Noisy Location Strategies?

[Joel Aaron]

How are you guys dealing with noisy locations for dramatic work?

EVERYWHERE outdoors around here has jets flying overhead every few minutes. There's a lot of traffic noise too. I've seen days added to shoots because the crews kept waiting for sound perfection on every take, every angle. I want to avoid this and keep moving along to the next shot and am trying to figure out a workflow.

Right now I've got a Sennheiser 416 for outdoors, would good lavs be any better? Or is ADR the only way to deal with this? What about ADR on the set? Or wild tracks?

Out of curiosity, does anyone have any idea what % of the average Hollywood movie is ADR'd?

Thanks for any ideas!

[Jon Fairhurst]

I'd say it's best to be prepared for all of the above.

If the shotgun picks up too much noise, use the LAV signal. If the LAVs are dirty, go with ADR.

One trick that I've read about is to record a starters pistol at each location. (Turn down the gain. Use a pad.) You can then use the inpulse in a convolution reverb to help match the signature of the ADR signal to the location sound.

And don't forget to record some clean "silence" that you can loop behind the ADR.

[David Tamés]

Originally Posted by Joel Aaron

How are you guys dealing with noisy locations for dramatic work? [...] I've got a Sennheiser 416 for outdoors, would good lavs be any better? [...]

Considering lavaliers typically omnidirectional, a well placed shotgun as close to the subject as possible is your best bet. I've sometimes used some foam sound absortion to the side or back of the mic in the path of a particularly troublesome external noise I have no control over to reduce it's effect on the recording, and it sometimes helps a bit. If worse comes to worse, you can always use your location audio as a guide track for ADR, but don't let that thinking prevent you from getting the best sound you can on location.

[David Tamés]

Originally Posted by Jon Fairhurst

[...] One trick that I've read about is to record a starters pistol at each location. (Turn down the gain. Use a pad.) You can then use the inpulse in a convolution reverb to help match the signature of the ADR signal to the location sound [...]

Way cool idea! Can you elaborate some more on this?

[Cole McDonald]

I've been looking at sound from a physics standpoint as a way to try to control it better in situations like this and have come up with a concept for a "noise sock" to eliminate background noise. microphone with airspace shock mounted in a pvc tube with airspace in side which is mounted inside another pvc tube with airspace. I say something like this being used at a golf tournament on tv and applied what I knew about making soundproof rooms to concept a way to cancel out all off axis noise. If the microphone is pulled back inside a little bit, the sound collection cone becomes narrower yet. Fur over this will get rid of any wind noise that could transfer to the mic. It'll be heavy, but could be a solution to shooting in really noisy environments. I've not built nor tested these concepts at all...if you do, post pix and recordings.

[Douglas Spotted Eagle]

In addition to the above suggestions, we're doing a series right now inside an airplane that has no door, and no real noise suppression. It's just plain noisy.
So, we're using a lav on subjects feeding a small mixer, all feeding to the left. I'm also using an AT 4053 on a short boom (read boom that belongs on a mic stand) due to small cabin space.

It works surprisingly well. We have no option of lessening noise. No room to maneuver, can't use wires in a jump plane (FAA regulation) etc.

[Douglas Spotted Eagle]

Originally Posted by David Tames

Way cool idea! Can you elaborate some more on this?

Not much more to tell. Find a room/area with the ambience you'd like to use in your recordings, fire a starter pistol or if you have really good clap, or anything else that is really percussive and has no tail, and make the noise.
Record the pistol filre and resultant verb.
Now drop that wave file into your convolution reverb from Sony, WAVES, Cakewalk, or other, and it convolutes it. Sony calls theirs the "Acoustic Mirror."
We did several ambiences for WAVES, a few for Sony, and a boatload of them for Cakewalk's newest reverb. Living in an area that is full of canyons and old mines is a reverberation dream. We even did one of the Grand Canyon from 30 feet down.

[Bob Grant]

Originally Posted by Joel Aaron

Out of curiosity, does anyone have any idea what % of the average Hollywood movie is ADR'd?

I believe upto all of it. Certainly for location.
Given the amount of time getting good sound is eating into your schedule it'd seem like a path to give thought to.

[Steve House]

Originally Posted by Douglas Spotted Eagle

In addition to the above suggestions, we're doing a series right now inside an airplane that has no door, and no real noise suppression. It's just plain noisy.
So, we're using a lav on subjects feeding a small mixer, all feeding to the left. I'm also using an AT 4053 on a short boom (read boom that belongs on a mic stand) due to small cabin space.

It works surprisingly well. We have no option of lessening noise. No room to maneuver, can't use wires in a jump plane (FAA regulation) etc.

Just curious if you've experimented with a second omni mic well separated from the subjects with the lavs recording a noise track that you can phase invert and mix in with the dialog tracks to cancel that aircraft noises there?

[Douglas Spotted Eagle]

Originally Posted by Steve House

Just curious if you've experimented with a second omni mic well separated from the subjects with the lavs recording a noise track that you can phase invert and mix in with the dialog tracks to cancel that aircraft noises there?

No. I'm quite confident it would not work. Too many variables, and while this does work in some instances, it's more the exception than the rule. Omni's and noisy cabins don't mix, I don't use an omni in this situation, ever. Tried once. That was enough.
Remember, I'm the guy that catches crap for preferring cardioids over omnis for the majority of lav situations.

[Steve House]

Originally Posted by Douglas Spotted Eagle

No. I'm quite confident it would not work. Too many variables, and while this does work in some instances, it's more the exception than the rule. Omni's and noisy cabins don't mix, I don't use an omni in this situation, ever. Tried once. That was enough.
Remember, I'm the guy that catches crap for preferring cardioids over omnis for the majority of lav situations.

I was wondering because that principal is used in a number of aircraft noise and vibration control systems as installed hardware in the airframe itself. Mics to pickup the engine noise, rotor vibration, etc and transducers to feed it back phase inverted.

[Emre Safak]

It won't work because the recordings won't be exactly the same due to lag and coloring. What you are describing is called noise cancellation, and it is used, but requires a more sophisticated approach.

[Jon Fairhurst]

Regarding recording impulses from a starter's pistol...

Originally Posted by David Tames

Way cool idea! Can you elaborate some more on this?

The trick here is to get an "impulse" signal that describes your environment and to use "convolution" to apply that impulse to your signal. By convolving a dry signal with an impulse, the result will sound as if your dry signal were recorded in the space where the impulse was made. The results can be stunning.

Musicians can use this technique to put a synthesized pipe organ into a cathedral. Unlike applying a simple reverb, this technique imparts the complexity of the cathedral's character into the result.

Convolution works like this: take a single sample from your audio stream. Multiply it by the current sample of the impulse. That's your first output. Take the same audio sample and multiply it by the next impulse sample. That's your next output. Let's say your impulse has 128 samples. Complete the cycle and your input sample is now spread over 128 samples of output.

Now do the same for your next sample. Offset the resulting 128 samples by one location, and add each sample to the previous result. Do that for every sample of your input stream, and you've now smeared your entire stream by 128 samples.

Let's look at a couple of examples: Say the impulse looks like a haystack. That will smudge your audio around. The high frequencies will get smooshed, while the low frequencies will still come through.

In another case, let's say the impulse has a big, narrow bump and a small narrow bump. That impulse represents the original signal and a dry echo. Make the second bump wider, and the echo will be smeared and have it's high frequencies filtered out.

By recording the impulse from a real environment, you capture all of the complex reflections, filtering and reverb from the site. It even captures the nature of the microphone, preamp and recording system.

If you can get impulses of microphones, you can even remove the signature of your original mic, and apply the signature of the mic you wish you owned. There are some limits to this, and the possibility of amplifying noise, but you can certainly get the flavor of your target mic.

In the musical sample library world, there are those who have recorded the impulse of piano bodies to give a natural resonace after lifting the keys or the pedal. Tascam's GigaViolin records raw strings being bowed, and lets you run the sound through a variety of violin bodies for different tones. Guitar amp simulators use convolution to model tone.

There are a number of products out there that perform convolution. For music, I use GigaPulse from Tascam. Sony's Sound Forge includes their Acoustic Mirror (PC) convolution plugin. There are also products from Waves, Altiverb (Mac) and others.

The thing is, it's really hard to hand tweak the EQ, delays and reverb to match ADR dialog to a real space. If you replace all of the original dialog, it's not a problem. If you want to replace a line or two, you want it to match up. Having impulses and the right tools makes this a snap.

[Joel Aaron]

Originally Posted by Jon Fairhurst

The thing is, it's really hard to hand tweak the EQ, delays and reverb to match ADR dialog to a real space. If you replace all of the original dialog, it's not a problem. If you want to replace a line or two, you want it to match up. Having impulses and the right tools makes this a snap.

Great idea. A lot of helpful ideas in this thread!

So would you record with the same microphone for ADR as you used on the set?

[Jon Fairhurst]

Originally Posted by Joel Aaron

Great idea. A lot of helpful ideas in this thread!

So would you record with the same microphone for ADR as you used on the set?

Not necessarily.

If you're using the impulse method, the set microphone will already be modeled. Also, if you were using a shotgun mic on the set, and your ADR room is a bit live, you'll end up with a "shotgun in a room" sound.

I'm no ADR expert, but you'll want something without too much coloration. If you have an impulse for your ADR mic, you can use convolution to remove its character from the signal chain. (GigaPulse has this feature, but not all products do.) Also, avoid the temptation to record too close to the ADR mic. The proximity effect may sound great, but it won't sound natural over a field shot, and any plosives may be hard to remove.

The most important factor is to record the ADR track as dry as possible. You can always add reverb to make it match up. Removing reverb is akin to compensating for the flaw in the Hubble telescope. It can be done, but only if you have a team of scientists.