naturerecordists
[Top] [All Lists]

Re: Digital clipping; transfer to CD; increased gain

Subject: Re: Digital clipping; transfer to CD; increased gain
From: Walter Knapp <>
Date: Tue, 25 Jan 2005 21:04:29 -0500
From: Syd Curtis <>

> But it means that I sometimes finish up with a recording that may be most=
ly
> below -24 dB.=20

Is this a average, or the peak value? What you need to look at is the
peak values. A site with a few loud peaks, the average may be quite low,
just as it was at the site. A site with fairly even sound the difference
between peak and average may be very small.

> I use Peak LE to copy a digital line out signal from a Tascam DAT to a Au=
dio
> Media III Card in a Mac G3 to make an AIFF file, and then Toast to burn a
> data CD with a writer that uses TEAC technology.  My hope has been that
> after the Tascam has made the initial A/D conversion, I am retaining the
> original digital data unchanged through to the CD.
>
> Question 1.  Is that a justified assumption?

It should stay within the digital domain. Depending on what you do to it
if it's the actual original values.

If the DAT is recording at a sampling rate of 48 khz, and CD is a
sampling rate of 44 khz there has to be a resampling somewhere between
DAT and CD. You won't have the original samples, but derived ones. You
would need to run the DAT at 44 khz sampling.

You would also have to verify that the audio media card is not doing a
resampling. Some computer sound cards are based on a 48 khz chip and
resample both input and output to the computer. Even though everything
says it's just doing 44khz. I have no idea on your card, the problem
turns up more in PC cards.

> For pleasant listening, a recording below -24dB is too soft.  Peak allows=
 me
> to increase the gain, and for an audio CD I usually aim to increase the g=
ain
> so that the loudest sounds are at about -6dB, with maybe just a few notes
> registering -4dB.

In peak's gain adjust dialog there is a button for clipgard. If you
click that it will calculate the margin between the loudest sample and
0dB. You can then set for whatever margin you want, but never more gain
than is shown by clipguard or you will get clipping. I often choose -2dB
for my adjusted file. By simply subtracting 2 from the value shown by
clipguard.

Note that not all recordings should be at the same level. A quiet site
should really sound quieter than a noisy one. Choosing a fixed number
for your top level kind of negates this. Best way is to listen to the
set of tracks and see how they work together. I use Roxio's Jam for
burning CD's, and once the files are set you can have it play the CD
before you burn it.

> Q. 2.   Am I right in thinking that the intrinsic noise produced by my fi=
eld
> recording equipment is not totally dependent on the recording level, and
> will be relatively worse if I set the R L in the field unnecessarily low?

A preamp has a zone in the middle of it's gain when it gives the best
sound. Set way low, or way high and the sound won't be as good. If you
are setting at the extremes to get the reading you want you will need to
look at other ways of getting the reading.

In addition, if you set real low you risk cutting off the low end of the
dynamic range of the site. Not as much damage as clipping, but still not
desirable. For most nature recording dynamic range is not a big problem,
the original source site won't have all that large a dynamic range.

In general, digital recording has a lot of dynamic range to play with.
Far more than the analog tape had. With analog tape you had to keep the
signal at the top to keep from losing some of the dynamic range off the
bottom. With digital the equivalent to recording with tape set to 0 dB
would be a meter reading around -20 to -30 as far as preserving dynamic
range. Probably closer to the latter.

I should note that I record with my metering set for most all the high
levels showing on the meter to fall at -15 dB or below. That does not
mean the highest value in my recording is that. Meters are only telling
you a average of many samples because your vision is not capable of
keeping up with the thousands of samples per second (think of TV, where
your vision sees it as smooth movement with only 30 frames a second).
It's not at all unusual to take that -15 dB recording into peak and it
says the highest value is -5 dB. Or to put it another way, if I'd have
set the way we used to for analog tape, I would have had 10 dB of
clipping. That's what we are trying to avoid.

I don't always use -15 dB, that's my starting choice based on my past
experience. But, I'll listen first to what's out there. I know, by
experience, that at some times I may need to lower the reading even more
to prevent clipping. Or can raise it a bit if the sound is very even. If
you consistently see that your recordings have peak values that are well
below zero, then you can adjust your gain setting upward to get your
peaks closer to 0 dB. (and if Peak says your raw digital has a max value
of 0 dB, you set the gain too high) It depends a great deal on what you
are recording, and your equipment. And how much you are willing to risk
clipping. For instance, I don't set the gain to prevent clipping when a
truck passes while I'm recording some frogs. It will clip massively, but
I don't want that part of the recording anyway. I set so it won't clip
with the parts I want.

Note on the Portadisc I'm paying attention to the Peak Hold bar. This
gives me a little more chance at what the true peak is, but it's still
not the true peak value.

> Q. 3.   If I use Peak to increase the gain in the computer AIFF file befo=
re
> writing a data CD, am I negating my desire to have the digital data of th=
e
> sounds unchanged from Tascam to CD?

The data will be changed, but by a even amount. It will be changed in
terms of it's mathematic value, but not in how it sounds (except for a
volume shift). With the restrictions I've noted.

As I've noted, due to the sampling rates involved, you are not
preserving the digital data unchanged.

> Q.4.  Is is possible to explain in terms understandable to a non-technica=
l
> person, what happens to the digital data when I use Peak to change the ga=
in?

The entire recording is shifted by a fixed value, louder or softer
depending on what you choose. It's a fairly simple addition to all
samples, everything will retain it's relative loudness, unless you use
enough gain to clip, or reduce enough to drop some of the dynamic range
off the bottom. If you are within those bounds, you should be able to
drop the gain by the same amount you raised it and be back to the
original sound.

> I should explain that increasing old-age hearing loss, means that I can n=
o
> longer rely on what I hear as a satisfactory quality assessment of sound,
> and need to rely on what I cam see on computer screen or recorder dials.

I'm sure I'm not as far along as you on this problem, but I already
depend a lot on sonograms and recorder meters. And sometimes having
others listen. Even if I could hear perfectly, I'd use them.

Walt




________________________________________________________________________
________________________________________________________________________

<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