RE: Re: Changing volume of MP3s

["Mark Harrison" <Mark.Harrison@xxxxxxx>]

Hmmm...

OK just re-read what I wrote, and realised that this really is my
"specialised subject" given that I have a degree in Maths/Comp,
my wife has a Masters in Engineering, with a lot of Signal Processing taken
as advance options, and I used to work as a recording engineer :-)

I'll try this in plainer English :-)


In order to listen to sound, it needs to get converted into electrical
signals in the brain. This happens in a part of the brain connected to the
inner ear. The ear is basically a sense organ that turns vibrations in the
air into electrical impulses in the brain. In order to get the air
vibrating the way we want for sound replay, we apply many years research in
fluid dynamics and discover that the best way to do this is to wobble
around a relatively rigid "thing". This "thing" is
called a speaker cone, and is, in turn moved by virtue of being firmly
affixed to a large magnet - the magnet in turn sits inside a coil, through
which electricity is passed (passing the electricity through the coil
creates a magnetic field which causes the magnet to move, taking the cone
with it.)

The electrical signal applied to a speaker is of the order of up to 50v.
The conventional way to do this is to pass a smaller signal - up to about
2.5v into an amplifier, and rely on the amplifier to convert an input of 0
to an output of 0, an input of 2.5 into an output of 50, and everything in
between in proportion. This isn't as easy as it sounds, because the act of
driving a loudspeaker's coil through an electromagnet "pushes
back" and changes the electrical characteristics of the load that
we're trying to drive! Nonetheless, this works with different levels of :
ultimate output power, exactness of this scaling, and ability to drive more
"difficult" loads - as we move up and down the price ladder.

In the old days (TM), this input signal of up to 2.5v was gained by reading
some continually-varying thing, such as the alignment of molecules in a
magentised medium (reading from tape), or the difference in
inclination/depth of a physical groove in a plastic platter (reading from a
record), with a little bit of aplification to get the answer up to 0-2.5v

"Continually-varying" is another way of saying
"analogue".

These days, signals are transferred as "digital signals" - a
strea, of 1s and 0s, with some agreed way of determining how to convert
them back into a continually-varying voltage.

There are two disavantages to digital, compared to analogue, and one
massive advantage.

The first disadvantage is that there's no continuous-ness in the voltage.
In analogue we could have 0v, and 1v, but also .01, .0123, .01232, .012321,
.0123214, .01232138 v - all distinguished from each other (sort of.) In
digital, there were only a certain number of "steps" you could be
on between 0 and 2.5, and, say, there was NO WAY of being mid-way between a
step, and therefore no way of telling the differnece between .0123214 and
.01232138 - both got rounded to the same step. This is called
"quantisation". In practice, this wasn't as bad as it might sound
- because once you've converted into this voltage, you discover that your
components aren't perfect and the "noise" - ie the random
fluctuation by - introduced was probably bigger than the differences that
subtle. In 1980, when the CD was coming onto the scene, precision in
manufacture was such that 16 bits of data was enough to guarantee that the
data-precision wasn't the limiting factor. These days, 24 bits seems to be
en
ough.

The second disadvantage was similar, but in the time-domain rather than the
voltage-domain. You can't take a running picture in digital. You have to
take a series of snapshots. This process is called "sampling". In
1980, it was felt that 44,100 of these per second was enough, and given
contemparous sound technology, it probably was. These days, 192,000 is
better... The problem is that you need a sample rate TWICE as high as the
maximum frequency that you want to replay. This means that, on CD, the
highest frequency playable is 22kz - which is just within the hearing range
of some people, and most people can tell the difference when nothing above
it is there.

However, the big ADVANTAGE of digital is that it's (compared to analogue)
very easy to make perfect copies. You just have to ensure that you gets the
1s and 0s in the right order, and you can check by comparing back. In
analogue it turns out to be an immutable law of physics that you _can't_
make a perfect copy.

What "normal" playback equipment does to amplify to _your_ choice
of volume is to take this stream of 1s and 0s, then convert them into a
voltage between 0 and 2.5, then amplify that as above. The problem is that
the "big voltages" cause magnetic currents WITHIN the amplifier
which can cause subtle corruption of the low-level input signals.

What "high-end" stuff does is takes the 0-2.5, and
"pre-amplifies" it to between 0-2.5v (ie - the same range, but up
or down as per YOUR choice of volume.) Then it passes it to a separate
"power amp" - normally in a separate box, which does the
conversion from 0-2.5 to 0-50, but at a fixed gain. Turns out this makes it
easier to isolate the interference, so better sound quality.

What high-end playback equipment does to amplify to _your_ choice of volume
is take these DIGITAL signals - the stream of 1s and 0s, and do some
calculating based on your choice of volume to come out with the very
digital stream that, when passed into the "digital to analogue"
convertor will produce the output signal INCLUDING the preamplification.
This is then passed into a power amp (again - in a separate box.)

What this MP3 tool is doing is that digital trick from high-end equipment.
To take the numpty ripping (or for that matter, original recording) that
the MP3 file arrives with, and recalculate what it should be so it
"sounds" at the same level as every other MP3 file you've got.
(With the proviso that, on albums, some tracks might be intended to be
quieter than others, so it takes this into account.)

This digital conversion process ISN'T loss-less. The act of having to force
the output back into 44.1Kz, 16-bit (or whatever) means that you introduce
a whole other set of quantisation errors by doing the calculation.

HOWEVER, this level of error is less audible than you'd get by converting
to analogue, amplifying up/down, and converting back to digital... Hence I
described their description as "cute."

M.



-----Original Message-----
From: andy_powell_is [mailto:ukha@xxxxxxx]
Sent: 25 May 2002 10:15
To: ukha_d@xxxxxxx
Subject: [ukha_d] Re: Changing volume of MP3s


--- In ukha_d@y..., "Mark Harrison" <Mark.Harrison@e...>
wrote:
> Andy,
> 
> Good call! I've been wanting something to do this for a while...

:D

> 
> They're being a little cute saying it's "lossless". What
they 
actually mean is that they're doing amplification upstream in the 
digitial domain (presumably at a high resolution, and then quantising 
back, a la Meridian.) Not technically lossless in the sense that MLP 
is, but "free from analogue conversion noise".
> 
> M.

I think I speak a different language... I don't understand this at 
all :(

Andy





For more information: http://www.automatedhome.co.uk
Post message: ukha_d@xxxxxxx 
Subscribe:  ukha_d-subscribe@xxxxxxx 
Unsubscribe:  ukha_d-unsubscribe@xxxxxxx 
List owner:  ukha_d-owner@xxxxxxx 

Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/



________________________________________________________________________
This email has been scanned for all viruses by the MessageLabs SkyScan
service. For more information on a proactive anti-virus service working
around the clock, around the globe, visit http://www.messagelabs.com
________________________________________________________________________


For more information: http://www.automatedhome.co.uk
Post message: ukha_d@xxxxxxx 
Subscribe:  ukha_d-subscribe@xxxxxxx 
Unsubscribe:  ukha_d-unsubscribe@xxxxxxx 
List owner:  ukha_d-owner@xxxxxxx

Your use of Yahoo! Groups is subject to the Yahoo! Terms of Service.