[Top] [All Lists]

Re: Re: Digital Distortion

Subject: Re: Re: Digital Distortion
From: Walter Knapp <>
Date: Mon, 23 Jun 2003 13:48:58 -0400
> In a message dated 6/20/03 10:48:55 PM Pacific Daylight Time, 
>  writes:
>>There's still a lot of recording advice floating 
>>around that's straight from analog recording and ruins digital 
>>recordings. We need to root that out.
> Dear Walt,
> We (you) have touched on several of these before. But, it would be helpful if 
> you could reiterate then in one message for those of us who need help!

I'm sorry I've not had the time on this. Keep tripping over something 
missing on my system to dig out of backups, download, etc. And I'm also 
busy stuffing my new Palm Tungsten C. Up around 200 frog photos in it so 
far, and plenty to go.

The biggest focus is on how we set level readings on a recorder (or any 
other place in the chain where we can set them). Standard analog 
practice was to set the levels so that they only occasionally hit the 
0dB mark, or sometimes even more aggressive than that. When the signal 
clipped in analog, it would mostly just limit volume or dynamic range, 
but still be pretty list enable. And analog had a lot less dynamic range 
than digital, so it was necessary to do this to get acceptable dynamic 

Digital, on the other hand is not at all forgiving of anything over the 
top, even a single wave cycle. So setting the recorder the same way as 
analog resulted in lots of clipped off wavetops. When these are fed 
through the output to convert back to analog to play, the output 
circuitry sees all kinds of changes in the waveform that it interprets 
as other frequencies. And generally loud. Thus you get all sorts of 
output sound that was not there. On top of whatever error the input side 
did with the overlimit waves.

So, most people learn to back off the level indicated to avoid hitting 
the 0dB level, often using the peak functions to try and find the level 
of the peaks. This can be done and still have plenty of dynamic range in 
digital because you have more dynamic range to work with, a extra 30dB 
or more, compared to analog.

Unfortunately, you can run into the limits of metering. A metering 
system is averaging, for nearly all even the peak indications are a 
average. The true highest wave may be still higher. And even one wave 
clipped can mess up the sound. This is where experience comes in. You 
have to use the recorder in different situations enough to know just how 
far off the indication is, and then set a pad to cover that, as well as 
any unexpected loud sound that you might want.

It's this second sort of problem that it's common for people to not 
catch. A few wavetops may not sound all that bad clipped, but they do 
change the sound. If you are not sure this is happening with your 
recording, get it in waveform display and expand it enough to see the 
individual waves. Then scroll through the sound and look for clipped 
waves. Naturally the most likely place is during the loudest parts.

One should also note that not all equipment clips at a indicated 0dB, 
you have to figure it out for your equipment. Some have a built in pad, 
and a combination system may have still a different pad. Stick in a 
preamp and you may clip when the preamp is not near indicating full. You 
have to look at every stage in your system leading up to where it's 
converted from analog to digital.

As I've noted, it's not always possible to eliminate all clipping. You 
may need to get good recordings of all species present, and the only way 
to get the quiet ones is to tolerate clipping the high intensity ones. 
This is a common problem in surveying mixed frog chorus sites. As a 
example, frogs like cricket frogs put out very high energy, short 
"clicks". Recording what else is there will very often clip the cricket 
frog sounds. (and those "clicks" are very bad for not being indicated 
properly on the meter) It's not that clipping should never happen, but 
it's that you should decide when it does.

Now, this is just the most common problem moving from digital to analog. 
There are other changes down the recording or processing chain. But 
start at the input, learn to get that right first. Then fuss over the 
others. We can get into them another time.



<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