flash based audio player (I'm
talking HOURS and HOURS of material here)

At which point one would ask, why would I use Joint Stereo (instead of mono)

(What does the "joint" mean in this context [or rather "Joint"]?)
[]
--
J. P. Gilliver. UMRA: 1960/1985 MB++G.5AL-IS-P--Ch++(p)Ar@T0H+Sh0!:`)DNAf

.... the older [studio] has a fixed position camera at five foot ten. I'm five
foot four-and-three-quarters, so I have to stand on a box, or I'd never reach
the Hebrides. Helen Young (BBC TV weather presenter), 10/2000.

In message , Bill in Co
writes:
J. P. Gilliver (John) wrote:
[]
By number of tracks, I mean stereo or mono: sure, encoding mono material
to a stereo mp3 bitstream _should_ not require more bits, but it does
seem to. So if you know it's mono, encode it as mono.

It does if it makes two tracks out of one, even if they are identical,
naturally.
IOW, two tracks should use twice the bitrate as a single track mono file to
get the same sound, since there is twice as much information being encoded,
regardless of whether or not its identical information (which it would be,
in going from single track mono to dual track mono)

I was under the impression that most compression CoDecs, when
compressing a stereo signal, took at least some account of the
difference (or rather similarity) of the two channels, producing
something that required fewer bits if the tracks were similar, than it
would if they were two completely unrelated tracks. Is that not so?
[]
--
J. P. Gilliver. UMRA: 1960/1985 MB++G.5AL-IS-P--Ch++(p)Ar@T0H+Sh0!:`)DNAf

.... the older [studio] has a fixed position camera at five foot ten. I'm five
foot four-and-three-quarters, so I have to stand on a box, or I'd never reach
the Hebrides. Helen Young (BBC TV weather presenter), 10/2000.

"J. P. Gilliver (John)" wrote in
:

(What does the "joint" mean in this context [or rather "Joint"]?)


It's a mix of two forms of stereo, it switches between them in an effort to
select the type most likely to sound good while reducing the data needed to
represent stereo. The two types are standard stereo (2 channels) and mid/side
stereo, which can reduce to the equivalent of mono during any instant where
the 2 channels are the same, which 2-channel stereo can't do. I don't think
one form can be converted perfectly to the other, if it could then mid/side
would be best as it allows smaller storage and more freedom in editing
apparent separation after recording by adjusting only the amplitude of the
mid part. But using it wherever possible allows the best compression.

"J. P. Gilliver (John)" wrote in
:

(though more likely to abandon it or look for a better recording elsewhere)

I always try to to do that. Used to, anyway, these days I tend to just
wait rather than go hunting. For example, the BBC now do Radio 4 Extra, and
it's likely that many things will be rebroadcast.

I very much doubt they care if we copy them either. By their own admission,
they're almost desparately keen to recover stuff they lost, by getting it
from anyone who had the foresight to record it when originally broadcast. It
would be extremely poor public relations if they went after anyone willing to
help archive stuff for any purpose, given their own dependency on it. One way
or another, the best of whatever persists from the last decades will be
openly available to the point where the BBC would likely be glad if they knew
they didn't have to pay to distribute it anymore.

Ok, so maybe they have lawyers who would beg to differ but if so, that's
stupid.

"J. P. Gilliver (John)" wrote in
:

I was under the impression that most compression CoDecs, when
compressing a stereo signal, took at least some account of the
difference (or rather similarity) of the two channels, producing
something that required fewer bits if the tracks were similar, than it
would if they were two completely unrelated tracks. Is that not so?


It is. I just posted about that. Mid/side, used as part of joint stereo...
There's more to it than I know, I can't remember exactly why true stereo and
mid/side aren't perfectly interchangeable, or how codecs decide when to
switch between the two.

Lostgallifreyan wrote:
"Bill in Co" wrote in
:

Why would one do that for old material? To add some stereo effect,
that's why. :-)


When I mentioned processing the sound soem time back, you resisted the
idea,
and of all plugins that do anythign other than restore sound.

Yes, in the context of special effects, NOT noise reduction. OK, I would
add pseudostereo in to the mix, which is a special effect, I suppose. But I
don't put it quite in the same category as adding flanging, etc. :-)

This came up
when I mentioned using NR and HR (noise reduction, harmonic regeneration)
to
increase clarity. I'd much rather do that than add stereo.

Noise reduction, ABSOLUTELY. And I've mentioned I've done that before as
part of my restoration work. Harmonic regeneration, not so much.

The extra bit
requirement is likely smaller because you don't need that much extra HF,
just
enough to repair obvious deficiency in old recordings and broadcasts,
after
reducing the backgound noise. I'm sure stereo is nice, but increased
clarity
is the very finest ear candy, I kid you not. The only time I adjust stereo
is
if the sound already IS stereo, and is either too narrow, or was that
horrible Phase 4 recording method, in which case (purely as experiment on
a
couple of tracks) I used an Acoustic Mirror reverb to make the sound feel
like it came out of a natural room instead of a diabolical excercise in
panning that had the subtlety of an Apollo thruster control.

I've played with that too but to me it destroys the music. It's fine if you
want to create a sci-fi music file, I guess.

Reducing noise

Yes! Agreed.

and boosting harmonics

Nah, I'll pass on that one.

is THE single best restorer I found. Not that there's
any program designed to do it all in one go, as far as I know. There
should
be. It would be like Dolby in that it could restore an original after
storage
or distribution on a low bandwidth medium. I don't think any codec has yet
standardised a method like that.

Using PSP pseudostereo with the right settings, you can get a good stereo
effect that is reversible, if you want (i.e., if you don't like it later,
you can recombine the 2 channels into mono, and it's fine; that is NOT true
for all pseudostereo plug-ins, however. Some are, and some are not, mono
compatible (i.e., without having consequent artifacts when being recombined)

J. P. Gilliver (John) wrote:
In message , Bill in Co
writes:
Lostgallifreyan wrote:
[]
Even if a bad-sounding 128 or 192 source exists I still don't mind doing
this
at risk of slight further loss, as 80 kbps for joint stereo and 48 for
mono
gains more than it loses. I hate when a big file sounds bad, it might as
well
be small.

The resulting files would still be a bit too big for me to store on my
portable player at those bitrates (which I'm trying to cram as much as I
can
on). :-)

Out of curiosity, why are you trying to do that? For a portable player,
unless it's _very_ small in memory, even with modern cells, you're not
going to get anything like enough play time out of the battery to get
anywhere near playing it all, even at the top bit rates anyone uses.

I can get up to 50 hours on one AA cell on my old Cowen iAudio 2 GB player.

But this is again missing the point I was trying to make. Believe it or
not, you can go all the way down to 32 kbps with *Joint Stereo* (2
channels)
using WMA (but NOT mp3), and still get very good results! And that is
really nice for old radio broadcasts (which have music), if you are
trying
to fit a whole slew of these on a portable flash based audio player (I'm
talking HOURS and HOURS of material here)

At which point one would ask, why would I use Joint Stereo (instead of
mono)

(What does the "joint" mean in this context [or rather "Joint"]?)
[]

"Joint Stereo" is a superior method (i.e. makes more efficient use of
compression) when used at these lower bitrates, as it only encodes the sum
and difference channel information, instead of encoding each L and R channel
independently at full bitrate. Why is this? Because the difference
channel information is really small, and doesn't therefore require a high
bitrate to encode it, so that more bits can be used to encode the sum
channel information. Of course, in the end, at the receiver, both channels
are recombined to get back L and R.

Something a bit similar to this was done for FM stereo broadcasts, but for a
different reason (mono compatibility with old mono receivers back in 1960)

Think of it this way, using a 128 kbps stereo file as an example:

If there were NO difference channel information (meaning identical
channels), (which is the extreme case), a full 128 kbps could be used for
the sum channel, instead of splitting it in half for each channel.

IF you use "Stereo" instead of "Joint Stereo", however, each channel gets
half the bitrate, which is a bit limiting.

Or, to put it another way, using stereo and not joint stereo encoding:
A 128 kbps STEREO file would have the same sound quality as a 64 kbps MONO
file, when played back. (That's if one uses Stereo, not Joint Stereo).

Lostgallifreyan wrote:
"J. P. Gilliver (John)" wrote in
:

(What does the "joint" mean in this context [or rather "Joint"]?)


It's a mix of two forms of stereo, it switches between them in an effort
to
select the type most likely to sound good while reducing the data needed
to
represent stereo. The two types are standard stereo (2 channels) and
mid/side
stereo, which can reduce to the equivalent of mono during any instant
where
the 2 channels are the same, which 2-channel stereo can't do. I don't
think
one form can be converted perfectly to the other, if it could then
mid/side
would be best as it allows smaller storage and more freedom in editing
apparent separation after recording by adjusting only the amplitude of the
mid part. But using it wherever possible allows the best compression.

I can convert a stereo file to joint stereo or vice versa as needbe. But
the true advantage of Joint Stereo over Stereo, *especially important at low
bitrates*, is the much more efficient encoding of the information, as I
explained in my just posted response.

Again, this is true since most of the information in any stereo music is
present (and therefore will be encoded) into the L+R channel, and not the
other L-R "difference information" channel (which will get fewer bits
allocated to it), leaving an effectively increased bitrate available for the
L+R channel). Of course, in the end, both are recombined back to get the L
and R channels.