Acceptable Recording Levels for SD702T

[Philip Perkins]

As a postie I'll work with whatever the production soundie gives me.  I will probably change the relative level of every piece of audio in the show anyhow, so if you like it way hot for whatever reason then knock yourself out.  What you can't do is CLIP, and fast clips are sometimes very hard to hear when one is in a high-ambient environment and/or is distracted by booming or mixing or whatever.  Pushing your levels up against the top puts you a little in danger of clipping, and dealing w/ clipped audio is a drag in post--some clips can be fixed but some can't, both take time.  I would say that if you are in any form of limiting on your rig all the time that you are prob too hot--a few limiter hits as you go is fine (I have to compress it all anyhow for TV).  If you DO record way hot, also keep in mind that your tracks may clip on some picture editors' sound systems (because their gain structure is screwed up and/or their gear is crappy) even though your tracks are clean.  Another phone call from post....who needs that?

phil p

[Wyatt Tuzo]

If the distance from direct sound to mic is correct then gaining down the track reduces the background sound and the voice (direct sound) becomes more prominent.

Como? I don't mean to sound rude, but how do you figure. Maybe I'm just misinterpreting your wording.

Another benefit of leaving plenty of headroom, is that the digital bus of the NLE won't get overloaded due to the adding of multiple tracks. In a 24 bit system, there is more than enough dynamic range to leave yourself enough headroom without the fear of "not using enough bits".

-Wyatt

[berniebeaudry]

Como? I don't mean to sound rude, but how do you figure. Maybe I'm just misinterpreting your wording.

Another benefit of leaving plenty of headroom, is that the digital bus of the NLE won't get overloaded due to the adding of multiple tracks. In a 24 bit system, there is more than enough dynamic range to leave yourself enough headroom without the fear of "not using enough bits".

-Wyatt

 

I don't know what Como means (did you mean come on?) but I don't take your answer as rude at all.  I'm not recording 24 bit, the cameras I'm usually recording to are 16 bit.  Also, its reality type stuff with none pro talent (HGTV, DIY Network etc.) except of course the host of the show whom I always have on their own channel.  I have my the limiter on my 442 set at -17 so I still have 3db of headroom after limiting.  I strive not to hit the limiter but with hammering, drilling etc. it happens occasionally.  Looking at the camera meters I am hitting peaks of -10 and sometimes a little more and often a little less. 

As for my statement about loud backgrounds getting turned down when the production track is gained down, I think I'm accurate.  If the ratio of direct to background sound is correct and you're gaining down the track the ratio is the same but the levels of both are reduced.  The direct sound seems to become more prominent then.  Maybe its just a perception but it seems to work that way.  I shot a segment in a sugar factory in Hawaii that was the loudest place I've ever been in.  Talent was miked with a lav, and was talking loud  (almost shouting) in order to hear herself speak.  I had a strong level but not excessive and when it went to post and got gained down, the roar of the factory practically disappeared and it made you wonder why she was almost shouting.  I had the same experience at some very loud sporting events.  Maybe my physics are flawed or the way I'm describing it but that's the way its working for me.  Again the editors I work with would rather gain down than up, and as long as I'm not clipping and the levels are not extremely hot they're happy.

I can see how on a movie set with very dynamic dialog and potentially many tracks, 24 bit and leaving plenty of headroom would be desirable.

Best,

Bernie

[Alan H. Chang]

I believe "como" means "come again?"

Bernie, I believe I would have to agree with you on the loudness perspective between ambient/dialogue levels.  AMB would most likely be underneath the dialogue in most situations thus their ratios against the DX.  One thing I've learned is that our hearing perspectives are skewed.  Even when adjusting the gain of the track with both the AMB/DX at the same time, our hearing becomes more sensitive to quieter sounds than louder sounds.  As our hearing reaches the threshold of pain, our ears become less sensitive to the sound pressure at a small linear rate.  For example, the AMB is registering around -50dbfs and the DX is around -10dbfs.  We make an adjustment of -10 on the fader.  Now everything is at -70/-20 and should sound just about the same, but our loudness perspectives are different dependent on the SPL of the material.

[Philip Perkins]

The level relationship between a voice and the BG given all other factors being the same is identical no matter what level you record it at....right?  Just because you turn the channel down in your mixer or headphones doesn't change that relationship at all.  If I bring the level back up in post to hit whatever the show dialnorm level needs to be, that voice to BG relationship will stay the same as it was when you recorded it.  The only wildcard is if you are recording in a way that your loudest sounds (presumably the voice most of the time) is hitting a compressor or limiter, and then you ARE driving the BG up into the voice.  This is another reason to not record ultra hot--the compression or limiting you are doing to stay out of the clips may be narrowing the diff in levels between your voice and the BG....

phil p

[berniebeaudry]

The level relationship between a voice and the BG given all other factors being the same is identical no matter what level you record it at....right?  Just because you turn the channel down in your mixer or headphones doesn't change that relationship at all.  If I bring the level back up in post to hit whatever the show dialnorm level needs to be, that voice to BG relationship will stay the same as it was when you recorded it.  The only wildcard is if you are recording in a way that your loudest sounds (presumably the voice most of the time) is hitting a compressor or limiter, and then you ARE driving the BG up into the voice.  This is another reason to not record ultra hot--the compression or limiting you are doing to stay out of the clips may be narrowing the diff in levels between your voice and the BG....

phil p

So its only a perception that the voice becomes more prominent as the track is turned down in post?  Is that what you were saying Alan? Phil, I don't record ultra hot, nor do I hit the limiter except on occasion.  The person is usually speaking loudly because of the noisy BG therefore I don't have to open the fader as far to get a solid signal to tape and the mic picks up less of the BG as a result.  Of course proper mic placement is key in recording in heavy noise and keeping a good voice to BG ratio.

 

[Philip Perkins]

So its only a perception that the voice becomes more prominent as the track is turned down in post?  Is that what you were saying Alan? Phil, I don't record ultra hot, nor do I hit the limiter except on occasion.  The person is usually speaking loudly because of the noisy BG therefore I don't have to open the fader as far to get a solid signal to tape and the mic picks up less of the BG as a result.  Of course proper mic placement is key in recording in heavy noise and keeping a good voice to BG ratio.

Given the same voice and BG levels, recording the signal hotter or less hot (w/o compression of any kind) will not change the level relationship between the two.  Asking talent to speak up in a noisy environment is a good idea (sometimes) and will help your s/n ratio.  Turning the whole mix down will not.

phil p

[peter775]

What I've always wanted to see in DAWs and NLEs was the option to set a reference level for audio when you start a new project. Then the meters could be toggled between a referenced scale, where the user sees 0 (and aims for it - with the headroom displayed more appropriately) and a dBFS scale when required.

Well. Unlike logic (which isn´t exactly famous for being used for post production) the AVID´s metering can be calibrated with a user-selectable head-room and shows both the absolute digital scale as well as the relative scale of your calibration. It´s just not done most of the time.

My recommendation is to set a fixed auditioning speaker level and not mess with it during work. Give your editor a clip with a narrator´s voice recorded with an average level of -20dBFs. Play that back in the cuting room and adjust the monitoring volume (NOT THE CLIP GAIN!!) so that the dialog sounds at a decent listening volume. Then lock the volume knob of the entire monitoring chain and begin the work and ask them to always monitor at the same level.

By doing this all the elements during work will naturally fall into place just by listening and adjusting the individual levels in the time-line. Editors need to get used to a standard listening volume instead of constantly gaining their clips up and down depending on random volume setting of their b-chain.

The more accurate way of doing it is to import a clip of pink noise at -20dBFs and set the clip gains to unity. Then set the speaker´s volume so that you get a read out of around 79dBSPL© on a radio shack SPL meter. After that lock all the volume knobs. 79 is a good volume for a small edit room.

Once the editor adopts to this method they will have plenty of head room for louder FX if needed and their dialog will always be at a decent listening level when recorded at around -20dBFs. AFTER calibration they can feel free to adjust levels inside their sequence to their taste. But as a result their work will be much less maxed out and compressed. Nice dynamics without everything being slammed to the max.

After this you won´t get any more calls asking for hotter levels unless you are really recording too low.

[Petros Kolyvas]

My recommendation is to set a fixed auditioning speaker level and not mess with it during work. Give your editor a clip with a narrator´s voice recorded with an average level of -20dBFs. Play that back in the cuting room and adjust the monitoring volume (NOT THE CLIP GAIN!!) so that the dialog sounds at a decent listening volume. Then lock the volume knob of the entire monitoring chain and begin the work and ask them to always monitor at the same level.

Hi Peter,

That's a great suggestion! I've always wanted to find a way to lock-off the volume in our "Edit B Suite" (it's a room with a workstation - who am I kidding.)

As for software, Logic was just my example, but Final Cut Pro, Premiere CS5, SoundTrack Pro, Samplitude/Sequoia (last time I used them) and more don't have the ability to set a referenced scale, but I'm glad to know Media Composer and/or ProTools does. And to me, it's in Final Cut where the real problems occur since it's in a tool at work for a varied group of users, some of whom might be talented cutters, but not have much of a clue about anything else.

Anyway, thanks for the great post (about post). I will be testing out the pink-noise/metered levels today and at least marking on the volume knob where it should set!

All the best,

Pk

[berniebeaudry]

Sometimes I use productions SD 442 after someone else has used it.  Often the meter has been switched to vu instead of peak hold which I prefer.  I think if you rely on vu scale metering and don't look to see what actual program is doing at the camera its possible that your tracks would be recorded hotter than you would expect.  The peak hold correlates to the camera meters much better on digital cameras in my opinion and gives a truer indication of level.  On some older analog  cameras with sluggish vu meters I've gotten berated from the camera op for the levels being too low.

Bernie

[peter775]

Sometimes I use productions SD 442 after someone else has used it.  Often the meter has been switched to vu instead of peak hold which I prefer.  I think if you rely on vu scale metering and don't look to see what actual program is doing at the camera its possible that your tracks would be recorded hotter than you would expect.  The peak hold correlates to the camera meters much better on digital cameras in my opinion and gives a truer indication of level.  On some older analog  cameras with sluggish vu meters I've gotten berated from the camera op for the levels being too low.

Bernie

Please note that what I wrote might not always be a good idea for direct to camera ENG work because in this field sometimes a very different approach is required to serve the needs of daily ENG.

What I suggested is what I´d recommend for non ENG work.

[Wyatt Tuzo]

To get back to the perceived level of BG vs. dialogue part of the discussion.

I think what you may be experiencing, Bernie, is the Fletcher-Munsen Curve (or Equal-Loudness Contour) in action. This is the principal that (basically) states that at controlled levels, the human auditory system perceives frequencies between 1k and 5k as "louder" than those surrounding frequencies. The exact loudness perception over the course of the frequency spectrum varies based on the level of reproduction. The flattest human auditory response occurs at a listening level of aprox 85dBSPL. At levels higher than that, the ear and the brain begin compressing certain frequencies in order to protect themselves.

The principal is far more complex, but thats it in a nutshell.

-Wyatt

[Alan H. Chang]

To get back to the perceived level of BG vs. dialogue part of the discussion.

I think what you may be experiencing, Bernie, is the Fletcher-Munsen Curve (or Equal-Loudness Contour) in action. This is the principal that (basically) states that at controlled levels, the human auditory system perceives frequencies between 1k and 5k as "louder" than those surrounding frequencies. The exact loudness perception over the course of the frequency spectrum varies based on the level of reproduction. The flattest human auditory response occurs at a listening level of aprox 85dBSPL. At levels higher than that, the ear and the brain begin compressing certain frequencies in order to protect themselves.

The principal is far more complex, but thats it in a nutshell.

-Wyatt

[Glen Trew]

I often get production audio in post that was recorded as if we were still in the DAT ages. Believe me extra-hot levels like dialog always around -10dBFs or more WILL get gained down in post. Recording average levels at 10dBFs or hotter will ONLY give you less headroom. Your sound will not sound better. The only thing you´re doing is give post a hard time. Less headroom in post for DSP processing. Constantly clipped levels after EQ is applied etc.

On one of my last projects I had to manually and destructively gain down every single clip because It wasn´t possible to work on these files because they were already maxed out all the time. Result: Dialog was back at normal levels but lots of wasted space (headroom) that could have been used if it had been recorded 10dBs lower in the first place.

Leave your average level at around -20dBFs and allow room for surprises instead of slamming everything into your limiter.

It´s a general misconception of many picture editors that hot tracks equal "good" sound. It´s complete nonsense. If they want more volume they need to grab the knob on their monitor system.

Hot tracks equal less headroom and more limiting.

Keep in mind that in cinemas the monitoring is calibrated to 85dBSPL@-20dBFs per speaker. Average dialog level on a final mix is around -20dBFs. It doesn´t make sense to record 10dBs hotter because it WILL be gained down in post anyway otherwise you will end up with bleeding ears.

Again: You gain nothing from recording extra-hot.

What I usually do before a new project is to go to the edit room. Set monitoring levels at around 79dBSPLand tape down the volume knob and ask editors to work with this level.

Asking the recordist to record hotter to compensate for lousy monitoring setups is a bad idea.

I apologize for the lengthy response coming up, but this topic keeps coming up, and is too often debated with generalities and misconceptions.

I dont' think anyone was recommending recording "extra hot", which I assume would mean often hovering next to 0dBfs (max) and, therefore, consistently going over max from time to time. As mentioned above, modern 24-bit records (or 16-bit DAT recorders, for that matter), in no way benefit from this practice. However, much more than that needs to be said regarding recording levels.

The original question asked what "acceptable" recording levels are. Too often, recordists/mixers feel that once over -20 the fidelity of the recording gets progressively worse until maximum is reached. But the fact is that a sound recorded at 0dBfs (max) is as high quality as the same sound recorded at -10 or -20. Also, an original track that does not record it's highest peak at 0dBfs (max) has wasted the available range of the recorder. But don't get me wrong... it is much better to waste a reasonable amount of that range in exchange for the safety of allowing for peaks of unknown levels that, sooner or later, will be encountered. That's why it's called "headroom", which could also be called "margin of safety".

It should not be a production mixer's concern as to whether or not their digital tracks will be re-recorded lower in post. If a recording exceeded digital max when it was recorded, the resulting distortion will not be changed by re-recording it lower. (rare exceptions noted to allow for upward EQ adjustments).

Once it is known that reasonable steps should be taken to avoid digital overload (anything over 0dBfs), then the discussion is about dynamic range (the difference between the low levels and the high levels of a recording) and how to achieve it. With 24-bit digital recording of an original dialog track, a very good argument can be made for setting the levels so that the highest peak will be at 0dBfs, pressing the record button, then putting your hands in your pockets to record the natural dynamics as determined by the actor. In fact, this actually is how prefader isolation tracks should be recorded. However, in the real world this is usually not a practical solution for a mixed track. The loudest peak of a dialog scene is never known until after it has already happened (too late), so it is best to plan to record the highest peak lower than you think it will be, for those times it's actually higher than you think it will be. With multiple mics going into the mono mix track, skillful compressing is usually needed, such as using limiters and/or riding the faders to get a more natural sounding blend. For those who have a problem with compression of original dialog tracks (and I'm sometimes one of them), remember that manually riding gain is nothing more than a crude form of compression with a very long attack time.

Where the term "recording hot" came from and why...

In the days of analog recorders such as the Nagra, recording original tracks "hot" was the best tool in the constant battle to overcome the ever-present tape hiss. That's one reason the 0VU reference point for these recorders was only 12dB below it's maximum rated record level. This was in direct exchange for loss of dynamic range, but considered a reasonable compromise. But analog tape also allowed recording "hot" to the point of consistently recording dialog above the max level, because the distortion created by overloading the tape would increase gradually and be much more gentle than the messy spatter of exceeding digital max on modern recorders. Overloading the tape was usually much more pleasant than the limiter on the Nagra.

To fully cover this topic would require a book. But, briefly, one more reason to not let your recording levels be lower than reasonably necessary is because some editing systems can provide only a limited amount of additional gain (12dB for some), which can be a bit of a nuisance.

In short, probably the best practice is to use skill and experience to reasonably avoid digital max, then use your ears for the rest.

Glen Trew

[hemmerlinj]

Recently it was brought to my attention that the editors on a show that I have been doing had some footage that had audio that was recorded too low and that they had to do a large gain increase with normalizing the tracks.  Luckily it wasn't from any of the days that I was working on, but it definitely was cause for concern.  I personally couldn't find any reason for this occurence. 

I have noticed in the past that the meters on cameras can be grossly different from how it will show up on a different machine.  While doing a docu-reality shoot a few years back.  We shot on the HDX-900.  Tone set to -20 on the cameras.  Mixed so that levels on camera were hitting around -20 on average and peaks at around -10 max. 

After shooting we often would check tapes on the decks we had back in the office.  The other mixer and I noticed that the levels on the decks were appearing about 10 db lower than on the camera.  We couldn't figure this out, except that something had to be off somewhere along the lines.  We never had any questions from post about anything so I would assume everything was fine.  We had like 4 different decks and it was consistently off on all of them.

Jason Hemmerlin

[Eric Toline]

Had any of the decks in question ever been calibrated against a known standard like the old Ampex analog alignment tapes that were used for analog decks?

Eric

[hemmerlinj]

Had any of the decks in question ever been calibrated against a known standard like the old Ampex analog alignment tapes that were used for analog decks?

Eric

I highly doubt it.  They were rentals from VER.  Who knows how much mistreatment they were given.  The only thing that was really giving us a reason to possibly worry was that the meters on the decks had the same difference from the meters on the cameras.  Like I said it was a few years ago and the show has long since gone to air.  We never heard anything from post about anything being a problem.

It makes you wonder with some of the older cameras that people are still using if those meters are showing what you think they're showing.  Is it possible that even though I might set my level on the camera at -20 is really at -20 or could the meter be out of whack? 

Jason Hemmerlin

[studiomprd]

"  could the meter be out of whack?  "

obvious

[Marc Wielage]

My standard is still -20dBfs nominal average with peaks to -10dBfs, and with occasional peaks 2 or 3 dB above that as needed and only barely hitting the limiter (set for about a -4dBfs) with an occasional errant peak.

That is exactly what I normally see in post for 90% of the A-pictures we've worked on in the past 20 years. This is on Tektronix peak-reading 760 digital scopes. (And that includes tracks I've worked on from Mr. Trew [October Road] and Mr. Wexler [*61], all of which sounded fine.)

I have rarely heard an editor complain about production tracks that were too loud, but I have occasionally heard complaints about production tracks that were too soft. To me, the safe bet is to keep the isos another 3-5dB lower than the mix tracks, and I have yet to run into any headroom issues or problems from dialog editors.

To me, the key is to make sure the levels are consistent day to day, so that people in dailies aren't whining or having to change volume during playback.

--Marc W.