naturerecordists
[Top] [All Lists]

Re: Some basic processing questions

Subject: Re: Some basic processing questions
From: "Avocet" madl74
Date: Thu Jan 19, 2012 11:47 am ((PST))
> 1. How do I decide how much to increase the volume by?

Peter,

Standard levels are always a matter of debate. In the days of
analogue, you had to minimise background noise like tape hiss, so
recording levels peaked nearer 100%. With digital recording they seem
all over the place, but many recordists aim at peaking to -15dB
to -18dB, leaving that spare headroom. Bearing in mind that your
recordings are most likely to be listened to on an analogue system,
the final volume shouldn't be too low. So that the peaks don't
overload the playback system and to give a margin of safety, I usually
peak to -6 to -3 dB. On Audacity, you set this on the Normalize
window.

> 2. If the above will result in clipping of the loudest parts, what
> should I do about them? Audacity has a Compress function which looks
> like it might help bring them down a little without affecting the
> quieter parts.

Unfortunately we have got too used to compression on broadcasting and
music and regard it a "normal". Compression nake ist sound "louder" at
the same peak levels. Ads are often compressed more than programme
sound. Compression acts on a range of levels, for instance reducing a
24dB range to 12dB (Ratio of 2)

In my mind and ears, compression destroys the naturalness of the
sound, especially wildlife sounds and I can hear it almost every time.
I think it should be avoided with all nature recording, especially in
the first generation where it is unnecessary with digital. Recordings
cannot be decompressed successfully because of timing restraints.

Checking out figures with 16 bit digitising, the first 2 bits are
taken up with "dither" and one is +/-. If you record low, say -18dB
peak (another 3 bits) you have a signal to normal peak noise ratio of
10 bits or 60dB. Playing this back in a quiet domestic 32dBA noise
level, normal peaks are 102dBA above that or well above the threshold
of pain.

In practice actual signal to background ratios are much lower than
this and you are doing well with 40dB, Anything lower than this will
suffer after compression which lowers the signal/noise ratio. With
high background noise, the last thing you want is compression making
it worse. The Audacity "Noise Removal" acts a bit like an expander - a
reverse compressor - but it is not a cure. It's bast to avoid
compression altogether on recording.

Limiting

Limiting is peak level compression only, or compression with a high
compression ratio, often 10:1. Used properly, it should mimic the way
our ears limit loud sounds, for instance thunder and door slams.

Beware of the Audacity "Hard Limiter" which is a clipper, not a
limiter, producing hard distortion. Every digitiser has a clipper as
without it the signal goes negative. Most recorders also have a
limiter which produces a "hole" after a peak. You can't repair this
hole, so it's better to record low.

Compression.

If there is a single or occasional event in an edited recording which
peaks well over the average, you don't want to drop the overall level
for that one sound, so limiting that mimics the natural ear is called
for. On the Audacity compressor, set the ratio to 10:1 and the
threshold to something like -6dB to -12dB, Check it out by listening.

With any compression, there is a recovery time after the compressor
has acted, confusingly called "Attack" time in Audacity. This produces
a "hole" after a sharp peak which is audible. I set this to 0.2sec
which is near the recovery time of my ears. Anything much longer
leaves a very audible hole, when the level rushes back up after a
pause. Try it.

> 3. Is there a standard length for fade in times?

No, full stop. The reason for a fade in is to prevent a "bang" start,
and to fade out after the presented sound ends. I use a fade in time
of 0.2 or 0.3sec if I am demonstrating an effect or constant sound,
and a slightly longer fade out.

With a natural sound it is a matter of judgement. A second or longer
can cause impatience unless it is a gentle track, and in general a
fade in should introduce the sounds or calls about to be heard. For
instance if a blackbird song about a couple of seconds long is being
"introduced", keep to that time scale, which would be too long for a
fast wren song. With waves washing in at 10 sec intervals, you can
introduce these with a fade in to match. Start at 1 sec and you can
easily undo that and try again with Audacity.

> 4. I've noticed that the default vertical scale in Audacity runs
> from -1 to +1 (linear?)

I display in "Waveform (dB)" (in the down triangle) which gives a
decibel scale. The waveform is very odd, but it shows up quiet parts
very well, including noise levels. If you drag the cursor over the
scale, you can select a narrower dB range and reverse this with right
clicks.

David

David Brinicombe
North Devon, UK
Cogito cogito ergo cogito sum - Ambrose Bierce








<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