Audio Meter says ok, but clipping is still happening?

[Douglas Spotted Eagle]

For broadcast, we're hitting peaks @-6dB, for replication/non broadcast, peaks are hitting -2dB, and as time goes on and oversampling becomes less common, we'll start pushing to fractions.
Our local bureau won't take anything less than -10dB at peak.
If they're stems to be mixed later, they go out at -1dB so that they're huge, and will be mixed to lower levels later. We keep things hot in house to preserve res, and then allow it to go down when it's ready for mix.
Kind of a hold over from the days of 1630, I guess.

[Michael Nistler]

Originally Posted by Arthur Kay

Ok, so what you mean is gain staging mismatch. Now I get it, thanks!
<clip>
So most of you guys use consumer stereo amps? Sorry, if this sounds snobby, really not my intention, but I come from the studio world, so I am not familiar with what the average videographer uses.
Cheers
Arthur

Arthur,

So in your studio world, where do you/they set the sliders on the mixers?(let's assume the audio engineer could also adjust a pre-amp)

A. Above unity
B. Below unity
C. Dead-on unity gain whever possible

Michael

[Arthur Kay]

Hi Micheal,

Not sure I understand what you mean. Maybe I'll try to explain how I record on location, so you get a picture - pardon the pun - on how I do stuff.
If you mean when recording the audio on location, I never record the audio on camera, but seperate (used to use a Nagra, now I use a portable PT rig - macbook pro and 002, connected to a Apogee stereo pre/converter). Maybe not the most efficient to haul around, but it is in my comfort zone (it is what I know to use) and it works *grin*

As far as gain staging on location, the PT is at unity, the preamp gain is set whereever needed, the Apogee converter is calibrated at 0VU = -18 dBFS.
The preamp gain on the Apogee is set to peak anywhere between -12 and -6. MEtering is done with a very good plugin on ProTools from Roger Nichols Digital called Inspector XL. There I use two meters, one VU and one PPM.
So in fact, I record pretty hot for broadcast terms. The final level on the deliverables is done in post. I will aim for an average of -18 dBFS (EBU) and the stations I deliver to allow peaks of -10, if not too many.

In the studio (audio production) the signal chain is as follows (at least where I work - btw, my work place can be seen on www.icpstudios.com).

- Mic > Preamp > compressor > eq > ProTools HD

At mixdown : PT HD > console > Master deck (most of the time a Studer 2 track)

This is pretty simplified though, in reality, depending on the audio source, I will use different things in the chaiin if the source asks for it.

So in audio, I will set the preamp to average at -12 to -10 and peak at -6 to -4. Since I record direct to disk, PT HD is at unity and the gain is set on the preamp.
The final master that goes to the mastering house peaks at -3 to leave some headroom for the mastering engineer.

As far as calibration of the converters is concerned, it depends on the console I am using. In the SSL room (a SSL 4000G) the converters are set to -14 (they are modded, since normally the 192 i/o doesn't go higher than -15 - but we found the signal chain on an SSL sounds better when the converter is set at -14).
On the Neve 88R in our studio B, the converters are set to -16, again, we found that to be the optimum soundwise for the Neve.
On the vintage EMI Neve in room C, the converters are calibrated to -18.

Sorry if this post is a bit long, but I thought I give you a bit of background info, so you might better understand where I come from, and since English is still a foreign language to me, I want to avoid misunderstandings.
I found it is easy on a BBS to be misunderstood, since you cannot hear my voice or see my facial expression when you read my ramblings *grin*.


Have a great weekend

Cheers
Arthur

[Steve House]

Originally Posted by Douglas Spotted Eagle

For broadcast, we're hitting peaks @-6dB, for replication/non broadcast, peaks are hitting -2dB, and as time goes on and oversampling becomes less common, we'll start pushing to fractions.
Our local bureau won't take anything less than -10dB at peak.
If they're stems to be mixed later, they go out at -1dB so that they're huge, and will be mixed to lower levels later. We keep things hot in house to preserve res, and then allow it to go down when it's ready for mix.
Kind of a hold over from the days of 1630, I guess.

Would never debate with your level of experience, Spot, but I thought I understood it yet now I'm getting confused so if you could clarify I'd appreciate it. Where is reference tone/average program material & dialog running when your peaks are that high? I understood that it's common for some of the analog stages in the broadcast chain such as conventional videotape as well as traditional theatrical houses only have about 8-10dB headroom and a master set up with a -20dBFS=0VU reference-level that has an average to peak ratio exceeding 10dB risks driving something down the line into clipping when it finally airs, although digital video, HD, DVD, and some theatrical material can sustain up to a 20dB ratio. That's kind of born out by Arthur's post above where he writes "I will aim for an average of -18 dBFS (EBU) and the stations I deliver to allow peaks of -10, if not too many." Those figures were also quoted in a thread on delivery specs in another venue a few weeks ago where Jay Rose cites -20dBFS/-10dBFS in the network contract requirements he usually works under and it's also echoed in my reference material from Holman ("Sound for Digital Video"). So what accounts for the discrepencies in the numbers - are we talking different meter calibrations (dBVU/dBU versus dBFS), peak versus averaging metering, different delivery destinations, or what?

[Bob Grant]

Spot,
just a minor correction. 0dBFS is actually 0, not 2^16 or 2^24. In the digital realm the quieter the sample value the bigger the number. That makes bit depth conversion very simple and explains why you cannot exceed 0dBFS. Meters can sometimes display over 0dBFS due to successive values of 0. You can as you mentioned also get clipping due to intersample interpolation pushing an interpolated value over the top.

Steve,
the confussion relates back to the old days of analogue and VU meters, typically the meters were set for 20dB headroom but they're average reading meters whereas today typically digital audio is measured with peak reading meters. The problem with peak reading meters is they tell you nothing about how loud something sounds. The problem with VU meters and modern digital audio is they can be thumping the stops when nothing is actually clipping. If your peak reading meters on digital audio are only hitting -10 or -20 you're wasting a lot of headroom.

Back to the original problem. Even though the meters on the camera were not showing clipping mics can be overloaded and wireless transmitters can be overloaded too. I've come accross the latter quite a bit, it pays to setup a wireless mic transmitter carefully. One trap is you can set it up for reasonable levesl at the transmitter with the normal speech levels and then the guy stand if front of a foldback monitor. Nothing you can do at the camera end will help, in fact turning it down can make it even more impossible to fix in post.

To see what's happened look at the waveforms. Digital clipping if things stay digital will always have a flat top, a nice straight line. Analogue clipping has some rounding due to saturation. Well unless it's really extreme clipping but in general you can easily see the difference.

And lets not overlook that sometimes what you're recording can be clipped before it hits your mics.

[Ty Ford]

Originally Posted by Dennis Stevens

........

We had the volume meter on the camera, and we were using DVrack.

What's DVrack? OK I googled. Does the system have a different adjustment for RMS (average) and peak? If so you could have been reading rms and clipping peaks.

Originally Posted by Dennis Stevens

We cranked the volume on the camera all the way. That made it loud, but the meters on the camera and dvrack did not show any clipping.

You cranked the volume all the way? Do you mean the record levels?

Originally Posted by Dennis Stevens

When I play it back on my computer ( I bought decent speakers for it) it sounds fine. Maybe a few bits where the person gets a little emphatic it sounds a bit too loud for one syllable, but not bad. I play it back on Adobe PP2.0, and the volume meters in the NLE don't show clipping.

Playback from what?

Originally Posted by Dennis Stevens

I burned it to a dvd, and someone watched it on their tv. They said they heard clipping.

Can you hear clipping when you play the DVD on your TV?

[QUOTE=Dennis Stevens]
I have to do a lot of research, of course. I'm wondering whether I really have clipping going on, or whether I just need to lower the gain on those bits in post? Do I need to change settings on my camera to get an accurate read of the volume?[/QUOT

Depends on where the clipping ocurred. Listen to the DVD your client has a problem with on your own DVD system (not just the computer) and see if you hear it.

You may not be as sensitive to clipping as your client. If the clipping occurred at the wireless mic during excited moments, there's not a lot you can do. Yoc can clip he wireless and not see overly loud levels downstream, but if you zoom into the waveforms you see the flat topping.

I've had this happen to me. Set the levels and then the person on camera unesxpectedly gets a little nutty and crashes the wireless transmitter. The take away message is to consider lowering the input sensitivity of the wireless transmitter. You'll have more system noise, but more headroom.

Did you end up using the lav track? Compare the wireless and shotgun tracks and see if the shotgun does better.

Another page in the "Good audio is not trivial" case book. :)

Regards,

Ty Ford

[Douglas Spotted Eagle]

Originally Posted by Steve House

Would never debate with your level of experience, Spot, but I thought I understood it yet now I'm getting confused so if you could clarify I'd appreciate it. Where is reference tone/average program material & dialog running when your peaks are that high? I understood that it's common for some of the analog stages in the broadcast chain such as conventional videotape as well as traditional theatrical houses only have about 8-10dB headroom and a master set up with a -20dBFS=0VU reference-level that has an average to peak ratio exceeding 10dB risks driving something down the line into clipping when it finally airs, although digital video, HD, DVD, and some theatrical material can sustain up to a 20dB ratio. That's kind of born out by Arthur's post above where he writes "I will aim for an average of -18 dBFS (EBU) and the stations I deliver to allow peaks of -10, if not too many." Those figures were also quoted in a thread on delivery specs in another venue a few weeks ago where Jay Rose cites -20dBFS/-10dBFS in the network contract requirements he usually works under and it's also echoed in my reference material from Holman ("Sound for Digital Video"). So what accounts for the discrepencies in the numbers - are we talking different meter calibrations (dBVU/dBU versus dBFS), peak versus averaging metering, different delivery destinations, or what?

Ref tone/print tone is always -20dBFS, FWIW, I've written about the asurl=http://www.creativemac.com/2003/02_feb/tutorials/analog_dv_levels.htm] spec[/url] and dialog averages -10dB. Our pieces tend to be much more than dialog, however.
As far as meters, everything is measured on Dorrough PPM meters, which are pretty standard on the broadcast side.

[Jack Smith]

Well, this is a great discussion.A lot of excellent info.
I liked my analogy
" Kinda like driving your car with your foot to the floor on the gas and riding the brake to control speed."
Dennis, did you determine what your situation problem is(was)?

[Stephen Pruitt]

I'm no expert on much of anything, but I'm pretty sure that the clipping involved in the recording had nothing to do with the digital world. I suspect an analog origin or, at worst, origination from the analog portion of the analog to digital conversion somewhere.

I do not know anything about the camera involved, but I'll bet that the camera's meter was showing the DIGITAL RECORDING signal level. Since the analog inputs were maxed out, I suspect that the "clipping" was probably just transistor preamp distortion at the front end. As we all know, tubes distort gently over a wide range whereas transistors or the circuits in ICs (also, of course, transistors) clip catastrophically ONLY when the headroom of the circuit is exhausted. If the meters were showing the RECORDING level rather than the output of the preamps, then it would be very easy to max the transients on the preamps (which were set to their highest level) without ever knowing it.

I've seen this very same thing happen in my own audio studio. I heard clipping all over the place, but I NEVER saw a single red-line event. It turned out the clipping was resulting from me overdriving the analog circuitry of my Lucid 9624AD converter (which did not have any associated metering) on quick transients from drum hits. The digital meters showed fine levels, but that's because they were only watching the digital realm. Except by listening, there was no way for me to "see" the clipping that was taking place in the analog circuits of the converter.

Anyway, that's how I see it.

Stephen

[Bill Davis]

Originally Posted by Douglas Spotted Eagle

I know what you meant, Bill...didn't mean to sound like I was picking on you. I think we all have our pet-peeve words or terms. "Over modulate" when it's not transmission related is one of mine, and then there are the odd-ball descriptives.
Then again...there are those here that wouldn't know a 7.5 ips from a 15ips either... :-) For many, the analog world is an alien concept

Just because I know you'll appreciate this...

Several years ago I shared a cab to the airport from NAB with a couple of 20 something audio software engineers from Silicon Valley.

Don't know how it came up but I mentioned "cutting tape" - their eyes got big as saucers, their jaws dropped, and one of them said something like "you actually used to physically cut tape???!!"

I wanted to tell them that the first TV station I hung out in still had a splicing block for QUAD videotape. It was obsolete even back then, but how quickly the stuff of editing passes - leaving only the skills behind.

Nice reminder of what I teach over and over in my training products, - "editing" is doesn't really take place in oxide or silicone - it takes place a couple of inches behind an editors eyes.

FWIW

[Zulkifli Yusof]

All these talk about setting proper gain settings for recording, appropriate listening levels for playback/mixing and levels for broadcast/theatrical release raises some interesing questions, solutions and tips about audio work.

I've been doing audio for awhile and I dont consider myself highly experienced in the field, but experienced enough to at least do the job. Now, I need to know a couple of things to set my mind straight before I take on another audio project.

Audio mismatch. I understood this completely since I had a very bad personal experience with it before. The problem could be because the microphone output to the recorder was too hot, a case of a bad cable, improper audio connections, or the recording gain setting was too low/too high.

Usually this is preventive as long as you pay attention to your cans and the levels of the recording. During monitoring, headphones should not be giving out its own distortions. If it does, either tone down the listening level or get a better pair of cans.

So am I right to say that you are monitoring the analog signal thus analog clipping (in your headphones; not the levels) are soft transient distortions which are fine, digital clipping is unintelligible and unusable?

Next is the issue of reference tone. I do my work on Pro Tools and as usual, finish it on DAT for layback. So, with my reference tone at -20dB in Pro Tools, at which level should I set the input level of the recorder to if I'm sending the signal through analog connections (-20dB as well or 0dB)?

Lastly, you guys mentioned peaking at -2db, -6db, -10db etc.....what is the usable range for broadcast/theatrical releases so that my loudest sound is safe and my softest sound is still audible?

[Bill Ravens]

Interesting philosophy used by video vs audio people. Any music audio recordist in the industry knows that the push is to maximise volume on playback. This means peaks not to exceed -0.1 dB and RMS values around -12dB. These numbers are after compression/limiting. Curious that the video broadcasters use a -20dB reference. At -20dB, playback requires cranking up the volume control MUCH more than musician is used to.

Not being critical, just noting that beauty is in the eye of the beholder.

[Ty Ford]

-20, of course, is for average (rms) levels with peaks usually up to -12dB.

That's the network, NPR and PBS spec. If you're not submitting to them, you can do what you want unless your client wants something else.

Regards,

Ty

[Douglas Spotted Eagle]

It is indeed the spec. Then there is the concept of "pushing the envelope."
I can count on one hand, the number of projects we've had sent back down due to pushing the required levels by a few dB. New York, Vernon Johns Story, Last Queen, The War that Made America, The Way West...all PBS shows, all Emmy, DuPont, or Peabody winners, and all with pushed audio. There is a difference in acquisition, which is where 99% of videographers screw up due to misunderstood threads like this one, and delivery, which is screwed up also due to misunderstanding the differences between averages and peaks.
Acquisition, you want as HOT as possible without exceeding 0. It's common practice, and common sense.
Delivery, you can do whatever you need, but at least you've got the bit depth to have anything to do with, what you wish.

[Bill Ravens]

Very well said Douglas. You can't put back what wasn't there to begin with, but, you sure can take something out you don't want. Brings to mind an old axiom, "you can't take it with you if you don't have it when you go." There's only one rule of digital audio acquisition....don't exceed 0 dBFS. And it's a fundamental concept to grasp the difference between acquisition of source material and delivery of a product to a customer.