naturerecordists
[Top] [All Lists]

Re: artifact in spectrogram analysis

Subject: Re: artifact in spectrogram analysis
From: "rock_scallop" rock_scallop
Date: Tue Jul 17, 2012 11:18 am ((PDT))
David,

While it is true that audio compression could as you say leave, "white patc=
hes where the background level used to be," MP3 also does that, however by =
a different mechanism.

Here is an article I googled that exlains it pretty well.
http://www.soundonsound.com/sos/may00/articles/mp3.htm
Look down the page under "MP3 Encoding":
"The encoder examines the contents of these frames, and attempts to determi=
ne where masking in both the frequency and time domains will occur, and thu=
s which frames can safely be allowed to distort."

If a sound is loud enough to completely perceptibly mask the frequencies be=
low or above it, those frequencies are altered, and in some cases completel=
y deleted - and in this case leaving the white patches. The lower the bitra=
te used, the more coarse and audibly noticeable such artifacts become.
This approach is also common to the Sony ALTRAC compression scheme.

John Hartog
rockscallop.org




--- In  "Avocet" <> wrote:
>
> Oh dear, and I've been trying to unravel muddled thinking in this
> thread.
>
> "Compression" means both audio volume compression where the gain is
> varied according to the input level, and a way of reducing the bit
> count on a digital recording (digital compression). Totally different
> from each other.
>
> The original recording is highly audio compressed which is a complete
> explanation to what is heard in the original recording and what is
> seen in the spectrogram. Doubters please believe a pair of ears which
> have professionally listened to his stuff for decades.
>
> The white patches on the spectrogram are when the parts of the
> spectrum which are at high level, and audio compression from these
> peaks has reduced the entire volume level leaving white patches where
> the background level used to be. End of. Tha's what spectrograms do.
>
> Digital compression is a quite different techique. WAV format digital
> compression compresses the digital version down to 16 or 24 bits. This
> compression is inaudible and we tend to regard WAV as "uncompressed"
> which it is not. Just so good that it is inaudible.
>
> MP3 and other types of digital compression offer a wide range of
> qualities. A bit rate of 320kb/s is very high and in practice
> indistiguishable from WAV digital compression. You can test this by
> subtracting a WAV compressed file from an MP3 version by sutracting
> one track from the other. The difference is lower than than a good
> background noise level and as such is inaudible.
>
> BBC radio MP3 podcasts are digitally compressed at 64kb/s which they
> regard as acceptable for speech, but it does sound a bit crunchy with
> audible artifacts and a "reverberation" effect. Check this out by the
> subtraction test between a WAV version of a good recording and a
> 64kb/s MP3 version..
>
> The one thing which is easy to refute is that MP3 digital compression
> does not alter the amplitude levels which is what is clearly happening
> on the original recording. Even people with cloth ears can hear the
> background noise jumping down and back up.
>
> As far as spectrogram displays go, even high definition ones will not
> show up any artifacts the ear cannot hear. If you want to check this
> out try audio compressing a random noise like wind in trees at 32kb/s
> or 24kn/s. The artifacts sound like running water.
>
> > Is clipping the same thing as aliasing (referred to in the Raven
> > manual)? If so, then I know what you're referring to, but if not
> > then I'm still not sure what exactly that is.
>
> No. Clipping is distortion of the peaks of the waveform by clipping
> them off and will appear to reduce the volume level because most
> meters only measure the peaks. Clipping makes any clipped recording
> sounds rough and it cannot be undone. The answer is to record at a
> lower level, nothing else.
>
> Technical note: All digitisers clip or audio compress the input level
> just below 100% to avoid the digital signal form turning "inside out".
> 100% plus just 1 bit digitises as 0% plus 1 bit.
>
> To be boringly repetitive, digital recording is so good, bringing up
> the volume later will not produce more noise as tape recording would.
> Old habits die hard.
>
> David
>
> David Brinicombe
> North Devon, UK
> Cogito cogito ergo cogito sum - Ambrose Bierce
>











"While a picture is worth a thousand words, a
sound is worth a thousand pictures." R. Murray Schafer via Bernie Krause.



------------------------------------------------------------------------
Yahoo! Groups Links

<*> To visit your group on the web, go to:
    http://groups.yahoo.com/group/naturerecordists/

<*> Your email settings:
    Digest Email  | Traditional

<*> To change settings online go to:
    http://groups.yahoo.com/group/naturerecordists/join
    (Yahoo! ID required)

<*> To change settings via email:
    
    

<*> To unsubscribe from this group, send an email to:
    

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

------------------------------------------------------------------------

<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