naturerecordists
[Top] [All Lists]

Re: Digital Distortion

Subject: Re: Digital Distortion
From: Klas Strandberg <>
Date: Sun, 22 Jun 2003 19:33:17 +0200
It makes sense. Thank you.

It is not only that analog changes might be "musical" (which I think is
true, many times) but also that I can predict the errors of the analog
changes I make. I can understand both what I am doing and - perhaps - how to
avoid it.
With the digital processing I don't understand anything. 

Of course one must use ones ears, I agree to 100%. But sometimes it is
difficult with digital equipment.

An example: I had a 5 minute long wav-file. I wanted to test which MP3
conversion that worked best. But to convert a 5 minute wav file into a good
MP3 may take 15 minutes!

So I took 20 sec of the 5 minute file, saved it as a wav and then made 3
different possible conversions into MP3. Then I listened to them all. After
listening, I could descide to use "X" conversion.

Then I used "X" MP3 converion on the 5 minute original file, listened to the
first 20 sec to check that everything was fine, and saved it. And - of
course - deleted the original wav file, as I had it on tape anyway, and
didn't want to waist harddisc memory.

The day after, by accident more or less, - I happened to listen to the
entire MP3 file. The bird itseld was good, just as I hade checked, but some
bypassing thrushes some -20 db below, sounded just terrible!

Which means that if you do something digital, you must check the entire
file, as some part of it may be very good, while another part may sound bad,
totally unexpected!

Thans again,

Klas.





>At 21:58 2003-06-21 -0000, you wrote:
>Klas,
>
>There are a lot of distortions during digital processing, IF you 
>consider each and every unexpected change of the signal distortion.
>I do. 
>
>When I boost a signal with 0.33 dB I want the signal stronger; I 
>realise that the noisefloor will also be boosted with 0.33 dB, after 
>all the dynamics are expected to remain the same.
>
>But in digital there is always rounding of the bits, so each volume 
>change distorts the signal, whether this is audible is not a fact, it 
>depends on the kind of change and on the number of changes applied to 
>the digital signal.
>
>I do know that as a rule of thumb, most unwanted analog distortions 
>are far likely to be musical, or audible pleasent. And at the same 
>time most digital distortions clearly stand out of the original data, 
>thus creating unmusical, and unpleasent audible parts.
>
>Sometimes digital processing is done in a way that is in theory 
>better then the old fashioned analogue way, but is that always more 
>musical? I know that all analogue EQ provides both a volume change 
>and a phase change. Most digital EQ leave out the phase change. This 
>way the whole EQ-ing is different from what you know, what you are 
>used to, and what you expect. 
>But is it wrong? I think each person must judge for his own what to 
>do, but I have heard that it is better to use as little of the same 
>kind of processing as possible: it is better to do 1 volume change of 
>30 dB instead of doing 100 volume changes of 0.3 dB. (Because of the 
>digital rounding).
>
>There must be many more unknown and unexpected things that digital 
>processing can do to a signal that one have never even thought about.
>Digital is in no way a perfect solution, neither is analogue.
>
>I must say that the above is my personal opinion :)
>Not hard facts. I cannot prove it; maybe some-one else can?
>
>Regards, Evert 
>who studied a little* DSP technigue during my first 30 years of my 
>life.
>
>* In fact I studied it a lot, even was a teacher, but that only means 
>I know it from books, not from the real world :)
>
>--- In  Klas Strandberg 
><> wrote:
>> 
>> Let me clearify the question:
>> 
>> Will someone claim that digital processing is 100% perfect? At all 
>levels,
>> all time durations (transient response) - all frequenicies? All 
>mixes of
>> sounds? No interence, ever? 
>> 
>> I don't think so. Not being an expert on digital processing math, I 
>guess
>> that nobody is prepared to say it is "perfect".
>> 
>> So, if not perfect -  what errors are there? That's the question. 
>What
>> errors are "natural" and logical for a digital process when I, say, 
>apply a
>> 10 kHz 12 db/octave high pass filter and then normalize the output 
>to -6db
>> peak??
>> 
>> We know all about analog distortion of all kinds, but what about 
>errors in
>> digital processing? It seems to me as there arn't any!?
>> 
>> I think some digital recordings "itch" in a strange way, and I feel 
>that
>> they itch more and more, the more steps I do, digital processing.
>> 
>> Suppose I use a lot of different digital processing. Filters and 
>dynamic
>> processors? Suppose the finished product is digitally manipulated 
>20 times?
>> Then the question is: What errors are "natural" and "logical" in 
>such a
>> chain of 20 links of digital manipulation?
>> 
>> I am almost ready to abandon all analog units which I use, - 
>filters, EQ's,
>> noise gates etc, in favour of digital software. It is sooo much more
>> practical, having all of it in my PC. But should I? 
>> 
>> I know all about the danger pressing analog gears too much, but 
>noone seems
>> to talk about the limits for digital processing.
>> 
>> Arn't there any?
>> 
>> Klas.
>> 
>> 
>> 
>> 
>> 
>> 
>> 
>> 
>> 
>> 
>>  
>>  
>> Telinga Microphones, Botarbo,
>> S-748 96 Tobo, Sweden.
>> Phone & fax int + 295 310 01
>> email: 
>>        
>
>
>
>To unsubscribe from this group, send an email to:
>
>
> 
>
>Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/ 
>
>
>
Telinga Microphones, Botarbo,
S-748 96 Tobo, Sweden.
Phone & fax int + 295 310 01
email: 
       




________________________________________________________________________
________________________________________________________________________

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

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