clipping \ digital distortion ?

Omb · Post by **Omb** » Sat Dec 13, 2003 6:14 pm

hello geeks,
all i know is that clliping is when the sound is over 0db and it can couse distortion if i do that ( correct me if im wrong ) .
now,
i have heard here and there that in digital processing or in some pluginns that i use with SFP or SX (like compressor etc.)i cannot get this clliping and this is a real advantage .
1.is clliping is another word for distortion ?
2.what is clliping ?
3.what is digital clipping ?
4.what the plugin do that prevent this clliping or how does it do that ?

i'm realy confused on this issue so answers or links for explanation will be welcomed .

have a nice day where ever you are on this planet .

Omb

<font size=-1>[ This Message was edited by: Omb on 2003-12-13 18:30 ]</font>

astroman · Post by **astroman** » Sat Dec 13, 2003 7:20 pm

On 2003-12-13 18:14, Omb wrote:
1.is clliping is another word for distortion ?
2.what is clliping ?
3.what is digital clipping ?
4.what the plugin do that prevent this clliping or how does it do that ?

1 & 2:
analog clippling of signals (imagine a sinewave for simplicity) is achieved by 'cutting off' the peaks of the waveform. You first amplify a signal but then limit the entry level of the next stage.

The resulting sound is perceived as 'distorted', the original sine almost becomes a squarewave if you cut off enough.
That why it's called 'Rectifier' for guitars.

There are different technical ways to do it with various results, but all sound rather 'good' in some way. The analog clipping produces additional harmonics which are characteristic for the sound of the unit.

3. digital clipping is a completely different thing. There's also something 'cut off', but this time it's a number.
Which is of course impossible - acoustically.

If you count from 1 to 10 that's straight forward til 9. Then you need another decimal place.
Same with the encoding of the analog signal in the computer, up to a certain level.

But WHAT happens if your signal crosses the invisible border of the maximum number which can be encoded ?
It doesn't stay at max but instead causes an overflow, which effectively drops the signal from max to zero within one sample

That's why it sounds so terrible and is to be avoided under all circumstances.

4. there are plugins which simply look ahead in the audiostream before it gets processed and either limit the signal or compress it, so it doesn't 'cross the border'.
Often this is combined with the simulation of analog circuits, tapes or tubes as those units (see above) sound pleasing to our ears.

cheers, Tom

Omb · Post by **Omb** » Thu Feb 26, 2004 5:58 am

its been a while but i finely got it

cheers
Omb

interloper · Post by **interloper** » Thu Feb 26, 2004 11:09 am

I'd also like to add to Tom's excellent explanation, that you can probably get away with clips that have a very short transient time signature. Some people, while normalizing tracks for cd masters use the change gain function and purposely clip some of the signal. It should be noted that these transient event don't last more than a few milliseconds, thereby not producing audible artifacts.

Immanuel · Post by **Immanuel** » Thu Feb 26, 2004 11:13 am

And it is said (I have not tested it), that a source like drums are less sensitive to a few "overs", that i.e. a piano.

interloper · Post by **interloper** » Thu Feb 26, 2004 4:06 pm

Yes, because their transient time signature is very short.

astroman · Post by **astroman** » Thu Feb 26, 2004 4:37 pm

tnx that you liked the text above

maybe some additional words for that special situation when the signal approaches top level:
the result totally depends of the processing and as such a proper 'clip' at max level will distort the waveform (like an analog guitar distortion) and may indeed be acceptable.

But usually 'digital clipping' means that the signal overflows (unprocessed) and flips from top to the reverse direction within one sample.

That is clearly noticable on whatever background, and a few milliseconds equal a gap several dozen samples, followed by an immediate rise with the same one sample transient when the signal gets back into 'regular' values near top level again.

But I guess today most apps are aware of this situation and take proper precaution.

cheers, Tom

Immanuel · Post by **Immanuel** » Thu Feb 26, 2004 4:46 pm

On 2004-02-26 16:37, astroman wrote:

But usually 'digital clipping' means that the signal overflows (unprocessed) and flips from top to the reverse direction within one sample.

For once, I think you got it wrong Tom. I believe Digital Clipping refers to having 2 or more samples after each other at maximum amplitude - be it negative or positive phase. As I read what you write, you say that digital clipping is defined be going from max amplitude +/- phase to max amplitude of the oposit phase within one sample.

This will, however, only acour after MUCH clipping (as I am used to it being defined) has been going on, or if you have a 22k (at 44k sample rate) sine tone a 0dB.

astroman · Post by **astroman** » Thu Feb 26, 2004 5:57 pm

On 2004-02-26 16:46, Immanuel wrote:
...As I read what you write, you say that digital clipping is defined be going from max amplitude +/- phase to max amplitude of the oposit phase within one sample. ...

yes, I refer to the 'classical' context when 24bit recording wasn't yet invented - or applied

When all bits (representing the signal) were used up and the processed value increased, an overflow exception occured and if I remember right the register containing the data was either zeroed or the data was interpreted as minus max value.

Which case applied was depending on the current program's code (the smart ones would just keep the current value, like a limiter), but there must have been lots of stupid (or careless) versions around, so this term of 'digital clipping' got established.

It has a very special sound that can hardly be messed with something else, except some sync problems.
If you manage to convince a CD player to process a data disk: that's the extended version of that sound

cheers, Tom