[Top] [All Lists]

Re: Digital Distortion

Subject: Re: Digital Distortion
From: Walter Knapp <>
Date: Wed, 25 Jun 2003 11:27:41 -0400
Klas Strandberg wrote:
> Yes, that's what I mean, Walt. There is no meaning in setting a fixed
> 0-level, but rather a "fitting" level.
> But I still might need to change the level!
> Comparison: I can have a .tif picture. I can change the size of it (bigge=
> OR smaller) and clearly see how the sharpness go away. I can "sharpen" it=
> and clearly see how tiny white lines grow around the egdes. For example.=

> Is that what happens to my sound files too?? More or less?

My impression is that changing gain levels in a soundfile is essentially
lossless, though you will get a small roundoff error in the
calculations. As long as doing so does not push some part of the sound
over (or under) the limit. There are changes, but they are often a
result of our sound reproduction system we play it with.

When I first started using computers I was used to getting 9 significant
places in my calculations using a HP calculator. I found I got different
results in a computer with the same calculation. Back then I learned
that computers were good, if lucky, to about three significant places.
Even today, to get precision carry of significant places in a computer
you have to do fancy programming. Using greater bit depth in
calculations for sound helps, and this is where it's more important to
use 24 bit or whatever, not when we record.

There is a tradeoff between speed and precision in a computer. Sound, by
the nature of the size of it's files encourages programming for speed.

BTW, I have found another parallel between graphic processing and sound
processing. Often the best results are obtained by applying some effect
lightly numerous times instead of one massive hit. A gentle hand in
sound filtering can often get you more. The same thing may also apply to
a entire loop of sound filters.

> I might get provocative now, but it feels like these matters are terribly
> basic! In the analog world it's "beginners level"! I'm actually a bit
> chocked that there seems to be no answer in the digital world. Or is it j=
> too complicated?

If you try to understand every step the programmer put into your sound
programming, it gets very complicated. And often the requirements of the
computer mean it's not being done the way that simple math formulas
would have you believe. Focus on the sound, that's what counts. It does
help with computers to do a lot of what if playing. Try things, try
variations of things, even when you don't have a goal. Nearly every time
I listen to some new soundfile I'll piddle around with it, even though I
have no intention of changing it. That's how I'm learning this stuff.

One problem folks have is the worship of that bit about digital being
bit perfect copies. Yes, if just copying a file, it's possible to
approach bit perfect copies. But, in sound processing, each processing
step is calculating a new waveform. You don't care as much that it's bit
perfect as that it approaches sound perfect. Do not get worried that the
sound changed, just make sure it changed the way you want. After all, if
you were doing it analog, every component it passed through would be
changing the waveform.

> Or can I be sure that the digital errors are always below audible level?

No, you can not. You must listen to the sound. Note as I said above,
defining any change in the digital world as a "error" is part of the
problem. You loose focus on the sound. You don't really care what
happened in the "black box" of the filter, you care what the input and
output are like.

> Digital recording is by all means better than analog recording! That's no=
> the issue here. The issue is if it is better to go analog with all change=
> and let the sound card only create the final digits?

I'm of the opinion that it's time to move away from analog in sound
processing. Digital can do things that analog has no hope of doing. With
time you can become just as familiar with what your digital stuff can do.

One part of this in the pro world is that many have trained their ears
to like the particular errors of the analog processors. So, when digital
does not make the same errors they think something is wrong. That analog
got it right and digital did not. Analog messed it up one way, digital
messed it up another way is more the truth.

> The issue is wheather I can trust a WaveLab software install (looking lik=
> hardware on my screen) to do the job as good as an analog likewise?

I don't know the WaveLab software, so can't comment on trusting it.

I'm not a fan of the user interface that makes a digital filter look
like the analog counterpart. This reinforces the idea that it should
make the same sound as the analog. People never learn to live with what
it does then. You learned what twiddling a knob on a analog setup did,
twiddle the settings in a filter the same way to learn what it does.

> My son asked me the other day that "why do you care about those things, d=
> people won't hear it anyway because of the computer fan..."

Or the cheap headphones, tiny junk speakers, etc. Certainly I despair
about the trend toward poorer and poorer average players. About how few
care anymore.

But, I don't sweat over something because of someone else. It's my own
personal standards that drive me. I constantly ask myself if I did the
best I could.

Chris, before he died once said to me. "You have saddled me with the
curse of quality, I can't just go out and get any old tool, I have to
get a quality tool, and then I have to use it well". He was not
complaining, just stating how it was.

When asked why he climbed mountains, a mountaineer once replied,
"because it's there". We don't need more than that for why we do what we do=



Your use of Yahoo! Groups is subject to

<Prev in Thread] Current Thread [Next in Thread>

The University of NSW School of Computer and Engineering takes no responsibility for the contents of this archive. It is purely a compilation of material sent by many people to the naturerecordists mailing list. It has not been checked for accuracy nor its content verified in any way. If you wish to get material removed from the archive or have other queries about the archive e-mail Andrew Taylor at this address: andrewt@cse.unsw.EDU.AU