Hi, I make event DVDs for playing on
all types of home TV sets.
There's spoken word and music.

I've been having my levels average about -9dB
and keeping in the range of going up 6dB (to -3) and
and down 6dB to -15.

I'm tired of the effects of heavy compression and
limiting and was considering shooting for an average
of -12dB, and keeping in the range of up 9 and down 9,
so keeping levels between -3db and -21dB.

Do you think the levels would end up being too low.
Someone here menioned that TV programming is
usually kept in the area of 3dB up and 3dB down.

Dave,
Peaks on DV should be around -12db (equals 0db in analoge)- I try to keep my average around -15 or so but most importantly I work at keeping peak at-12. At most of the seminars I do that's generally fairly easy to do it's the DJs and bands at receptions that play havoc.
Sometimes with the camera set to peak at -12 the average drops to about
-20 and then I have to boost it a bit but I'd rather do that than have to play with compression to try to save a clipped piece. But that's just me!

Don

Dave,
Peaks on DV should be around -12db (equals 0db in analoge)- I try to keep my average around -15 or so but most importantly I work at keeping peak at-12.
Don

Actually, this number is dependent on the gear's calibration, not any baseline spec. This is why the ATSC determined a few years back, that -20dBFS is the equivalent of 0 dBVU.
Content dependent and calibration dependent, -12dB is a reasonable averaging point, but if you've got more than 12dB of dynamic range, this means you're hitting 0dBFS, and unlike analog, you can't go past that point. it truncates/blows up/shatters/brickwalls/whatever you want to call it. Keeping your PEAK at -12dB means you're throwing away a lot of bits.

What constitutes a 12dB dynamic range?

Measuring from the most quiet point to the loudest point is how dynamic range is determined.

Hey Spot,
Thanks for the info-I've always looked at -12 as my peak level and try to stay around -15 to -20 for average but now I need to look at that again. But maybe not-you know I've got it going pretty much where I want it now using those levels (took me some time to get them figured out-about 5 years ;-0) so maybe I shouldn't mess with it.

BTW, I don't remember if you were the one who recommended the AKGSE300/C93 hypercaroid - if you were thaks a milion-great sound for indoor run and gun and receptions and such-sounds so much better than the ME66 and the AT897. (if you didn't reco it thanks for all the audio help anyway)

Don

... -12dB is a reasonable averaging point, but if you've got more than 12dB of dynamic range, this means you're hitting 0dBFS...I think you must have meant 24dB dynamic range?

I could have said that better, sorry.
If your quietest point is at -12dB, and you have say...14dB of dynamic range, you're screwed.
If your average is -12dB, then you have 12dB of dynamic range. Thanks for catching me on that, Fred.

Measuring from the most quiet point to the loudest point is how dynamic range is determined.
I understand that, but I was looking for a real-life example of a situation where you would encounter such dynamic range. Basically I am trying to get a feel for the numbers.

depending on the person's speaking voice, you *could* encounter a dynamic of 20dB, if they were passionate or pulpit pounders, and the spoken word contains a lot less dynamic than the singing voice, which is why compressors are all but required in most singing-voice situations and many spoken word situations.
Additionally, you might have two actors with a wide range from the most quiet point of one actor and loudest point of another.
Television has a dynamic range of around 8-10dB, which is better than it was just a few years ago, and getting wider bit by bit (no pun intended). Technology has the ability to provide a lot more range than that, but at the same time...you gotta be careful about how much range you put into something. Walter Murch says he's squeezing 25dB of dynamic into some motion pictures. That's a lotta range! I'd like to be that good and still leave the audience without bleeding ears. :-)
Anyway...recording normal, average dialog expecting around 12dB of dynamic range should be plenty, but as a side tour, this also explains why you need compression, and good compression, for most projects. I'd be seriously in trouble without my UA100 or WAVES tools on the software side, and various compressors on the hardware side.
Somewhat related... This is a great read (http://www.filmsound.org/articles/sergi/index.htm)

For the record -

Classical music usually has much more dynamic range than that. And it's rarely compressed. Requires good equipment for playback, and a lot of skill for recording - and good sound-checking! A soprano has a tremendous range, and a peak can easily jump 30db above averge for lots of pieces!

Percussion, applause, etc. have lots of big peaks, too, but are often forgivable in a way that other sounds aren't; you can sometimes let those go. But listening is the only way to know!

Sorry, audio newcomer here. How do you impose DVD-appropriate levels on your audio using something like Sound Forge 8.0? Using Normalize (peak or RMS?)? or Dynamics (which settings?).

Sorry, audio newcomer here. How do you impose DVD-appropriate levels on your audio using something like Sound Forge 8.0? Using Normalize (peak or RMS?)? or Dynamics (which settings?).

Normalizing raises the overall gain of the track so the peaks are just below clipping. It's like turning up the volume control. Compressors such as Sound Forge's Dynamics plugin, OTOH, selectively change the gain, raising quiet passages while leaving the loud ones as is. There's no single set of settings that would be universal. It depends on the type of music, etc, and personal taste.

I discovered some "tips and tricks" documentation in the product information section of the Waves web site - www.waves.com - that can help you with starting points for settings to try. Even though they are discussing their own products (which will work with Sound Forge, BTW) those pages can help you understand the principals.

Hi Steve, thanks for your answer. I've gone through some books and videos and have a general idea of how to play with normalization and dynamics/compression. I think my confusion at this point is more one of terminology and how to find certain quantities using something like sound forge. When someone says they keep their average levels at -15 dB, are they referring to an RMS of -15 dB over the entire sample? (I know they don't really mean 'average' because that should be -Inf, right?)

For the record -

Classical music usually has much more dynamic range than that. And it's rarely compressed. Requires good equipment for playback, and a lot of skill for recording - and good sound-checking! A soprano has a tremendous range, and a peak can easily jump 30db above averge for lots of pieces!

Percussion, applause, etc. have lots of big peaks, too, but are often forgivable in a way that other sounds aren't; you can sometimes let those go. But listening is the only way to know!

Of course it does. But you cannot put uncompressed classical music into a broadcast or into a DVD stream, for the most part. Audiophiles *hate* this, of course, but when mixing/mastering, you've got to consider John Q. Public and not the guy that has top end Martin-Logan, tube or super high end amp/preamp, etc.
I would shudder to think anyone might use a DV camera or HDV camera for primary audio source from an orchestra or symphony....Yuk.

I think you must have meant 24dB dynamic range?


This is what I was thinking, too. If you are keeping average
levels around -12dB, then you have 12dB dynamic range above
that, and 12dB (or thereabouts) below, so that means you
have 24dB of dynamic range.

I see how Douglass' definition of "lowest to highest" is
"dynamic range", but here we're talking
about "average level" being -12dB, not lowest level being
-12dB. So, if you're average level is -12dB, shouldn't your
dynamic range be much more than 12dB?

I see how Douglas' definition of "lowest to highest" is
"dynamic range", but here we're talking
about "average level" being -12dB, not lowest level being
-12dB.

It's not my definition, it is *the* definition.
"Dynamic range describes the ratio of the softest sound to the loudest sound in a musical instrument or piece of electronic equipment."
"A workable definition of dynamic range is the ratio of the highest (lightest) signal which a scanner can record to the lowest (darkest) signal..."

[QUOTE=Dave Largent]So, if you're average level is -12dB, shouldn't your
dynamic range be much more than 12dB

Yes, and no. If your average is -12dB, and you have an upward swing of 14dB, then you would be +2dB over 0, but there is no +2dB over 0dBFS, which is the point I'm putting across (and apparently badly, if it's not being understood). Yes, you have a dynamic range which may or may not exceed or be equal to 24dB with an average of -12dB.
I can't speak for anyone else, but I'm more about peaks and keeping my average slightly hotter, which is why good compression is important to me. I'd much rather have a slightly hotter signal with light compression on the top than leave a lot of bits on the floor simply because the source wanted to get loud for a passionate moment.

Remember, while acquisition is different than delivery, you can never recover lost resolution, and it's all about using up those numbers. Why go 30 in a 55 lane?

I'm sure others feel differently about it, but that's my method of working.

...I think my confusion at this point is more one of terminology...When someone says they keep their average levels at -15 dB, are they referring to an RMS of -15 dB over the entire sample?...They simply mean that they set their levels so there is about as much content over -15 dB as under -15 dB. At least, that's what I mean. "RMS" does not have any meaning with respect to dB, which is just a way to express large ratios with small numbers.

...(I know they don't really mean 'average' because that should be -Inf, right?)You are correct that -15 dB is not the arithmetic average of the actual signal that ranges from, say, -25 dB to -5 dB. But you are in error about -Inf. That woudl equal a voltage level of zero.

They simply mean that they set their levels so there is about as much content over -15 dB as under -15 dB. At least, that's what I mean.

Yes, that's what i mean when I talk about having
an average dB level.

Now, here's what I was wondering about (in my
original question): If I will run into any trouble
by making a DVD with a dynamic range of
from -3dB to -21dB, which is 18dB range.
Now, Douglass says TV only has a range of 8-10dB.
Why is TV's range so low? And how will this
work out with making a DVD (with a range of
18dB) for TV play?

Thanks for the clarification about 'average levels'. This has also become more clear to me after downloading Ozone 3 today -- great program -- the meters and graphs in Ozone 3 seem to provide more understandable information (to me at least) than I was able to obtain from Sound Forge alone.

You are correct that -15 dB is not the arithmetic average of the actual signal that ranges from, say, -25 dB to -5 dB. But you are in error about -Inf. That woudl equal a voltage level of zero.

I guess I'm speaking more as a mathematician when I say this. If the signal has correct DC-offset, then the mathematical average should actually be -Inf, yes? (The statistics tool in Sound Forge confirms this -- it reports the "average" to be -Inf in both channels.)

Now, Douglas says TV only has a range of 8-10dB.
Why is TV's range so low? And how will this
work out with making a DVD (with a range of
18dB) for TV play?

It's not a question of TV having a low dynamic value, it's the range in which the levels are kept deliberately. Too loud, everything becomes squished down at the broadcaster anyway. Too soft, the viewer can't hear and they turn up their set, but when a commercial comes on it blows them out of their seat. Stretch it as far as you will, but realize that if it has too much dynamic range from actual audio to loudest point (meaning from most quiet sound you want audible vs silence) then you'll either have viewers straining or recoiling. This is another benefit of 5.1, of course, is you can have a mix with much more happening, since the dialog is usually in its own channel, and you've got more locations for sound sources.
Borrowing from Murch, when we started working for theatre, we built a "popcorn loop" just like they did for Apocalypse Now, and still use today. It really helps you figure out what is being heard where when you've got that sort of a source. In a theatre mix, you can have monstrous dynamic range. Too much. But, most mixes are kept to the 20dB range. As I mentioned earlier, Walter Murch, Brian Keane, and a few other mixers are pushing 25dB worth of range. Dynamic range is abused by most engineers, both in video and music industries. Pushing everything hot doesn't leave the viewer any space to breathe and you'll wear their ears out and they become numb. Too much range, you'll leave them straining to hear, but recoiling when you make a punctuation mark. back to the original question or point...Television *can* offer a broad range. But the question isn't how broad it is or isn't, it's how audible or listenable the audio becomes if you stray too far from general standards. Experiment, it doesn't take long to figure it out.
Bear in mind as well, now we're talking about delivery, which has very little to do with acquisition. It's much easier to record a solid signal and pull it down in post than it is to record a quiet signal and attempt to recover it in post.

I thought Dynamic Range Control in Dolby Digital existed to address this problem.

I thought Dynamic Range Control in Dolby Digital existed to address this problem.

Do *you* want to control what the viewer hears, or do you want metadata embedded in the stream to instruct whatever variable their playback device might have to deal with it?

Once you start undergoing the process of submitting audio to Dolby to be certified, you quickly start figuring out means of circumventing the process as much as you can.
LFE management, Dialog management/normalization, equalization, etc are all bypassable to a limited extent, but yes, you can ignore it all and allow your encoder to deal with it for you via the metadata inserted and processing. Most engineers prefer to not do it that way, but it is there if you want to fall back on it. Kinda like having an auto TBC for your video though, and you never know for sure what the results might be. It might not be that important to you though. Some folks are thrilled at their mixes on plastic computer speakers and they're very happy with Radio Shack mics, too. It boils down to what you're happy with and what you're exposed to.