So I was recording this dialog. The subject was recovering from a cold/respiratory ailment but was for the most part excellent in front of camera.

She was wired with a AT822 wired lav, and for safety I had a Rode NTG3 boomed in as close to edge of frame as possible. Lav paned hard left, shotgun hard right. Levels were set normal for lav, and lower for the shotgun for some room ambience if I wanted that later.

So here is my problem, though not huge, but my sensitive ears are making me go crazy. There were a couple of words through out this video series where the first word from her would slightly "distort" as you will hear from my sample. It happened so quickly, I may or may not have heard this while monitoring the audio. Pitfall of being a SPC, but did the best I could. Levels on my mixer never exceded 0 dB.

To me, it sounds more like her voice breaks up just a bit. The audio levels never "spiked" suggesting clipping, but it kind of sounds like it.

So my questions is: what do you think about this, how to prevent or think about this in practice for next time, and finally can it be fixed? I tried to carefully "roll back" the audio for just that one word but that just sounds like the audio dies for a split second.

Thanks for you advice in advance.

Jonathan

PS- I am having trouble uploading a simple .mov file, so here is a link to the file. May have to let it load for a few seconds. Short clip:

http://www.levinphotography.com/ICU.mov

I occasionally get a harsh/gritty sounding word or sound, regardless of input stage and record volume settings... (though if levels are hot, inter-sample clipping is not unheard of, but that is usually very brief and with percussion instruments and transient sounds.
In any case, I normally use the Sound Forge Pro's "Smooth" process (an enhancer with a negative setting). The end result is usually seamless, even to 'trained' ears. I sure some kind of spectral tool could be tweaked to work as well.

I've heard similar sounds live on both speech and singing - I think it's her voice, not the recording. However, I agree that it would be better to smooth it out if possible.

If you get it on both channels, it's probably her voice. If you get it on only one channel, you can track it down to its source through normal debugging.

I don't think it's the voice - sounds too "electronic" to me. If it only happens on the first word after a silence could it be AGC rearing it's ugly head somewhere? Sounds like she puts a lot of sudden stress on the first word and a weak link in the chain is "ringing" until it settles down.

What did you use to record this, besides the mics? Are both channels represented in the file - I see two audio tracks but the two mics certainly wouldn't be giving you stereo. I do notice the problem regardless of which track is muted.

Wish I had a good answer instead of more questions.

I've been staring at the first word in Izotope - I think the first word looks rather strange. In fact I think the whole thing looks a bit odd.

I've attached a screen shot.The striation pattern in the orange is quite noticeably different for the first word. Not sure what this is trying to tell me.

A few more questions to add to the equation:

Which mixer are you using? How was the input trim set for both of the mic inputs? Among the many places where distortion can happen and not show up on the meters is at the mic inputs if the mixer input trims (or gain depending on how the mixer is labeled) are set too high. Some mixers will flash a peak LED or if you activate the level trim indicators on some mixers you'll see how much or how little headroom you have left on the input as you make the adjustments. Some mixers don't have either of these features.

The same is true for the camera input if the mixer output is too hot for loud transient sounds.

Also what was the recorder or camera used? It could be a limiter or compressor setting that isn't reacting fast enough for the first word.

Lastly, which lav? The AT822 is the older unbalanced stereo mic, related to the AT825 balanced stereo mic.

OK - I tried to more or less mouth the same words with a heavy emphasis on the word "Last"

A few things stick out

1) I'd make a lousy voice over guy (sorry)

2) The Izotope plot looks more like I'd expect vocals to look. Much smoother and less "striated" energy distribution at the lower frequencies, and relatively little energy at the higher frequencies as well as a bit less abruptness at the start of words. You can clearly see the way the frequency distributions shift upward during the pronunciation of the first word compared to the flatter more parallel nature of the plot of the original recording.

You can also see that I was sitting a bit to my right of the mics while speaking

Setup:

M/S Schoeps figure 8/wide cardioid into an SD 702 at 96kb 24 bit

Something is filtering/limiting/generally messing up the original but I have no good idea yet what it could be.

FWIW I guess that's all I can do at the moment.

Thank you all for your advice!

Rick-
can't quite afford Sound Forge right now but I did look into that. Holy crap, as if sound couldn't get more involved, I almost forgot about all that post production "sweetening" that goes into the end product.

Colin-
Though almost fully recovered, there were a few raspy utterances here and there. When there was shorter dialog, I usually did more than one take. But some of the dialog went on and on, even with breaks, and time was limited, especially with teleprompter operator.

Bruce-
It is indeed on both channels. I thought maybe some weird artifact from either the lav or the shotgun that I could then just swap out the one word(s) real quick.

Jim and Jay-
In the middle of this project I switched from using a Shure FP33 connected to a Marantz 660 recorder. I was also recording a scratch track to dslr. I gain staged everything (if I am using that term correctly). Even the scratch track has this sound, so that leads me to think it is just her recovering voice. For the second part of the project I acquired a Sound Devices 633 which I absolutely love! The difference between 16 bit and 24 bit files is amazing. And I CAN hear the difference. I am wondering if this would have been an issue using the 633 right from the start? Noise floor on 633 is almost un-noticeable whereas the FP33/660 combo is really to noisy for my taste. What ever that graph/chart thing is, it sure is pretty!

Jonathan

Can't believe the sound was due to a cold. There are clearly frequencies missing/ overemphasized. I don't think a human voice can sound like that absent some recording problem. Sounds sort of like "The Chipmunks" of decades ago.

The first thing I notice is that both channels of your .mov sample are exactly the same audio. Nothing is panned hard left; nothing is panned hard right. I see this visually on the waveforms provided by Andrada. When I analyze the file on my own system, I can see that the levels on L and R are exactly the same: (L-R) = 0. (Doesn't anybody else notice this?!) So there's no way to make a comparison between your two mics.

I don't know whether your sample is just one mic, mixed to both channels, or whether it's a mix of the two mics. But since you want us to evaluate and compare the two mics (to try to identify the problem) I think you've done something that you didn't intend to do.

The fact that this is mixed also makes me wonder what else was done to the audio, between your original recording and the file you provided for evaluation. Might you have inadvertently applied some EQ? Some limiting? Who knows what else? I tend to agree with Mr. Andrada in that I suspect AGC might play a role in this.

I would like to hear this same section of the track, but with the two mics on separate channels, and with several seconds of silence before and after this section (if, in fact, such silence exists on the original).

Other than that, there is a pronounced noise before the very first syllable. Sort of a subdued buzz with a fair amount of HF content around 250Hz, 500Hz, 750Hz, and at a lower level 1,000Hz. The higher frequencies are not present later in the track, although the 250Hz is. Perhaps this noise is mixing with the voice frequencies and producing some IM distortion.

Please try to sort out why your provided sample is not the raw original, and then provide that instead ... that will be a lot more useful.

Jim- I know....

Greg-
I was having trouble uploading my .mov file for some reason. I went through a couple of Apple Compressor settings to try and get it to upload, thus the reason for the hyperlink. Not sure exactly what "version" I uploaded.

The original recording had one mic (lav) on one input/channel and a second boomed mic (shotgun) just out of frame from top. Lav panned hard left, SG panned hard Rt.

If someone can tell me what I am doing wrong. I don't think I can upload a ProRes 422 .mov file, even that short clip would be monstrously giant. I think I could get a seperate file for each channel, and I also have a few other instances of this, particularly with the word "Apply" Let me work on this.

Thanks all for looking into this.

JL

Also give us a clue on how the signal got to your dSLR for the scratch track that also has the problem.

You didn't specify which part of the project this portion came from, the Shure mixer and Marantz recorder part, or the Sound Devices you used later.

I'm hoping it was from the earlier part. It wouldn't be hard for me to imagine some interaction with the Shure and Marantz being the problem.

Jay and everyone:

I just uploaded a new sample that better illustrates whats going on. Today I was able to upload. Weird.

There are two files: Scratch, audio from Nikon D800e and LavLeftNT3Right, audio that was recorded from Shure FP33 to the Marantz 660.

BTW- I had a cable going from the headphone out on the recorder to the audio input on the Nikon, probably has a lot to do with the shitty sound.

And yes, the audio from the SD 633 is glorious.

The word, as if you won't hear it is "APPLY..."

Thanks again.

Jonathan

Jonathan wrote:
<<The original recording had one mic (lav) on one input/channel and a second boomed mic (shotgun) just out of frame from top. Lav panned hard left, SG panned hard Rt.>>

The files you just provided still are not lav panned left and shotgun panned right. Both tracks have exactly the same audio, which is either one mic or the other, or a mix of both. You really need to review how you are sending the tracks to the recorder, because it seems they are not making it there in the manner you describe. Perhaps you can give a step-by-step, connection-to-connection description of how you've got things set up.

As far as the word "apply" goes, it may be just my ever-more-ancient ears, but I don't hear anything weird on that word other than that upper midwest, slightly nasally accent sneaking in.

"can't quite afford Sound Forge right now but I did look into that."
- The lower cost Sound Forge Audio Studio 10.0 has the Smooth/Enhance process, but may not be worth it for one process. It's certainly useful for many other audio tasks as well though.
Sound Forge Audio Studio 10 Tech Specs (http://www.sonycreativesoftware.com/audiostudio/techspec)
I'm sure there are free or low cost 'smooth' plug-ins available that would do the job as well. Even attenuating the offending frequencies via EQ would likely work..

Jay: " It wouldn't be hard for me to imagine some interaction with the Shure and Marantz being the problem."
> I had a 660 for a while and would concur.
.. If the FP33 was feeding the 660 recorder, how was it hooked up? (XLR> XLR, mic or line? Tape out to 660 1/8" in? The 660's input stage clipped very easy.

Rob and Rick,

Thanks. Listen toward the end of the word apply. Maybe I'm way nit-picking this thing, but sounds like shit to me. And I'm willing to bet my ears are aged more than yours!

As far as the audio files, I am not sure they are transferring the way you mention. In my editing program everything is just the way I mentioned.

Set-up: Audio Technica 899 lav going to input 1 on FP33. Rode NT3 into input 2. Input 1 panned hard left, 2 panned right.

Mic level out from mixer to mic level in on 660 via xlr. Headphone out from 660 to 1/8' audio input on Nikon D800. Headphone out on D800 to Monitor In on FP33 to monitor audio. Gain Stage everything.

In post, FCP X, raise level of Lav and lower NT3 just for ambience.

Jonathan