Someone says "anything over -10db in digital video is distortion" and you say....

Someone made the statement that "on a digital video deck you want to keep
the levels below -10db because above that you're into distortion." Further
they state that 0db is the same as clipping. I was under the impression that
attempting to *exceed* 0db yields clipping, since there isn't anything over
0db. Soundforge definitely makes a distinction between 0db and "clipped".

What say any of you?

Thanks for all shared wisdom.

 

I think this person is mixing-up two things. Yes, you keep your levels
below -10dB in the digital domain, and you do this in order to avoid
them reaching 0dB, because that is where they clip. The area between
-10dB and 0dB is what is called "the headroom", and is there for
louder sounds (plosives, cough, whatever).
-10dB is quite high actually, if you don't know exactly what you're
doing. You can pick your choice between -12dB, -14dB, -18dB and even
-20dB. All these values are used as safeguard against clipping, and
each soundo has his own standard. I use generally -12dB for normal
dialogue, and -18dB if I don't know what could happen. But you will
have to find your own comfort-zone, because mine does not need to be
yours

cheers

-martin-

 

You are correct. I think the intent may have been to say that metering,
being what it is, an average of -10 (or -8 or -18 or ??) is a good place to
place the average peaks, because there will often be peaks that exceed the
average by 10 db or so. Therefore, to avoid clipping set a level that will
acommodate all peaks so that none exceeds the brick wall 0 dbfs.

Steve King

 

And also the short plosives or other short transient sounds may not
register fully on the meter or led's so they may be well over 0 dB but
don't register over 0 dB.

I actually thik people should pay more attention to the AVERAGE meter
reading rather that the peak. Pick a level say -15 dB and try to set
the level so that 1/2 the time the meter is above -15 and 1/2 the time
it is below. If you do it this way, the "loudness" i.e how loud it
actually sounds ... will be more consistent.

Mark


Mark

 

This person is mixing levels....-10dB is a line level voltage for
certain equipment, not a reference point for where clipping begins. 0dB
is the absolute ceiling of the loudest volume that can be recorded on
digital equipment (ie. 16 1 bits for that sample). It's possible for a
-15dBu signal to cause clipping on some equipment, and a +5dB u signal
to be below clipping on other equipment.

Many people and software will consider the point at which you reach, but
not exceed 0dBFS to be clipping, simply because they are both
represented the same way in binary (all 1's)

Other people, like myself, believe that clipping may occur, but only
matters when you can hear it. Some equipment handles clipping really
well, and if you're recording a thrash distorted guitar part you might
not even notice lots of clipping. On the other hand, if you're recording
clean vocals, you probably won't be fine with even minimal clipping.

But yes I think you're right, clipping occurs when they are no more
zeros to turn on to represent the extra signal, or there are no more
metal oxide particles in a tape to represent the xtra signal/information.

Jonny Durango

 

Most digital equipment clips someplace in the last 1 dB
before FS. Up until that point, digital equipment is
generally very clean.

However this begs the question - how are you measuring
levels? There's really only one totally reliable way to
measure levels in the digital domain, and that is to record
a sample and then look at it with a DAW.

What you actually record is where the rubber hits the road.
Everything else is an estimate.

In many cases, clipping of real world digital equipment
takes place a bit below 0 dB, IOW -0.1 dB or maybe even as
low as -0.5 dB.

Furthermore, clipping might be frequency-dependent. If
clipping is frequency-dependent, the clip point is probably
the lowest at the highest frequencies. Most equipment will
hit its highest undistorted levels at 10 KHz, but there may
be increasing losses at higher frequencies.

Most of my cautionary comments apply to the less expensive
digital equipment.

Fact is, a lot of digital equipment clips at some point
above -1 dB, but below 0 dB.

Theoretically, 0 dB signals are not clipped.

 

Jonny Durango wrote:

You'd hear the clipping at playback with D/A conversion, correct?

Then wouldn't it depend on the converter? If the converter created a
smooth peak (possible?) would that explain why some equipment handles it
better?

 

I think I should clarify, this is referring to material that's already been
recorded, not a question of where to set the levels during recording.

I was under the impression that the waveform and VU meter in Soundforge etc.
"shows all". That there are no peaks that don't show up on the wave form and
the VU meter tells you exactly what the peak levels are up to 0db, unlike
analog meters which show an approximation of the peaks, which are likely
somewhat below the actual levels.

Yes? No?

 

In an environment with digital and analog VTRs "tone" always is set to read 0 on
an analog VU meter and -20 on a digital meter. I am referring mainly to Beta SP
(analog) and Digi-Beta (digital machines) With properly recorded audio the level
on a Beta SP deck will fluctuate around the 0 level marker but sometimes read a
couple of db over. The same audio will read maybe 10 db above the -20 marker on
the digital machines due to it's "peak reading" tendencies.
In the old days (with tape) when there were only VU meters an engineer had to
"just know" that certain sounds would read much lower on a VU meter than they
really were. An extreme example would be orchestra bells or a triangle. If you
let the level approach -10 on a VU meter the sound would be distorted. This was
especially important on the first generation of the recording. Some of the
extreme peaks would saturate the tape but in a gentle way. Later in the "mix"
you could boost these same sounds and avoid the distortion because the extreme
peaks in the overtones were now gone.

 

Assuming the meter shows the true instantaneous peak level, this is utter
crap.

On a digital recorder, the higher the recording level the _lower_ the
distortion -- until you exceed 0dB, at which point you clip.

 

Soundforge.

In case you haven't seen my previous posts I'll give you a quick rundown.
Made a Firewire transfer of a musical performance from a DVCam deck to my
Sony TRV-240 Digital8 camcorder. From there through a Firewire port to my
H/D. When looking at the still unprocessed file I found that the highest
peaks were either right at 0db or registered as clipped. All the highest
peaks were in one particular song which was somewhat higher in level than
the rest of the performance.

I should probably mention that the people who recorded the sound at the
performance really messed up all over the place - they showed up late and
didn't get to do a sound check. Her vocal levels were all over the place in
the mix from hot and distorted to buried. Half the time she couldn't hear
herself in the monitors. Her mic died during one song. During the opening
bars of the performance (for TV), someone apparently hit the "on" button on
the onboard reverb - suddenly the sound is awash in ridiculously overboard
reverb and a couple of seconds later it suddenly goes away (I can just see
the board guy fumbling in a panicked scramble for the fader or pot.) It was
apparent from the shape of the waveform that for some reason the first and
last songs had compression/limiting applied to them - with a lot of
accompanying distortion - I assume with an onboard compressor/limiter they
had. Why just the first and last songs I have no idea. Given everything else
that happened, I have to wonder if they even knew they were doing it. Can't
imagine why the TV studio would do it.

Anyway, the fact is there's distortion all over the place, even in places
where the sound is nowhere near clipping. It's not caused by my gear, the
VHS copy the station gave her of the original tape sounds the same way.

I used Soundforge to add reverb via the Acoustic Mirror plugin. To do this,
I had to lower the overall levels by 2db to keep many more peaks from
clipping than already were once the reverb was applied. I figured I should
put it back to where it was, i.e. with the highest peaks (maybe 2 or 3
spikes where this ocurred) at 0db, which is what I did. Saved the whole
thing with the DV format, played it back to my camcorder via firewire and
then back to their DVcam deck.

When I mentioned to the Station manager about the levels issue, he seemed
incredulous that any of them were that hot. I explained that this is where
they were on the original file. He even asked if I had "adjusted" anything
during the Firewire transfer. I'm not aware that this is even possible. I
was under the impression that a Firewire transfer is utterly unlike going
from an analog source, that a transfer simply dumps whatever data is on the
digital tape to the h/d, no user intervention possible. At any rate, all I
did was set up the cam, hook up the Firewire cable and hit "capture" on the
Pinnacle Studio 9 software.

 

That would be a pretty bad digital video deck, however they may want
to keep the peaks 10 dB below full scale because they want to have
10 dB of headroom in their playback chain. It could be that the deck
that they play back from produces an analog output from a -10 dBFS
digital recording that drives their system into clipping, but that's
sloppy system engineering on their part.

In any case, it's convention, and a trade organization (SMPTE)
standard, to deliver recordings that don't peak above -10 dBFS.

That's not correct.

First off, 'dB' by itself doesn't mean anything when talking about
levels. Let's assume that we're talking about dB relative to digital
full scale, which is the maximum digital level. This is what 0 dBFS
is, and it's likely that this is what your misinformant meant by "0
dB." You can have a recording with peaks that reach 0 dBFS that isn't
clipped. And you can't exceed 0 dBFS. But if you put an analog signal
into an A/D converter high enough so that it would go over 0 dBFS if
it could, then you'll get clipping.

Most digital devices have a "clip" indicator, but those work in
different ways. (there's no standard for it) Some will light up when
the level gets to within a couple of tenths of a dB of full scale (say
-0.2 dBFS) just to give you warning that you're about to clip if
things get much louder. Some light up as soon as it sees a sample that
reaches full scale (this is becoming common for 24-bit systems).
Others light up when they see three (or some other number greater than
one) consecutive samples at the full scale value. That's a pretty good
guess that clipping has occurred, but if you're recording square
waves, you can have the clip light on all day and still be recording
what you put in.



--
I'm really Mike Rivers ()
However, until the spam goes away or Hell freezes over,
lots of IP addresses are blocked from this system. If
you e-mail me and it bounces, use your secret decoder ring
and reach me here: double-m-eleven-double-zero at yahoo

 

Agreed, but it's surprisingly common in the broadcast world.

Yup, gotta waste that last 10 dB by design. Sigh...

 

Doc wrote:

What's to adjust? ;-)

That's my experience.

I've done my camcorder transfers using real basic software
and hardware - namely Windows Movie Maker (WMM) and a
generic Firewire card. It seems to work like clock - I push
some buttons, wait for a while and there's an AVI file on my
hard drive.

If I need to play with the sound track more seriously than
what WMM allows, I've used Audition/CE to pull the sound
track off of the computer file that Windows Movie Maker
created. WMM supports adding a stereo audio track back into
the movie from my hard drive.

Since the audio is always kept in the digital domain,
there's really not a lot to mess it up. It is what it is,
until I start playing with it with one of the editors that I
use.

 

You are right.

-m-

 

I haven't read all the posts , but -10 digital seems to becoming a new
T.V. standard ,
From feeding cameras to the quality control inspector at discovery channel
sending shows back if there are peaks over that !

Maybe Will , will lend his experience , but it seems to combatting sloppy
field
work and over agressive transmitter chains .

regards Greg

 

FWIW, we set reference tone at -10 on the DV deck's meters when recording
dialog. For music, we usually use -12. But on dialog, we're nearly always
compressed at least 3:1 at the mixer, so the 10 is more like 30.

And for music, there's just too much the audio guy does to the 2-mix to
fully explain here, but for levels, there's usually a compressor on most
every input (some of them 2-stage). For critical events, there's a separate
multi-track recording for re-mix, etc.

Steve

 

Over here in the BBC, we use -18dbfs as our reference level and we set
our peak to 8db higher than that - ie -10dbfs. This ensure that
there's always 10db of headroom before the onset of clipping.

Steve

The Doctor Who Restoration Team Website
http://www.restoration-team.co.uk

 

The general standard for video in the US is (and has been for quite
some time) peaks at -10 and vu levels peaking between -18 and -24.
While this standard does fail to take advantage of the additional 10db
of headroom available in the digital realm, it allows for a seamless
signal flow between analog and digital machines.
This is not the case with DVDs, where there is not yet a clear standard
established.

 

That's true for digital peak levels, which are not indicated by a VU
meter. Do you know where the -18 to -24 range on the VU scale is?
Let's not get our meters confused.


--
I'm really Mike Rivers ()
However, until the spam goes away or Hell freezes over,
lots of IP addresses are blocked from this system. If
you e-mail me and it bounces, use your secret decoder ring
and reach me here: double-m-eleven-double-zero at yahoo