Mixdown Levels

[berniebeaudry]

Most of my work is for cable tv shows, and/or PBS. When I've had communication with post on any of my tracks, recording "hot" was never an issue. I did hear from several editors that they would like to see more level and I worked with them to get the tracks up to where their system was ingesting a level closer to what they considered appropriate. My son is an editor and as long as production tracks aren't distorted he finds it easier and faster to turn levels down in the mix instead of having to gain them up. I try to avoid limiting no matter how good the limiter in the chain is but I do try to get a good strong signal to the recording device. I've had "overs" on very rare occasions according to the meters I'm looking at, but a playback doesn't reveal any horrible distortion. I try not to do that but I think the fear of 0 might have some recordists being a little too careful when it comes to level. Glenn your explanations are excellent and I think I'm pretty much working the way you describe.

Bernie

[soundtrane]

<<What audio level do you find most appropriate to listen to movies like "The King's Speech" or "The Social Network"? Now how about "Sucker Punch" or "Batman: The Dark Knight"? You may be tempted to turn up the volume on the latter movies, but maybe not for the former. >>

This is mostly due to the DVD mastering process - the DIALNORM and compression presets being used.

I recently had major issues with DIALNORM value and compression being applied blindly to a mix that has been already tailor-made to suit the dynamic range of a DVD. There were more issues than just this - simply showed how careless they were, a Brit company which shows off "expertise" in this field.

I had to check 5 or 6 test runs and finally had to settle for a compromise in levels, better than what it was at the onset, but not quite up to standards. The production company had already announced the DVD release date.

While LKFS and the EBU move to curb drastic level changes between programs on TV (read commercials) are a good move, I don't think this applies to production sound.

However, what it could possibly do is to sensitize people in film post to come up with mixes that use dynamic range of the medium without necessarily testing the upper limits for 80% of the duration of the film. This however is a whole new and different topic.

-vin

[Tom Conen]

Glen!

You made two general statements I was referring to:

1. "[K]eep the loudest peaks approaching but safely below 0dBfs" and

"if the hottest signal is -6dBfs, then the tracks are at least 6dB lower than ideal"

2. A VU-Meter is "Virtually Useless" because it won't tell you where the peaks are. And you added,

"[that] is why, [...], we have stopped seeing VU meters used in field recording equipment".

I had to comment on this statements, because I know you are well respected and with a lot of experience. I'm sure, that you know what you're talking about and what you are doing. But I'm sure that readers with less experience will draw the wrong conclusions and will follow your advice without thinking about it.

My answer to statement #1:

This view is a wrong understanding of digital recording. There is no technical need for getting your peaks up to 0dBFS (using 24bit). You can do so, because it may not compromise your audio signal as long as you don't hit 0dBFS.

Imagine you are a dancer in a really big ballroom and the walls of this ballroom are 0dBFS. You could dance in the middle of the room and make all the moves you want with a save space between you and the walls - and deliver a great performance.

You could also try to dance the same choreography as close to the walls as possible. But if you bump into the wall your dance is ruined and you wish the room was bigger! ?!?

(BTW: What's the use of a mechanism that grabs you and puts you back in the middle of the room right before you hit the wall, when you could have danced in the middle right from the beginning?)

Let me quote from a Sounddevices Manual:

"The 302’s [metering] scale is designed for digital recording devices, providing maximum information between -30 dBu to 0 dBu which is where typical peaks occur (-50 to -20 dBFS). This allows the user to record with a full 12 to 20 dB of headroom, [...].

[...], the meter color changes to orange at 0 dBu and red at +8 dBu. These color changes correspond to -20 dBFS and -12 dBFS respectively, which are commonly used recording levels for today’s digital recording devices."

Manufactures give us the advice on how to use their products. In technical papers and whitepapers they explain why. I also recommend the book "Mastering Audio" from Bob Katz. He explains digital recording in not too technical terms.

If someone can show me any benefits (sonically or for workflow reasons) from setting my peaks close to 0dBFS, I will accept them. I recommend to leave a headroom!

My answer to statement #2:

VU-meters are no relict of analog times. They are still out there (e.g. in SD Products) and not just "for aligning digital meters with VU meters".

Let me quote a Sounddevices Manual (788T) again:

"VU (Volume Units)

Ballistics correspond closely to how the human ear perceives loudness and provides a good visual indication of how loud a signal will be. [...] While giving a very good visual indication of perceived loudness, VU meters give poor information on actual signal peaks and thus not ideal for use during recording. [...]

Peak Only

Peak-reading ballistics correspond to actual signal maximums, but don’t necessarily correspond to perceived signal loudness. A peak meter has a near-instantaneous attack to display maximum signal amplitude and a slow decay to allow the user to see them. Peak metering is essential for digital recording, since signal overload can cause immediate distortion. [...]"

Depending an the kind of signal, a VU-meter can be more or less accurate. You have to know, how to read them and when you can trust them. Always use your ears!

A VU-Meter gives you that visual reference you need, when you are

- inexperienced

- tired

- busy

- ...,

and helps you to meet postproduction's specifications.

So in conclusion: The best way is to use both meters. (And of course your ears.)

# My own experience:

Post wants dialog with a constant loudness and peaks not up to 0dBFS.

Because: Less work!

So again: No offense to Glen. I appreciate this dialog and I'm eager to hear other opinions. Workflows may differ without being wrong in any way.

As this would be off topic, I'll start a new one, regarding VU-meters and it's use.

[Tom Conen]

My son is an editor and as long as production tracks aren't distorted he finds it easier and faster to turn levels down in the mix instead of having to gain them up.

Of course it's easier to turn levels down. (Fader down) Especially when your production tracks have no headroom. In this case post is forced to compress your signal in order to bring levels up without hitting 0dBFS. - More work!

It's our job to minimize the work of post as good as possible. So it's best to ask them were the average levels of dialog will be in the postproduction process and record at this level. Also provide enough headroom (maybe you need limiting or compression), so post can bring the level up or down just by using a fader.

[Mark Orusa]

This is mostly due to the DVD mastering process - the DIALNORM and compression presets being used.

This only applies to DVDs using Dolby Digital, not DTS or stereo PCM, and the disc author must activate the dialog normalization option in the encoder. Popular Blu-ray audio codecs are DTS-HD Master Audio and Dolby True HD. While both offer the optional dialog normalization feature, they are lossless mixes that are equivalent to an uncompressed track. Some Blu-rays have uncompressed PCM tracks. Despite these advances in audio, some movies have been entirely remixed for home video, often reducing the dynamic range.

Here's a couple of good discussions regarding dialog normalization:

http://www.avsforum....ad.php?t=822297

http://www.digitalho...ead.php?t=69985

Mark O.

[Marc Wielage]

There's a very good, very long, involved discussion on mixing levels for theatrical features, TV shows, and commercials at this link on the Gearslutz Forum's Post section, with some excellent comments by Marti Humphrey (among several others). Suffice it to say, nowadays, it's very hard to do a mix purely by VU meters or even peak-reading meters and get it approved by the various distributors and networks.

I have seen perfectly good mixes (even from an Oscar-winning film!) get rejected in home video because of two or three peaks that went past their mandatory -10dBfs accepted peak level. I hated to slap a limiter on it, but that was the only way to get this particular title approved. It eventually went out on Blu-ray and nobody noticed -- it still got rave reviews.

Marti also has a good, very detailed discussion on speaker level set up for mixing rooms at this link on the Digidesign User Group. He makes a good case that the standard 85dB reference level is not optimized for home use, and recommends 82dB instead.

My personal experience is that I generally have no problem with TV mixes, but I feel that some theatrical mixes have much too much dynamic range -- mainly the low-level dialog is too low. You bring up the overall monitoring level for that, and suddenly you're hit with a 120dB explosion! I've heard a handful of remixed films and TV shows sound like this, too, but out of respect for the mixers, I'll refrain from criticizing them. But the gist of it is, there's no real audio level standard for DVD or Blu-ray... yet.

--Marc W.

[John Blankenship]

Of course it's easier to turn levels down. (Fader down) Especially when your production tracks have no headroom. In this case post is forced to compress your signal in order to bring levels up without hitting 0dBFS. - More work!

It's our job to minimize the work of post as good as possible. So it's best to ask them were the average levels of dialog will be in the postproduction process and record at this level. Also provide enough headroom (maybe you need limiting or compression), so post can bring the level up or down just by using a fader.

As someone who does both location and post sound, I don't agree it's our job as location mixers to "minimize the work of post." I believe it's our job to provide tracks they can use to create the best mix possible. An exception to that might be when post will have insufficient time to do their job properly.

It's the real world out there, after all.

[Glen Trew]

Glen!

You made two general statements I was referring to:

1. "[K]eep the loudest peaks approaching but safely below 0dBfs" and

"if the hottest signal is -6dBfs, then the tracks are at least 6dB lower than ideal"

2. A VU-Meter is "Virtually Useless" because it won't tell you where the peaks are. And you added,

"[that] is why, [...], we have stopped seeing VU meters used in field recording equipment".

.....

...I'm sure that readers with less experience will draw the wrong conclusions and will follow your advice without thinking about it.

As evidenced from this thread, readers have already drawn the wrong conclusions and have misconceptions about metering and levels. Because this thread is going back and forth between the recording of original production tracks and post production, I need to restate that my comments are about the recording of original production tracks that are, later, subject to post production or broadcast. I'll also restate that this topic is complex in terms of both technical and artistic considerations, and needs at least a small book to cover thoroughly. Well, there goes my Spring vacation.

If, while recording dialog, the highest peak was a forceful handclap that registered 6dB below digital max (-6dBfs), this either suggests that the dialog tracks were recorded way lower than they should have been, or that a compressor/limiter was over-used on more than just the handclap. Headroom is useless and wasted if there is no chance of using it. It would be much better to let the rare handclap splat than to needlessly compress or record low everything else. (Going back to the first example of the handclap in this thread, if a director told me to record like this, they wouldn't have to ask me to leave.)

The "ideal" digital recording would, indeed, have the desired dynamic range (an artistic judgement call made by the person in charge of the recording) with it's highest peaks at 0dBfs. Of course, while recording original tracks, it's impossible to know exactly how high the incoming level will be, so we use our experience and best judgement to give ourselves a safety net to allow for surprise peaks. This safety net is known as "headroom". In an ideal recording, the headroom would have been correctly predicted and caused the highest peaks to be at 0dBfs.

My answer to statement #1:

This view is a wrong understanding of digital recording. There is no technical need for getting your peaks up to 0dBFS (using 24bit). You can do so, because it may not compromise your audio signal as long as you don't hit 0dBFS.

First, you CAN hit 0dBfs without any degradation of signal. But more importantly...

The benefit of the increased dynamic range of 24bit digital recording is not to be able to record the peaks lower, but to record the lower levels lower. For example, if a 24bit recording has it's highest peaks at -6dBfs (6dB below max), then its dynamic range is approaching that of a DAT recorder (not exactly, but close). Related example: The output level of a 16bit DAT recorder that has a rated output of 0dBu is exactly the same output level 24bit recorder that has the same output rating. The difference is that the lower level content (and, therefore, the "average" or "general" or "typical" (neither of which are perfect descriptions) content can be recorded lower.

Let me quote from a Sounddevices Manual:

"The 302’s [metering] scale is designed for digital recording devices, providing maximum information between -30 dBu to 0 dBu which is where typical peaks occur (-50 to -20 dBFS). This allows the user to record with a full 12 to 20 dB of headroom ..., the meter color changes to orange at 0 dBu and red at +8 dBu. These color changes correspond to -20 dBFS and -12 dBFS respectively, which are commonly used recording levels for today’s digital recording devices."

The 302 manual in this quote is incorrect and incomplete. The statement "...between -30 dBu to 0 dBu which is where typical peaks occur (-50 to -20 dBFS)." should have added "...when metering in VU mode". But when in VU mode, the meter is no longer a dBfs meter. Reading further in the manual about the meters gives a more thorough explanation of the meters, but, like I said earlier, it would take a book to say enough.

Anyway, when in VU mode, and with a 0VU reference of -20dBfs, peaks will, logically, typically register -20 on a dBfs scale. But these same peaks will register up to 0dBfs on the peak meters used on digital recorders, and may even exceed max. This is another example of why, when recording original dialog tracks to a digital recorder -- as is typical for film and video production sound, a VU meter is nearly useless. I suppose that if the operator could not listen, the VU meter could be of some use in gauging desired dynamic range. But if listening is an option, use your ears to judge the desired dynamic range and keep the max level below 0dBfs.

Ironically, the Sound Devices 302 is an ideal device to prove my point because it has an option to display VU metering and full-scale peak metering at the same time. I've often used it to demonstrate why VU could stand for "Virtuously Useless" for original production tracks. Set the 302 meters to VU and turn the limiters off so the results are not skewed. Then, with a mic about 3 feet away (to simulate a boom setup) have someone talk into a mic, setting their level so that the VU indicator is peaking around 0VU (-20 on the scale). Then switch to peak meter mode. You will see the actual peaks displayed by the peak indicator to be well above the -20 mark, and, if the full dynamic range is used, the peaks will be up to 0dBfs. Do the same thing with finger snaps or a hand clap and the results will be even more dramatic. This drill proves that the actual peaks, as shown on a digital recorder’s peak meter, should be well above the -20 mark.

My answer to statement #2:

VU-meters are no relict of analog times. They are still out there (e.g. in SD Products) and not just "for aligning digital meters with VU meters".

I don’t recall reading or saying anything that suggests that VU meter have no place. It is true that they are just as useful (or useless) as they have always been in certain applications, but when peak meters are available (as they are on all modern recorders) it is also true that VU meters have no use in recording original production tracks that are subject to post production mixing or mastering.

However, I did say “…the only reason for a 0VU reference is to allow the alignment of equipment used in that recording (during production or post production) that may have VU meters. It is not, and was never intended to be a target for [original production track] recording levels” which is true.

One more drill everyone can do that is related to this topic:

With the mixer you normally use, play a tone at -20dBfs. Then, while listening with headphones, set the headphone volume control to where the loudness of the tone starts to feel uncomfortable. I’ll bet most everyone will see that the headphone knob is now set where it usually is. Then set the tone at -12dBfs and set the volume control to where the headphone loudness starts to feel uncomfortable, and you’ll see it’s not quite as high (people with analog Nagra experience will probably see that their volume control is now where it was a couple of decades ago). Now, with the headphone level set lower, ride gain on some dialog to keep the loudness where your ears tell you it should be while keeping the peaks from going past 0dBfs (max). You now have approximately the same dynamic range as during the analog Nagra years (except that the Nagra would allow recording above max without harsh distortion); the end result being that the peaks are the same, but the rest of the recording is, on average, 8dB louder.

Yep… it would take a book. But even the book could be summed up with the statement:

“Use your ears to judge desired dynamic range and keep the peaks safely away from exceeding 0dBfs”.

Glen Trew

[Scott Smith]

Interesting thread., As Glen points out, there are are some staggering misconceptions when it comes to judging headroom and optimizing gain structures for digital recording. The fact that various different levels for the "0 VU" equivalent in relation to "full scale" digital levels have varied considerably over the years, with different standards being proposed for different program production (such as music, which still seems to be in the area of -12 dBfs for "0 VU") doesn't help matters.

While levels for broadcast network specifications vary, this has no bearing on production tracks. Generally, the only complaint I receive from editors and re-recording mixers relates to level being too low on iso tracks, which being pre-fader, have to have more headroom than mix tracks.

Back in the day, most major studios had a standard for maintaining levels from production through re-recording, so at least there was some consistency. While it didn't always work, it was a step in the right direction. We tried the same approach years in Chicago, and published an agreed set of standards between various studios and post houses. While it worked pretty well for interchange of analog material, I shudder at thought of trying to hammer out a similar paper for digital standards, given the variety of different approaches taken by various manufacturers and studios. One can always dream, though....

--Scott

[Joshua Anderson]

I'll also restate that this topic is complex in terms of both technical and artistic considerations, and needs at least a small book to cover thoroughly. Well, there goes my Spring vacation.

Glen Trew

Whatever...I'll wait for the movie version...

These are really great posts, Glen. Thanks for taking the time to find quick and clear explanations.

[Scott Smith]

You were actually considering a vacation!?

--Scott

I'll also restate that this topic is complex in terms of both technical and artistic considerations, and needs at least a small book to cover thoroughly. Well, there goes my Spring vacation.

Glen Trew

[studiomprd]

There is another "wild card"in all of this metering discussion..

the meters themselves are not always "true", that is they may be labeled VU meters, but not really be proper VU meters.

This is common in not fully professional lower priced gear and even up to popular equipment (quick example would be the VU meters on Shure FP series mixers!) Not all of the peak meters are of the same accuracy, either, and the calibration of the dots varies widely...

Let me paraphrase-restate what Glen has said in summation:

The easy answer is that there is no easy answer!

[TheAudioSynthesist]

Okay, so there is a lot of information and misinformation going around in this topic. I typically would have to agree with Mr. Trew. As someone who does location sound and post (for both tv, web, and feature film material) I've come across this topic of discussion with plenty of other sound people out there and there are a lot of mixed opinions. I was once told my tracks were too quiet but they still sounded clean when they were gained or compressed. I've also had other mixers (video editors in charge of sound due to lack of budget) tell me my tracks were too hot. I usually keep my dialog nudging the yellow but never approaching the red and when I do post on sound that I recorded on set, I have very few problems.

For location sound I have only used Sound Devices and Zaxcom recorders so my experience is limited to those products. I learned to keep my levels a little bit hot in production so I wouldn't find myself gaining/compressing in post. I mean, isn't this the right idea? The louder you record your signal in production, the less noise you will get from your recorder right? So, when you're in post, if your production tracks are loud you can easily turn them down further reducing over all noise in your mix. Chances are, when you're recording this way, you won't get quiet tracks in post so that's not usually an issue. I always find that gaining or compressing production tracks in post adds noise to my mixes...

[Marc Wielage]

I was once told my tracks were too quiet but they still sounded clean when they were gained or compressed. I've also had other mixers (video editors in charge of sound due to lack of budget) tell me my tracks were too hot. I usually keep my dialog nudging the yellow but never approaching the red and when I do post on sound that I recorded on set, I have very few problems.

I think "yellow" or "red" are kind of arbitrary readings. Typically, I think what they expect are dialog levels peaking at about -10dBfs, but the problem is where it should average. That's kind of a judgement call.

If it's not clipping and it's consistent day to day, I don't think it's a huge problem. When in doubt, ask the post people prior to the shoot and see what they want. One compromise that I think can work is using a limiter on the mix track, and then record isos at slightly lower levels.

--Marc W.

[studiomprd]

" isn't this the right idea? "

it depends...

as you said: " but they still sounded clean when they were gained or compressed. "

" less noise you will get from your recorder right? "

not exactly...

digital recorders record a list of numbers, one's and zero's, and do not add or subtract anything. So it has a lot to do with the audio chain and A-D conversion, then later, the D-A conversion and audio chain...

The major weaknesses of a lot of the low cost recorders lie in these areas...

[Joel Pinteric]

Thanks a lot everybody, looks like this thread got popular with answers!

Just a reminder though, I was referring the matter specifically in post-production, as stated in my OP.

I.e. Ideal levels in mixing-down / dubbing, including peaks and rms.

This is not about production sound recording!

[Joel Pinteric]

What are your thoughts on compression (mainly dialogue) to achieve a desired loudness, since the peak output is now, in some cases, 10-12dB lower?

I try to avoid compression due to the low dynamic range or noisy environments of some dialogue, but on the other hand it is unfortunate that some mixes seem too quiet without it.

[studiomprd]

" thoughts on compression (mainly dialogue) to achieve a desired loudness, "

ya' lookin' to pick a fight ??

dialog compression?? well it depends, of course, and is mostly a matter of taste and technique (along with good tools) but I tend to prefer a bit of compression, judiciously applied to the dialog mix and mostly gentle, with a touch of make-up gain.

[Marc Wielage]

Just a reminder though, I was referring the matter specifically in post-production, as stated in my OP. I.e. Ideal levels in mixing-down / dubbing, including peaks and rms.

You mean for the final mix? Very complicated question. I think if you surveyed 10 different big-name re-recording mixers and dialog editors, you'd get 10 different answers. My gut feeling is that automated gain riding, a little peak limiting, and (sometimes) a little compression helps, plus EQ can match different mike placement issues. But every situation is different, and not every recording will work with the same approach.

Marti Humphrey's comments on standard mixing levels take into account the various standards (85dB program level for theatrical, 80-82dB for TV and home video), but how to achieve it is largely program-dependent. Two people in a quiet room will require a totally different approach from three people walking on a beach surrounded by waves, or a dozen people in a nightclub with a thumping band in the background. As long as the mixing environment is carefully calibrated, I think several different approaches can work, and at least you'll know the whole show will work in context.

I would do a search on the Gearslutz forum and look for specific comments on dialog compression settings. I know a few mixers that use a "secret sauce" set-up on the C4 compressor within Pro Tools, and this works (with some peak limiting and volume-riding). The trick for me is backing off and not letting the compressor's effects become too noticeable. It's very sobering to look at the huge problems just caused by actors or interview subjects who don't know how to consistently modulate their dialog, especially mumblers who "eat" the last couple of words at the end of a sentence and drop down by like 5dB as they run out of steam. The dialog editors and mixers who take stuff like this and make it work are absolute magicians, especially when they can avoid letting the audience detect the background noise rising and falling.

[Jay Rose]

[quick 'secret sauce' trick for dialog and v/o:

Set the threshold a lot lower than you'd think. But also set the ratio a lot lower. Monitor the GR meter to see that you're still getting just the amount of compression you want, and not too much because of the low threshold or too little because of the ratio.

Then tweak the time constants (ideally, in separate bands).

Assuming a relatively clean source, compression will be a lot more transparent.]

[Joel Pinteric]

Thanks a lot guys.

" thoughts on compression (mainly dialogue) to achieve a desired loudness, "

ya' lookin' to pick a fight ??

dialog compression?? well it depends, of course, and is mostly a matter of taste and technique (along with good tools) but I tend to prefer a bit of compression, judiciously applied to the dialog mix and mostly gentle, with a touch of make-up gain.

I was asking specifically if you receive very hot tracks before mixing, hence why I said "...since the peak output is now, in some cases, 10-12dB lower". For example if you recieved a dialogue track peaking at 0dbFS, but your final peaks must not exceed -10dbFS: you lower it, but then you find it too quiet. That was my train of thought, but I actually just went and tried it and the RMS seemed to follow through fine, I only added a little compression with big spikes.

If however there was the scenario where the attentuated signal was too quiet, but was also noisey, you'd still unfortunately have to apply compression to meet the spec...

[ccsnd]

Please see the thread about this in the post forum. All of these questions were answered. (by me)