Noise / Hiss Cancellation

  • Thread starter Thread starter asoussa
  • Start date Start date
A

asoussa

New member
Hey,

I have a wav file with 2 type of noise Background & Hiss/Hum plus the audio gain os low!! Complicated!!

Any way, I tried to use the demo version from CoolEdit PRO which looks it can do something good but lots of parameters which i got lost among it!!

Any one can help how I can proceed?? Any guide?

Appreciate your feedback
 
When I read the title of this thread, I was thinking Bias Sound Soap 2, but I see Benny already suggested that. I second his suggestion. It's a great program for removing noise.
 
SoundSoap and/or Audition 2.0. Both work best if the sound/noise source is a constant sound source. The idea is you record the noise by it's self then use it as a filter to run the recording against later. They both work quite nicely for very low level constant background noise or hiss but the harder you drive the clean up process the more artifacting you get and your recording starts to loose a natural sound and more digital.

I contacted Bias a while back when Soundsoap first came out. They didn't have any demo versions but they offered to run clean up tests on sound bytes that I sent them. You might contact their sales people and see if they would do something like that before you buy.

I've over-simplified the process and results a little but that's effectively how they work. I've used them both and they have their usefulness and their limitations.

Good luck.
 
I've had no luck with Sound Soap. In my opinion it's a complete waste and I felt totally ripped off once I used it. I'm not talking about the "Pro" version, that might be better. But the small version is flat out useless in my experience.

Much, much better software for noise reduction is made by a company called Arboretum Systems, and it's called "Ray Gun". It works extremely well, but for some reason is much less well known than Sound Soap.
 
true noise/hiss calcellation.

Let me describe true noise cancellation.

Feed into your good quality stereo soundcard, or external unit, at equal volume on both channels, the interstation hiss from FM radio. (Pink noise would be better but FM noise is easier to come by.) Add to the left channel, some speech but so that the speech is all but swamped by the FM noise and the words cannot be understood. Record both channels at a high level but without any clipping.

Now digitally phase invert one channel, sum both channels to mono and listen.
If you did everything right, the noise will disappear and all you will have is a clear voice with all the noise removed.

That is true noise cancellation and can only happen when you have a perfect copy of the offending noise as a separate entity with which to truly cancel itself out.

Now use the latest and the greatest noise reduction software in the world (Cedar, Sonic, I dont mind. You tell me) and see how you go removing the same noise from that left channel, to leave only the voice.

Tell me how you went. Give both methods a score out of 10.
The scores you get will sum up my views on digital, single ended noise reduction software and will importantly give you a benchmark as to what represents 100% success, because without a benchmark you dont know where you are.

If you're like me you will have learned something very basic, and will resolve always to work hard for a clean, noise free recording in the first place.

Cheers Tim G.
 
SonicAlbert said:
Much, much better software for noise reduction is made by a company called Arboretum Systems, and it's called "Ray Gun". It works extremely well, but for some reason is much less well known than Sound Soap.
Click to expand...

Hmmm, never heard of it. Probably Sound Soap is more recognized because it's from a more well known company. I'll download the demo for Ray Gun tomorrow and test it out, thanks.
Sound Soap has worked pretty well on a few projects for me. Especially for the money. Yes, it's not perfect...but I find hardly any noise reduction plugin is (Waves gives me fits at times). I think Cedar is still the industry leader.

It's all about the art of minimalism. Less is more. I tend to EQ as much as I can out of the original take, then add noise reduction. This helps me use the less is more idea by only concentrating on the the part I couldn't EQ out. Maybe throw two noise reduction plugins on the channel and different settings. I think no matter what you do or use there is always going to be artifacts to be heard. It's just the nature of the beast. Just learn to compromise between noise and artifacts. Or learn to cover it up with other tracks (ie. covering up artifacts added to dialogue with music).

Or just rerecord.


--edit--

Good interview of some of the greatest engineers who have had to deal with more nosiey audio than any of us:
http://www.filmsound.org/QA/nonoise.htm
 
Last edited:
Tim Gillett said:
Let me describe true noise cancellation.

Feed into your good quality stereo soundcard, or external unit, at equal volume on both channels, the interstation hiss from FM radio. (Pink noise would be better but FM noise is easier to come by.) Add to the left channel, some speech but so that the speech is all but swamped by the FM noise and the words cannot be understood. Record both channels at a high level but without any clipping.

Now digitally phase invert one channel, sum both channels to mono and listen.
If you did everything right, the noise will disappear and all you will have is a clear voice with all the noise removed.

That is true noise cancellation and can only happen when you have a perfect copy of the offending noise as a separate entity with which to truly cancel itself out.

Now use the latest and the greatest noise reduction software in the world (Cedar, Sonic, I dont mind. You tell me) and see how you go removing the same noise from that left channel, to leave only the voice.

Tell me how you went. Give both methods a score out of 10.
The scores you get will sum up my views on digital, single ended noise reduction software and will importantly give you a benchmark as to what represents 100% success, because without a benchmark you dont know where you are.

If you're like me you will have learned something very basic, and will resolve always to work hard for a clean, noise free recording in the first place.

Cheers Tim G.
Click to expand...


well, mistake #1 with that theory is your saying that the source noise is going to be in phase on both channels. Ain't gonna happen. Pink, white, whatever is just going to be random noise (at least specific to frequency or amplitude). So it won't cancel any noise out.

And this isn't about noise cancellation. It's about noise reduction. Reducing it so it's less annoying to the ear so we can hear the music, dialogue, whatever. And the ability to reduce it first depends on how much of a problem there is and how smart the technology is for doing noise reduction. Unfortunately it's not always possible to record clean, noise free audio in the first place. And I don't think the original poster is saying he's recording noisey audio and just wants to use the plugin to clean up every track he records. If that were so I'd agree with you and tell him to fix the problem first. But the problem is already there. Cleaning up crappy phone recordings, old record albums or cassettes...and more commonly, audio for video. Believe me, so much audio for video/film is recorded horribly and you just have to learn to live with it or try and correct it with a plugin/hardware if you can't re-record it in a studio (which mainly only big budget films can afford to do).
 
I wouldna said it if I hadna tried it first! Have you tried it?

It does work (in this rare case) because that signal is NOT random, related to itself, though it would be to anything else, including itself only a fraction of a second later in time. You didnt read what I said. I recorded both channels of the FM noise at the same time. That noise is mono.

And I did it to underline YOUR point about the otherwise randomness of such noise, which you so eloquently put.

The title of the thread (not my thread) was noise cancellation. I demonstrated what noise cancellation is. That's why I titled it true noise cancellation, to distinguish it from "noise reduction", the latter being a better description because as you say, and I was trying to emphasize, noise reduction is usually something different and more modest in result.

I agree with everything else you said. Much of my work is working with often unbelieveably crappy sound from mostly old recordings.

You might do better next time to not make the assumption you know more about the topic than the person writing the post. I encounter this attitude a lot more frequently than I'd prefer to, on this site.

Regards, Tim G
 
bennychico11 said:
Hmmm, never heard of it. Probably Sound Soap is more recognized because it's from a more well known company. I'll download the demo for Ray Gun tomorrow and test it out, thanks.
Sound Soap has worked pretty well on a few projects for me. Especially for the money. Yes, it's not perfect...but I find hardly any noise reduction plugin is (Waves gives me fits at times). I think Cedar is still the industry leader.

It's all about the art of minimalism. Less is more. I tend to EQ as much as I can out of the original take, then add noise reduction. This helps me use the less is more idea by only concentrating on the the part I couldn't EQ out. Maybe throw two noise reduction plugins on the channel and different settings. I think no matter what you do or use there is always going to be artifacts to be heard. It's just the nature of the beast. Just learn to compromise between noise and artifacts. Or learn to cover it up with other tracks (ie. covering up artifacts added to dialogue with music).

Or just rerecord.


--edit--

Good interview of some of the greatest engineers who have had to deal with more nosiey audio than any of us:
http://www.filmsound.org/QA/nonoise.htm
Click to expand...

Again, I agree with Benny. I've had good like with Sound Soap 2. And for $49 (I think) for the standard version, which can work as a stand-alone program, or a plug-in in your DAW, I think it's a pretty inexpensive tool to have in your arsenal. Plus the interface looks pretty cool too. I have no doubt there are better programs out there, and I haven't used Sound Soap Pro to know the differences, although I think Pro only works as a plug-in. But again, you don't have to break the bank.

Like Benny said, I think it's much more well known because Bias is a fairly well known company. I've used Peak and it's a nice little 2-channel editor as well.
 
Tim Gillett said:
I wouldna said it if I hadna tried it first! Have you tried it?

It does work (in this rare case) because that signal is NOT random, related to itself, though it would be to anything else, including itself only a fraction of a second later in time. You didnt read what I said. I recorded both channels of the FM noise at the same time. That noise is mono.

And I did it to underline YOUR point about the otherwise randomness of such noise, which you so eloquently put.

The title of the thread (not my thread) was noise cancellation. I demonstrated what noise cancellation is. That's why I titled it true noise cancellation, to distinguish it from "noise reduction", the latter being a better description because as you say, and I was trying to emphasize, noise reduction is usually something different and more modest in result.

I agree with everything else you said. Much of my work is working with often unbelieveably crappy sound from mostly old recordings.

You might do better next time to not make the assumption you know more about the topic than the person writing the post. I encounter this attitude a lot more frequently than I'd prefer to, on this site.

Regards, Tim G
Click to expand...

Sounds like what you were describing was more like how active noise cancelling headphones work. I think the title of the thread was maybe a little misleading, because he does say noise cancellation, but I think (unless I got it wrong) he's really looking for a way to clean up dirty audio.
 
Tim Gillett said:
I wouldna said it if I hadna tried it first! Have you tried it?

It does work (in this rare case) because that signal is NOT random, related to itself, though it would be to anything else, including itself only a fraction of a second later in time. You didnt read what I said. I recorded both channels of the FM noise at the same time. That noise is mono.

And I did it to underline YOUR point about the otherwise randomness of such noise, which you so eloquently put.

The title of the thread (not my thread) was noise cancellation. I demonstrated what noise cancellation is. That's why I titled it true noise cancellation, to distinguish it from "noise reduction", the latter being a better description because as you say, and I was trying to emphasize, noise reduction is usually something different and more modest in result.

I agree with everything else you said. Much of my work is working with often unbelieveably crappy sound from mostly old recordings.

You might do better next time to not make the assumption you know more about the topic than the person writing the post. I encounter this attitude a lot more frequently than I'd prefer to, on this site.

Regards, Tim G
Click to expand...

I think you need to be careful with your assumption that certain posts are written with an attitude. My post was just in reply to what you said and no attitude of my knowing more than the poster was implied. Sorry if it came across that way to you.

All I was saying is that taking pink noise or interstation radio hiss (which both come in forms of stereo...and you just said record the hiss/noise to both channels which anyone is going to assume is just take a stereo recording of noise)...and flipping the phase of one of the channels is not going to cancel that noise out. Because if you look at pink noise rendered to a stereo track (not just copying a single, mono render of the noise) you'll see they don't look a like at all. And as you said, both channels have to be exactly alike in order to cancel 100%.

But even IF you had two of the same noise channels that were identical, once you sum a voice to one of the channels it's going to change the harmonic content of that channel so much that you aren't going to be able to cancel the noise when you sum it to mono. Certain frequencies may diminish and the sound will change, but the noise will never cancel. If you need audio examples as to how it will sound, I'll more than happy provide it for you.

Like RAK said, the original poster may have asked about noise cancellation, but the posts following talked about noise reduction as we tried to make clear to him that it's much more about minimizing the artifacts than completely elminating them.
 
Tim Gillett said:
That is true noise cancellation and can only happen when you have a perfect copy of the offending noise as a separate entity with which to truly cancel itself out.
Click to expand...
And where, pray tell, does one get that? If I could get a seperate exact copy of the noise on an entire track that was seperate from the rest of the signal, that would mean that I already had the means of seperating the signal from the noise and therefore would not need to go any further :).
bennychico said:
I tend to EQ as much as I can out of the original take, then add noise reduction.
Click to expand...
This is an important step to consider when using noiseprint-style NR packages.

Sometimes the noise is sophisticated enough in nature where just removing the unprocessed noiseprint winds up taking out too much of the signal as well, leaving really ugly sounding artifacting. Sometimes, some targeted EQ can "simplify" the noise without adversely affecting the signal. Then getting the after-EQ noiseprint can still remove the noise without artifacting the signal the way it did.

This is one of the few times where I actually advocate the use of an RTA. One's ears are not great at isolating many important components of real-world, non-pink noise, and an RTA can help identify some more energetic frequencies that can be "surgically" notched out of the noise with minimal or no audible effect to the signal. Then the resulting simpler noiseprint in-turn artifacts the signal less.

It doesn't seem to make a lot of sense at first blush that it should work that way, but it often does. I just used this method about two weeks ago myself on some tracks I got from a synth composer. These tracks had some rather noticable hiss on them that was an electronic bleed that came in on his system when he was converting his MIDI to audio files. I tried using the Sony NR plug on it and it took out so much of the signal that some of his synth lines sounded like whales on acid instead of strings and horns. So I undid that and took a look at the noise signature from the quiet part of the file and ID'd a curious harmonic signal in the noise. There was a fundamental that peaked somewhere around 381Hz, with both even and odd harmonics in descending amplitude all the way up the scale, and even a single subharmonic to the fundamental.

So I attacked the track with a decending-amplitude harmonic filter (using the Elemental Audio Eqium plug) to take those harmonic peaks out of the noise. Because the filtering was very narrow Q and at "weird" frequencies as related to the synth signal itself, the effect on the signal was, for all intents and purposes, inaudible. Sure you could hear a slight difference when A/Bing, but not enough to affect the outcome of the production. But the important part was the noise print of the filtered noise now removed the noise without severly artifacting the signal the way it did before.

G.
 
bennychico11 said:
I love that plugin :)
Click to expand...
Ain't it nice? Definitely a workhorse plug in my toolkit. Was glad to grab it before Roger Nichols changed the skin and charged double for it :p .

G.
 
bennychico11 said:
I think you need to be careful with your assumption that certain posts are written with an attitude. My post was just in reply to what you said and no attitude of my knowing more than the poster was implied. Sorry if it came across that way to you.

All I was saying is that taking pink noise or interstation radio hiss (which both come in forms of stereo...and you just said record the hiss/noise to both channels which anyone is going to assume is just take a stereo recording of noise)...and flipping the phase of one of the channels is not going to cancel that noise out. Because if you look at pink noise rendered to a stereo track (not just copying a single, mono render of the noise) you'll see they don't look a like at all. And as you said, both channels have to be exactly alike in order to cancel 100%.

But even IF you had two of the same noise channels that were identical, once you sum a voice to one of the channels it's going to change the harmonic content of that channel so much that you aren't going to be able to cancel the noise when you sum it to mono. Certain frequencies may diminish and the sound will change, but the noise will never cancel. If you need audio examples as to how it will sound, I'll more than happy provide it for you.

Like RAK said, the original poster may have asked about noise cancellation, but the posts following talked about noise reduction as we tried to make clear to him that it's much more about minimizing the artifacts than completely elminating them.
Click to expand...

Thanks for your reply.
FM interstation noise IS mono. FM only becomes stereo when the HF carrier, present on a broadcast, gives you the stereo difference information for that broadcast. Interesting in that we have the same additive, subtractive process there too.

No you're wrong that adding the voice will compromise the noise imprint. It is simply added to it. Subtracting the noise will... subtract the noise.
Yesterday, as a confirmation for myself before I launched into print, I did the test. If you like I can send you the files.
Of course it will never be 100% perfect because of real world limitations but we're talking real word effectiveness, not absolutes.

Sure, real world noise reduction is quite different. I was picking up on the wording used and explaining what "noise cancellation" really means, and how much better it would be, if it were possible to do that, which it usually isnt.
Best wishes and I appreciate your post.

Tim
 
SouthSIDE Glen said:
And where, pray tell, does one get that? If I could get a seperate exact copy of the noise on an entire track that was seperate from the rest of the signal, that would mean that I already had the means of seperating the signal from the noise and therefore would not need to go any further :).
Click to expand...

My point exactly. I didnt tell you where you could find it because I knew it normally CANT be found. Again, my point, first to underline that noise cancelling (what I described and demonstrated to my own satisfaction with an actual test) is normally impossible, but if it WAS possible, it would be the first port of call, and would leave all other methods in the dust. More a point of clarification I guess.

Since joining the site I've been reading and appreciating your posts.

Best wishes,
Tim
 
Tim Gillett said:
My point exactly. I didnt tell you where you could find it because I knew it normally CANT be found. Again, my point, first to underline that noise cancelling (what I described and demonstrated to my own satisfaction with an actual test) is normally impossible, but if it WAS possible, it would be the first port of call, and would leave all other methods in the dust. More a point of clarification I guess.
Click to expand...
Gotcha. Unfortunately I think folks like Benny and I misinterpreted your original post as recommending an actual real-world technique. Apologies for the misunderstanding.

G.
 
Tim Gillett said:
No you're wrong that adding the voice will compromise the noise imprint. It is simply added to it. Subtracting the noise will... subtract the noise.
Yesterday, as a confirmation for myself before I launched into print, I did the test. If you like I can send you the files.
Of course it will never be 100% perfect because of real world limitations but we're talking real word effectiveness, not absolutes.
Click to expand...

Below are pictures of how I understood your test.
The before picture has the top track contain voice content that I had, and the two bottom tracks being identical pink noise rendered at -18dBFS.
Now in the after track I bussed the voice trak and one of the noise tracks to a new mono track...and recorded the summation of the two. As you can see, parts of the sound in the top track has changed. Notice the big peak that wasn't there at the beginning of the track? Parts of the rest of the file look similar to the bottom one, but I promise you it is not.
If I flip the phase of one of these tracks and play them both summed in mono...SOME frequencies have canceled or changed, but the noise is still there. Actually, I find that most of the lower frequencies are gone, but there is still quite a bit (if not more) HF noise. If you want audio examples, it'll take a little while to post them.

Remember what happens when you add two wave forms together? If they are both traveling in the same direction amplitude wise, they are going to sum together and INCREASE in amplitude. Hence as to why the after picture shows that peak at the beginning where the voice comes in at.

Again, I think we're straying from the original poster's question (as happens so much in these threads) :)
 

Attachments

  • before.webp
    before.webp
    20.9 KB · Views: 48
  • after.webp
    after.webp
    22.1 KB · Views: 60
OK. Thanks for the pics.
The way I did it was send the mono analog noise as L + R straight in to the stereo soundcard (with accurate L/R level matching) but with a voice track added to the left only. The only digital manipulation was inverting one track's phase.
In stereo playback you still got lots of noise, as you'd expect, and with huge stereo spread, as you'd expect, and the voice barely audible, let alone words intelligible.

Sum to mono and practically all the noise went. The voice stood proud and clear.

If only we could do that with some of the recordings we're given!

Cheers, Tim
 
You must log in or register to reply here.
Back
Top