Hi,

I have a piece of audio that has a hum or whatever this is called and I don't know how to remove it. I do have Izotope RX but can't figure out how to get rid of it. I am attaching a clip. If someone could tell me what exactly that sound is called and how to best remove it that would be awesome. If you have tips on how to remove it using Izotope RX that would be even better.
https://www.dropbox.com/s/7tjvmom6zwbdphi/Tom.aiff?dl=0

Kathy could it be your audio playback hardware? The audio clip sounds clean to me, I'm either going deaf or there isn't any hum?

Same here. Clean
"I'm either going deaf or there isn't any hum?"
Same here as well, I even ran it though a analyzer to check.

I hear a tiny bit of hum right at the very end when the subject finally takes a breath(!) But it seems VERY far down from the average signal levels. And it seems more like environmental buzz (like from a fluorescent lamp ballast or something) than a traditional mains power hum.

I wonder if we are hearing the same thing? With such a very short snippet, it isn't all that easy to judge, but it seems like it is so much lower than the dialog, it isn't worth bothering with. Perhaps a better sample would illustrate what your think the issue is?

If you are trying to eliminate a "buzz" (vs a "hum") then you must also go after the harmonics up the spectrum. For example Adobe Audition has a filter for 60/120 Hz and another for 50/100 Hz which include notches for at least half a dozen octave-spaced overtones above the fundamental.

Yes. It's that buzz sound I'm referring to. It sounds like fluorwscent bulb as Richard said.
If this is not worth rwmoving then I'll leave it as is.
Now next question. How do I make this sound better?

I don't have a good playback system available at the moment, but, to my ear, the three things that could be troubling are:

1) It sounds slightly over-driven. Dunno where along the audio chain that happened, and there isn't much you can do about it within a reasonable cost/benefit ratio. I would think it is the kind of thing that could be easily avoided if monitored properly during recording.

2) It sounds somewhat compressed. Perhaps a first-cousin of #1 above. Again, not so bad it would be worth taking extraordinary measures.

3) It sounds just a bit reverberant. But, without seeing the video and knowing what the expectations of the producers are, very possibly well within the limits of perfectly acceptable. There are new software tools becoming available to significantly reduce reverberation, but they seem fiddly and extremely expensive and limited only to acute cases until the cost descends to where most of us live and work.

4) It sounds a bit "tubby" (mid-low frequency bump) But that could very well be attributed to the crummy headphones I have at the moment. Tonal balanced is something that would be pretty easy to apply, and to listen to the results and "season to taste".

I've listened to the clip on my home computer which has a pretty good set of JBL consumer speakers.

If you thought it was bad enough to come to the forum, then I'd have to ask like Alex, could it be something in your own playback system?

I don't detect anything out of the ordinary beyond the minor characteristics that have already been mentioned.

I'm guessing this was recorded with a lavalier mic?

yes, lav mic. Well, I don't think it's bad but I am becoming anal about sound so I thought professionals would want to remove that buzzing sound.
by over-driven do you mean clipped?

No, if it were clipped you would know it without asking.

Without knowing anything about your audio path, we can't offer any speculation about WHERE it might have happened. But it was in some linear/analog circuit BEFORE it got digitized into ones and zeros. It might even have been the microphone itself.

We also don't know if/how the audio was monitored, so we don't have any forensic evidence of what happened during recording.

Can you list the complete audio path during the recording?

And any settings like compressors or limiters used in editing?

It looks and sounds to me like it ran up against a hard limiter at some point. Perhaps an all or nothing limiter switch on a mixer or the camera/recorder.

If that is what happened, then it's possible to keep pushing up the signal feeding into the limiter and it won't allow any more peak signal through, you're just pushing up the background noise and compressing the overall voice. Making it sound a little strident and over-driven but without hitting any point that actually clipped.

I've got studio reference monitor speakers and it sounds great. No hum at all and a very clean recording.

I removed the slight echo in the sample recording. Because I could. :-)

Grab the updated audio from here: https://www.dropbox.com/s/q6hwkqo90tzqvkb/Tom%20no%20echo.aif?dl=0

Andrew

PS. There is a tiny bit of hum / consistent looking noise that you can see on a spectral view of the audio, but it's waay down at the -56db level. Not worth worrying about.

Interesting noise. It's not a pure single-frequency hum, nor is it typical power line "buzz" with harmonics every 60 Hz. It is predominantly 120 Hz, 360 Hz, and a small bit at 960 Hz.

Peak level of the noise is about -48 dB. Peak level of the voice is about -4 dB. That makes the S/N ratio -44 dB, which is audible at high normal listening levels (confirmed by my ears).

Looking at a spectral view of the clip, the noise is also present in the small pause around 3.4 to 3.7 seconds, but you don't much notice it there because it's masked by the breath and other ambient noise.

As many have said, no way to guess where it originated, without knowing a lot more. However, it is very easy to remove using the noise sample process in Cool Edit or Audition. Since the level is so low to begin with, it takes very little removal to eliminate it entirely, without any audible artifacts.

Thanks for the responses.
The mic was connected to the camera using XLR cable. I am using this EMP4S: Ambient Recording GmbH (http://www.ambient.de/en/products/ambient-recording/mic-power-supply/emp/emp4s.html) adapter in order to connect the mic using XLR cable.

I would put that adapter pretty low on the list of suspects.
Exactly what microphone? What camera?
Where was the microphone placed?
Was the subject's speaking voice louder than average?
It would be almost impossible to reveal TOO MANY details here.

I would put that adapter pretty low on the list of suspects.
Exactly what microphone? What camera?
Where was the microphone placed?
Was the subject's speaking voice louder than average?
It would be almost impossible to reveal TOO MANY details here.

Microphone: Sanken COS-11D
Camera: Canon C100
Mic was clipped to his jacket on a right side, around 3-4inches down from his chin
I don't recall him speaking louder than average
I had peak limiter on.

What did the audio levels look like during recording? We can tell from the audio clip that the limiter was engaged. But it sounds like the audio recording level was set too high, forcing the audio up against the limiter nearly all the time.

The limiter is provided as a safety measure to avoid clipping. It is NOT an "auto-level" feature that allows us to ignore setting proper recording levels.

Richard,

I think you make a good point about excessive limiting. That is probably why the unwanted noise ramps up after the last spoken word. The limiter is releasing then, and gain is coming back up, bringing the noise level up with it.

"[/i]it sounds like the audio recording level was set too high, forcing the audio up against the limiter nearly all the time"[/i]
- Louder than normal PB levels as well.. Monitoring at very loud levels will produce all kinds of extraneous noise... and fry your ears into hearing things that really don't exist.
Calibrate your monitors to a known reference level and keep them there.

When I'm monitoring on headphones, and seeing the levels on the camera display, where should the level's fall?

The audio levels should ideally bet between -12 and -18 dB, though you should also check peaks to make sure that you're not exceeding 0 dB for more than a transient.

If you're using a mic that isn't terribly sensitive, you might have to use a fair bit of gain to get to that level, in which case you're likely to end up with some noise from the preamp in the recording.

If the recorded noise if fairly mild (I haven't had a chance to listen to your audio sample yet), you might even be able to take care of it with little more than a carefully tuned multiband compressor.

This sounds like ordinary "room tone" to me, something every mike will pick up if it is sensitive enough.

One of the first things I was taught, fwiw, was to always record 30 seconds of room tone when doing audio, for patching over excisions in the finished audio --- total silence in a room sounds unnatural, and would be noticed if you edited and just dropped out the total audio. It is also such a low level sound that I wouldn't be concerned about it.

It doesn't sound like a gear problem, just the natural sound of the environment.

"seeing the levels on the camera display, where should the level's fall?
I prefer peaks @ around -6.0dBFS, with a high quality limiter in place. But 12-18dBFS would be good, especially with 24 bit.

The audio sounds distorted a bit, like the input level on the transmitter was too high or the gain on the receiver. But nothing serious. It could sound a bit better with some high cut and some other tricks, but not much. Background sound and/or music would hide most imperfections I think.

When I'm monitoring on headphones, and seeing the levels on the camera display, where should the level's fall?

Generally: As HIGH as you can manage without WITHOUT going OVER the limit (0dB Full Scale)
In practical terms it depends on WHAT you are recording, HOW you are recording it, and WHERE (under what conditions). It is not something easily boiled down into a general command. You get the FEEL for where the levels should be after practice with your gear and your program material.

Now with SOME gear, you can increase the record level to the point where it would be clipping, but the limiter is artificially reducing the record level to prevent clipping. This has the effect of squashing the dynamic range, and also pumping up the background noise in gaps between words, etc. That is what it sounds like happened to the audio in your example. Without the limiter, it would be terribly clipped and you would know instantly that the record level was too high.

It would probably be instructive to try recording some things just for practice WITHOUT the limiter engaged so that you can see what the dynamics actually are. And you can experiment to see how the signal-to-noise ratio is compromised when you record too low, and how the audio is clipped if you go too high. There is a "Goldilocks Zone" where the record level is "just right".

Of course, you must also apply some amount of judgement to allow some "headroom" for expected (and unexpected) audio level peaks. Again, you must develop a "feeling" for this based on what you are recording. A "talking head" lecture from someone who has delivered the same speech 100 times before may be pretty predictable and leveled-out. OTOH, a live performance of something really dynamic (and/or amateur performers) is almost guaranteed to have some significant peaks over the average audio levels. THIS is what the limiters are there to catch. The limiters are NOT intended to be used as an "auto-level" function. That would be like abandoning steering and depending on the guard-rails to keep you on the road. You would probably arrive at your destination, but your car will be pretty banged-up from riding up against the rails all the time.

"That would be ilike abandoning steering and depending on the guard-rails to keep you on the road."
- Good analogy Richard!

OK, thanks for all the help so far. Here is another person recorded in the same space with the same equipment. Does it this recording show the same problems as the other one I posted before? What can I do to improve this voice? Or should I leave it as is?
https://www.dropbox.com/s/i0npkqf1dsyeztn/Jack.aiff?dl=0

Did you use some sort of noise gate on this? Or was it manually edited? For example between 8:02 and 8:38 seconds, after the word "ways" there seems to be a momentary dropout of the room tone around the breath noise. Then the room tone returns when the speech resumes.

For that matter, I feel as if the room tone (that predominant 120 Hz background noise) is pumping up and down in level throughout the duration of the file. Did you have any kind of AGC applied during recording (compression? limiting?) or applied in post?

Greg,

This is supposed to be "RAW" file but it is exported out of Final Cut without anything applied to it. The only thing that was set was peak limiter. I do hear that dropout so I am now wondering myself what it was caused by. I will go back to the recording again and double check to make sure there is nothing applied to the file when I export. I am exporting this out of cut sequence in Final Cut.

There also seem to be some extremely brief drops around 4.61 and 4.70.

Exactly what all was in the recording chain, starting with the mic?

Microphone: Sanken COS-11D
Camera: Canon C100
Microphone was wired to the camera with XLR cable
Mic was clipped to his jacket on a right side, around 3-4inches down from his chin
I had peak limiter on.

Are the levels of the exported file (loudest peak is ~ -13.5 dB at ~ 3.25 seconds) the same as the levels of the original file as imported into Final Cut? If so, then I don't think the camera's limiter is involved, because the manual says it starts to limit at -6 dB and you are nowhere near that.

Did you have gain control set to manual, or to auto?

---

Also I'm a bit puzzled about what you're giving us. I see both channels are exactly the same. If I subtract the channels to produce (L-R) there is absolutely nothing there. Every sample is exactly zero. If you had recorded two stereo channels, even with the same mic connected to both, I would expect some data in (L-R) because the two preamps would be slightly different, encoding would be slightly different, etc. So I'm curious about what transforms were applied to the camera audio, before you uploaded the file to the web. At the very least you must have created a mono mix in FC, with both both output channels = (L+R), which would be different from raw data from the camera. (Or perhaps I don't fully understand what this camera does if you give it just one input channel.)

Are the levels of the exported file (loudest peak is ~ -13.5 dB at ~ 3.25 seconds) the same as the levels of the original file as imported into Final Cut? If so, then I don't think the camera's limiter is involved, because the manual says it starts to limit at -6 dB and you are nowhere near that.

Did you have gain control set to manual, or to auto?

---

Also I'm a bit puzzled about what you're giving us. I see both channels are exactly the same. If I subtract the channels to produce (L-R) there is absolutely nothing there. Every sample is exactly zero. If you had recorded two stereo channels, even with the same mic connected to both, I would expect some data in (L-R) because the two preamps would be slightly different, encoding would be slightly different, etc. So I'm curious about what transforms were applied to the camera audio, before you uploaded the file to the web. At the very least you must have created a mono mix in FC, with both both output channels = (L+R), which would be different from raw data from the camera. (Or perhaps I don't fully understand what this camera does if you give it just one input channel.)

Hi Greg,

Sorry for the mess with this file. What I uploaded was not raw. Here is the file NOT from Final Cut. All I did to this one was trimmed the original to create a sample file in Apple Logic. This was recorded to one channel on C100.
I want to know if there is anything "wrong" with this recording, and how to make it sound "better". To me it sounds distant and echoey. I don't believe peak limiter engaged during the recording.
https://www.dropbox.com/s/nv50i8ft7gfwmdu/Jack2.aif?dl=0

OK, this finally sounds like raw audio, so we aren't analyzing something irrelevant.

I think I still hear some of the background hum, which was predominantly at 120 Hz. Of course it's hard to tell, because you have not included any silence (i.e. no speaking) with just room tone. That is always very useful when you're trying to evaluate a recording. I would bet real money that it's still there.

I believe I hear some clothing rub at various points. Of course the talent might be moving around some props, there's no way to tell without accompanying video. If it is clothing rub, there are ways to deal with that.

It does not sound especially "distant" to me, the presence is there. But the room does sound a bit reflective and echoey. I have heard a lot worse, even on network TV. I guess it depends on the nature of the finished product. If it's more or less documentary in nature, or even educational, of someone speaking in a classroom or a big room, IMHO the sound will be characteristic of what people are seeing, so it won't be a problem. It does not reduce intelligibility. Of course if it's supposed to be a close and intimate narration, it would detract a bit.

The only ways you can reduce that echoey characteristic are either to add a lot of absorption to the room, or to use a much more directional mic (let the discussions begin!), carefully aimed at the talent. Or to do both. Your choice will be somewhat determined by that kind of visual shots you need to get.

Here's a spectrogram of lower frequency region.

OK, this finally sounds like raw audio, so we aren't analyzing something irrelevant.

I think I still hear some of the background hum, which was predominantly at 120 Hz. Of course it's hard to tell, because you have not included any silence (i.e. no speaking) with just room tone. That is always very useful when you're trying to evaluate a recording. I would bet real money that it's still there.

I believe I hear some clothing rub at various points. Of course the talent might be moving around some props, there's no way to tell without accompanying video. If it is clothing rub, there are ways to deal with that.

It does not sound especially "distant" to me, the presence is there. But the room does sound a bit reflective and echoey. I have heard a lot worse, even on network TV. I guess it depends on the nature of the finished product. If it's more or less documentary in nature, or even educational, of someone speaking in a classroom or a big room, IMHO the sound will be characteristic of what people are seeing, so it won't be a problem. It does not reduce intelligibility. Of course if it's supposed to be a close and intimate narration, it would detract a bit.

The only ways you can reduce that echoey characteristic are either to add a lot of absorption to the room, or to use a much more directional mic (let the discussions begin!), carefully aimed at the talent. Or to do both. Your choice will be somewhat determined by that kind of visual shots you need to get.

Greg thanks! I was wondering how I can improve this recording after the fact. I am not looking for a solution to remove clothing rub. I am looking for an overall improvement. Can I de-echo this? Does it sound like there is too much bass?

Kathy,

I'm afraid I'm an "old school" guy. I think the way to have echo-free audio is not to record the echo in the first place. Fortunately I've never been up against the wall where I needed to try to remove echo after the fact. So I've never tried any echo reduction software. I did hear at least one demo that sounded somewhat convincing, but I honestly do not remember what program (or plugin) that was. The topic has been discussed here often enough, so hopefully someone else will jump in and give you a specific name. Or else use the forum search function and look for fairly recent posts, because IIRC I heard the demo fairly recently.

I've listened to this on a few different speakers, and on Sennheiser HD-280 Pro cans, which I think are reasonably accurate and flat (compared, for example, to the popular Sonys). I do not feel the file is the least bit bass heavy. Of course someone else might have a different opinion. What are you using to listen?

BTW, I do seem to hear a markedly unnatural quality to the HF portion of the audio. It really jumps out at me when I listen. It's very obvious in the ending "s" on the word "dynamics" from roughly 0:01.58 to 0:01.95 seconds (and in many later sibilants throughout this sample). You can see it in the attached spectral view, and also in the frequency scan, of this time slice. There seems to be a resonance around 5 kHz. More than that, there seems to be a significant notch from 2 kHz to 4 kHz. I don't know whether this was caused by the mic itself, mic placement (maybe reflections from a very close hard surface) or what's going on. In any case, I think I am hearing a predominant frequency around 5 kHz, rather than a nice smooth range up frequencies. To me, it sounds bad and what I hear seems to match what I see in those two displays.

Of course this could have been produced by some very badly applied EQ, but if you're sure this is a raw file, then we can rule that out. Try another mic? Clip on two mics at the same time, record one to each channel, analyze the raw stereo track?

EDIT: Are you sure this is raw? The camera literature specifies a 48 kHz sample rate. The latest file (Jack2.aif) appears to have a 44.1 kHz sample rate. If both of those statements are correct, and "Jack2.aif" is actually a converted file (not raw), then can we be sure that there wasn't some other transform that might have changed the audio quality?

Specifically, was the mic out in the open? Or was it buried under clothing?

Aside from that, how does this adjustment strike you?

Kathy,

I'm afraid I'm an "old school" guy. I think the way to have echo-free audio is not to record the echo in the first place. Fortunately I've never been up against the wall where I needed to try to remove echo after the fact. So I've never tried any echo reduction software. I did hear at least one demo that sounded somewhat convincing, but I honestly do not remember what program (or plugin) that was. The topic has been discussed here often enough, so hopefully someone else will jump in and give you a specific name. Or else use the forum search function and look for fairly recent posts, because IIRC I heard the demo fairly recently.

I've listened to this on a few different speakers, and on Sennheiser HD-280 Pro cans, which I think are reasonably accurate and flat (compared, for example, to the popular Sonys). I do not feel the file is the least bit bass heavy. Of course someone else might have a different opinion. What are you using to listen?

BTW, I do seem to hear a markedly unnatural quality to the HF portion of the audio. It really jumps out at me when I listen. It's very obvious in the ending "s" on the word "dynamics" from roughly 0:01.58 to 0:01.95 seconds (and in many later sibilants throughout this sample). You can see it in the attached spectral view, and also in the frequency scan, of this time slice. There seems to be a resonance around 5 kHz. More than that, there seems to be a significant notch from 2 kHz to 4 kHz. I don't know whether this was caused by the mic itself, mic placement (maybe reflections from a very close hard surface) or what's going on. In any case, I think I am hearing a predominant frequency around 5 kHz, rather than a nice smooth range up frequencies. To me, it sounds bad and what I hear seems to match what I see in those two displays.

Of course this could have been produced by some very badly applied EQ, but if you're sure this is a raw file, then we can rule that out. Try another mic? Clip on two mics at the same time, record one to each channel, analyze the raw stereo track?

EDIT: Are you sure this is raw? The camera literature specifies a 48 kHz sample rate. The latest file (Jack2.aif) appears to have a 44.1 kHz sample rate. If both of those statements are correct, and "Jack2.aif" is actually a converted file (not raw), then can we be sure that there wasn't some other transform that might have changed the audio quality?

Hi Greg,
Sorry for a late reply. I was away. I opened the file in Apple Logic and it got converted to 44.1 kHz. I just want to trim it so I wouldn't post 2hrs of audio. I don't believe anything else was applied to the file. How can I trim and still call it RAW?

Hi Kathy,

If you're sure nothing else was changed, then I guess it's close enough to call it "raw."

My concern is that there seems to be a lot of attenuation between roughly 2kHz and 4kHz. (I say that because the frequency content of the file drops off abruptly around 2kHz, but then seems to come back up around 4kHz with a bit of a peak around 5kHz.) Since the file was obviously "processed" in some way (as indicated by the change in sampling frequency), I wanted to be sure that you hadn't inadvertently applied some EQ in the process.

Of course I can't prove that those frequencies are really attenuated, compared to the actual live sound. But based on what I hear (comparing your "raw" file to my "adjusted" file), and compared to what I see in the two visual representations of the file, it certainly seems to me as if something went awry.

How does my "adjusted" sample (in my last post before this one) sound to you?

Do you have any clues as to what might have happened to that bit of the frequency spectrum? Was the mic out in the open, or was it buried under clothing? Things like that?

Any chance you have another mic of the same make/model so you could record a stereo test file, with each mic feeding a separate channel?

*** What does anyone else think of this file? How does the original file (from post #32) sound to you? Is it lacking in presence? Do you think it's weak in the 2kHz - 4kHz range, compared to lower frequencies and also compared to the peak around 5 kHz? If so, any idea what might have happened to that range of frequencies?

Hi Greg,

Sorry again for such a late reply. I listened to your sample and it sounds pretty good to my ear. Since my last posting I had to redo my video removing some parts and adding others. I think I'm all set with it. After adding a music track it all sounds decent.

Thanks

Hi Kathy,

Glad you got it straightened out to your satisfaction.

Did you ever have any thoughts about the seeming dip in the 2 - 4 kHz range?

Hi Kathy,

Glad you got it straightened out to your satisfaction.

Did you ever have any thoughts about the seeming dip in the 2 - 4 kHz range?

Hi Greg,

Yes, so what I was calling RAW was not really raw because it was coming from Final Cut X already edited, meaning that audio and video was cut already although no audio processing was applied obviously it was not the way it was recorded. So I think the 2-4 kHz range happened where the cuts were.
I appreciate your help. Audio is something I am really trying to improve and it's no easy.

Hi Kathy,

If by "cuts" you mean "edits" (as opposed to frequency cuts due to equalization) then I don't think that's the issue. Just making a simple edit, or a fade for that matter, will not affect the frequency response at all.

Of course without some sort of comparison testing, I can't say for certain that your file had been equalized. But when I listened to it, and looked at the frequency spectrum visually, it certainly appeared that the range from about 2 kHz to 4 kHz was lower in level than the rest of the spectrum. And, to my ear at least, the "presence" was lacking, which corresponds to the above.

Maybe I'm overly analytical, but I wish I could reach some conclusion about this, because it might make things easier for you (or make your audio clearer) in the future.

If you ever have time to just record a few minutes of test file, using the lav in question and another mic which you know is good, let me know, and I'll make some suggestions.

Meanwhile, too bad NYC is 5 hours from here, or I'd be tempted to run over there for a look-see.

Cheers!

Hi Kathy,

If by "cuts" you mean "edits" (as opposed to frequency cuts due to equalization) then I don't think that's the issue. Just making a simple edit, or a fade for that matter, will not affect the frequency response at all.

Of course without some sort of comparison testing, I can't say for certain that your file had been equalized. But when I listened to it, and looked at the frequency spectrum visually, it certainly appeared that the range from about 2 kHz to 4 kHz was lower in level than the rest of the spectrum. And, to my ear at least, the "presence" was lacking, which corresponds to the above.

Maybe I'm overly analytical, but I wish I could reach some conclusion about this, because it might make things easier for you (or make your audio clearer) in the future.

If you ever have time to just record a few minutes of test file, using the lav in question and another mic which you know is good, let me know, and I'll make some suggestions.

Meanwhile, too bad NYC is 5 hours from here, or I'd be tempted to run over there for a look-see.

Cheers!
Greg,

Thanks! How can I get you the RAW file without exporting it from somewhere? The whole RAW video is like 1hr

An uncompressed 1hr. video file would be freak'n huge! That said, an uncompressed WAVE 48k/16 audio file would be about 6MB per track minute.
Some file sharing sites have don't have the usual single file size restrictions.. pCloud for instance.. and offers up-to 20GB free. About 10 gigs upon registration.