Nick, I looked at the clip, and sure enough, it played too loud. This was because the volume control with a hidden slider was by default at maximum volume. Lowering the volume made a huge difference. Could this be the problem? Otherwise it sounds perfectly normal to me.

0.1 db would have been *way* too loud.
Dropping the level globally by 6-8db would have helped.

What's really needed when mixing sound is a properly calibrated PPM meter (peak programme meter)

The BBC now uses the relationship of -18db (full scale) = 4ppm. This leaves quite a bit of headroom still for those transient peaks which may peak at almost zero.

Ie. if you send a 1 khz tone through the meter at -18 db, it should read PPM 4 on the meter. Once this is set up, your sound should not peak higher than PPM6. Dialogue can hover anything from 4 to 5.5. Films soundtracks tend to mix dialogue lower, so that those lovely bangs and crashes (and of course the music) sound loud in comparison.

Commercials always tend to sound loud because they are quite heavily compressed, but they are still within the maximum allowed level.

Pages to look at:
http://www.ebu.ch/en/technical/trev/trev_297-spikofski_klar.pdf
http://www.kozco.com/tech/audiolevels.html
http://www.gearslutz.com/board/post-production-forum/107770-film-broadcast-levels.html

Thanks for the replies and when (if) I'm feeling intelligent I'll try and absorb the information in those links.

This is a revelation for me. Everything I have read has taught me that the peaks in digital audio should be as close to 0dB (but never exceeding it). Besides keeping recorded audio close to 0dB, I have always applied this theory to my final mixes for DVD or web delivery. I didn't realise I should actually be quietening my final mixes to some other standard.

lol...Yes, it's why Sf has the extended Vu/PPM meters, and Ithink I've managed to get Vegas to show these meters once before as well, but that last bit could have been a false-memory

Thanks very much for your help guys, I'm learning a lot here :)

So in Sound Forge 9.0e I opened up the channel meters window and ticked "Show VU/PPM" and chose the "UK PPM" VU/PPM scale (which is another name for the BBC one).

Then I used "Tools > Synthesis > Simple..." and generated a flat tone with the slider at -18.0dB and the frequency 1,000.00 Hz. When I played this the peak meters shows -18dB and the VU/PPM meters showed 4.2. This is obviously a little over 4 and I don't know why or if that is important. I'm not sure if that can be calibrated. Can it?

Then I rendered a WAV file out of my project in Vegas and played it in Sound Forge. The peak meters were peaking at -0.9 dB and -1.3dB and sure enough, the VU/PPM meters were peaking at 7.8 and 7.7 (above the 6 that Alistair mentioned).

So then I used the output fader in the Sound Forge hardware meters window to drop the output to -7.0dB. With that setting the peak meters were peaking at -8.0 dB and -8.5 dB and the VU/PPM meters peaked at 6.3. How about that for web delivery?

Am I on the right track with all this?

My main question now is... what's the most elegant way to drop the level of the final mix in Vegas by 7.0 dB before rendering? Just the output fader on the master mix? A volume master FX? There seem to be a number of choices. I was wondering about a method that I could use as a template for future projects, that automatically increases or decreases the output mix to peak at 6 on the BBC meter, but without adding any compression. Perhaps the volume maximiser component of wave hammer can do this?

Kenny, I can't get those meters to show in Vegas. Would certainly be great if that would work. Anyone know if it's possible? Can the output of Vegas be routed into Sound Forge? Or perhaps a third party meter plugin perhaps?

What you need to measure is "loudness" which is the way humans hear things. Measuring peak levels is useless except to make sure the peaks don't go over 0vu. The "normalize" function in Vegas is also absolutely useless for matching average levels.

See this thread for a discussion of peak vs. average levels:

http://www.sonycreativesoftware.com/forums/ShowMessage.asp?ForumID=4&MessageID=406093

John

Thanks John, that much I had already grasped, which is exactly why I'm measuring the "loudness" level in the Sound Forge VU/PPM meters.

What I'm really hoping for is more info on quantifying what appropriate levels should be for delivery on the internet and DVD.

DVD levels are fairly standardized, esp. if you are going for Dolby approval. It also depends on what you are mixing - feature film, music only, or narrative with a little background music (doc, instructional, etc).

We use K System metering in addition to PPM, along with Dolby standards on dialogue, etc to determine proper mix levels. For DVD film, K20 is the currently (somewhat new) recommended metering system. For internet, I would suggest K14. Basically this system gives you a green/yellow/red meter that takes average levels for different formats into account, based on digital full scale metering, but does require a little research into how to apply it to your end product. Bob Katz has an article on this (Google for bob, katz, k system, film, mix - I don't have the link handy).

Typically, I set all my projects to an RMS "average" loudness of -20db using Sound Forge. First, I go to "Process" > "Normalize" and scan the levels, noting the RMS level. Let's say it's -27.4db with a few peaks to -2db, I subtract 20 from that and end up with 7.4. I then get into the Waves L2 Maximizer and set the threshold for -7.4db and let it do it's thing. This raises the loudness to -20db and lops off a few peaks in the process. I end up with a program that has an average loudness level of -20db. In case the RMS level has ended up louder than -20db, I merely use the volume process and turn it down. The -20db number isn't etched in stone, I do make exceptions depending on various factors, but -20db works 95% of the time.

Of course, as far as the level relationships between VO, nat sound, effects and music, mixing is an art and I have simply used my ears and used the meters to make sure I haven't gone over 0db at any point in the process. :)

John

Thanks KDM, the subject of the K-system also came up on a thread I started on the Sound Forge forum. The link you are thinking of is here.

John, thanks for the description of your workflow. I tried your method on the track in question. It was -18.0 dB on the first RMS scan. I knocked it down to -20.0 dB with the volume process. Then when I played it back the peaks on the BBC PPM meter in Sound Forge still hit 7.4 and 7.1. As musicvid said on the other thread, there's a lot of compression on the stock music.

By the way the free PSP VintageMeter that Bob mentioned on this thread looks useful as a VST plugin on the master fx chain in Vegas. Click the name on it to access the controls.

John, are you doing any mixes in 5.1? I'm curious to know if or how you might apply your system to the discrete files.

Sean - I wouldn't recommend using a normalization/limiting process on stems for 5.1 mixing. 5.1 levels should be determined at the mix stage, not stems, in order to maintain the proper relationship between dialogue, music and fx. I will often run stem groups (working in Nuendo, not Vegas, so this is easier) and might run a little additional level automation there, but usually every track is set for the mix with little or nothing on the 5.1 bus.

Even in stereo DVD mixes (for Nick and others), you might find an average rms level that usually works well enough, esp. if you are working with library music (which is often highly compressed already). But as pointed out, library music varies in compression (and quality), and too much of it is way over compressed. When we get post projects where the producer/director has used library cuts that are overcompressed, it just either takes more work to get the track to sit properly in the mix with narration, VO, fx, etc, or I end up dropping it lower than I would with unmixed custom music stems, or an unmastered music track. Compressed music alone will push your perceived loudness for your overall mix up, even if rms levels seem to be similar ("perceived loudness" being a function of the level of the whole frequency spectrum, not just an rms level).

In general, be careful about just using rms levels to define a mix (esp. if running stats on a mixed file, rather than metering in realtime as you mix) - it's fine as a rough guideline when watching meters in realtime since rms is an average already, but your ears are your best ally (or a good audio post engineer :-).

Agreed---and I am an audio post engineer! My 5.1 surround room is calibrated, with the necessary variations between theatrical and DVD mixing. In a properly calibrated room, meters are essentially ignored in favor of "if it sounds right, it is right." But after so many years of stereo music and broadcast mixing, I have a hard time keeping my eyes from glancing at the meters---and the waveforms of the final mixes are so small compared to the blasting loudness we're accustomed to, I almost think I need therapy to cope with it :-).

lol - I know the feeling, Sean. :-)

Good comments regarding broadcast levels, but it might also be that Nick has to deliver a version for the Internet. Any tips there?

Perhaps SMPTE or even W3C should ratify some standard for web delivery, since it's kind of a broadcast medium.

lol. web dot2, now including audio standards :)

My suggestion for internet is to adhere to K14. There will be sites with louder audio (over compressed for sure), but K14 is most recommended for general music and that's where web audio should be as well.

The Roger Nichols InspectolXL metering system can do K12, K14 and K20 along with a bunch of other metering standards. It's the best, most informative software metering I've seen. I've got all his plug-ins and they're well worth the money.

http://www.rogernicholsdigital.com/nativeplugins.html

Inspector is free, InspectorXL costs money.

John

(Roger Nichols is Steely Dan's recording engineer.)

<<My main question now is... what's the most elegant way to drop the level of the final mix in Vegas by 7.0 dB before rendering?>>

I dont follow this - in my final sound mix in vegas on a stereo channel I will stick my Waves L2 plug-in - this limits the output to a brick wall -0.2db, and the sound goes quieter or louder depending on the threshold I set. Most time's I have used it to increase loudness. ((Obviously if I have done my previous multi-channel sound mixes right I will have a good dynamic range between quite and loud sounds)). Waves L2 is a brilliant plug-in and can raise my mix by up to 6db+ with no perceived loss in dynamics. For the life of me it seems crazy to "drop" your "final" mix by 7db. Did I read that right?

Before Vegas: The multichannel mixes I do in Cubase or Wavelab depending on the type of music. These are done as 24bit files which gives me LOTS of headroom so I dont clip and get nasty digital distortion. I use my ears and a reference DVD/CD to ensure I've got a reference of some kind at least. I try and get a good dynamic range - I find voices the most challenging (with music and foley I have lots more control) and again use my ears to get the right mix (and there is a rough cut video playing in cubase/wavelab simultaneously).

I then do final exports and dither to stereo 16bit and bring it into Vegas. Then I do the above, stick on Waves L2 and adjust thereshold etc to get the right level. I check my rendered DVD against another ref DVD.

(I know many people simply use vegas (and I do for some stuff) to put their music on. Same principles apply. Except if you just want to do it all in one go, ie, mix AND master, you need to put your mastering plug-ins last in the chain. I dont like this but for the simplest 2 stereo sound channels video)

Did your friends think the voice was too loud - perhaps it was uncompressed? ALL and absolutely all my voice stuff goes through a compressor plug-in - it would simply sound unprofessional without one IMHO.

If they thought the music was too loud - well there's no such thing; if your music didnt digitally distort then its absolutely fine. It can be as loud as it wants to be!

Hope this helps.
cheers

>> For the life of me it seems crazy to "drop" your "final" mix by 7db. Did I read that right? <<

Yes you did Sherif. My original thinking was just make it peak as high as possible without going over 0dB. Until I got these comments. I found that dropping the mix by 7 dB is what is required to get the BBC PPM meters in Sound Forge peaking at about 6, which is what was recommended by Alistair earlier in the thread.

I must admit my DVDs do sound very loud when I first put them in the player compared to whatever else plays through my hi-fi (DVD, TV, PC). But until now I was kind of proud of that fact :)

By the way there are no voices on this video. It's purely the stock music. When I do a voiceover I do add compression.

The full k-system is a bit difficult for me as it involves calibrating monitors. I don't own a sound pressure level meter and in any case I work late and have neighbours!

Anyway I'm now seeing if I can simulate k-14 metering without spending $250 on a plug-in. I've put the free PSP VintageMeter VST plugin on the master FX chain in Vegas, switched it to "VU", and set its 0VU reference level to -14dVFS and VU integer time to 600ms (according to 2nd post on this thread). With these settings the needle is generally around -4dB to 0dB with occasional peaks into the red (+1dB) but never near or over +4dB. So it seems to me it's already within k-14 guidelines and that in this case the perceived loudness is due to the amount of compression on the stock music.

Just about every 'modern' audio CD I've ripped hits the stops at 0dB FS on Vegas's PPM meters.
When recording audio one should aim for as close to 0dB without clipping, that gives you the best S/N. However that can be a pretty risky / difficult approach. I usually record at 24 bit and leave 10dB headroom just in case when I'm recording live sound.

Mixing for video / film / CD is a different matter to what you do when recording. The BBC might have it's own standards, commercial broadcasters, especially radio, pull all manner of tricks to get their station to sound the loudest, including adding DC offsets. Decades ago in the early days of digital audio we used to have a AM receiver hooked up to a VU meter. The national broadcaster wouldn't get much over -10dB, with some commercial stations the meter barely moved off 0dB.

There's no one right way. You need to factor in the probable listening environment, how you'd mix for cinema is different to how you'd mix for a TVC. Even for a DVD I try to factor in where it's likely to be listened to. If it's a boring training video that's going to be watched by a bunch of employees in some aweful room I compress the speech really hard, no room for subtle sounds that'll be lost in the ambient.

I'd strangly recommend a copy of "Audio Postproduction for Digital Video By Jay Rose from CMP Books ISBN 1-57820-116-0. Been a few years since I bought it but I think it's still in print. Covers pretty well everything from how to wire an XLR to how to mix for different formats.

Bob.

The free 'RMS Buddy' VST plug-in is handy for monitoring average & peak levels - [url=destroyfx.smartelectronix.com/extras/]

Tony