Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[krabapple]

I hope this is the right forum for this.

I'm discussing digital NR as a means of reducing hiss on recordings, over on another web forum (not a prosound forum).  Things like CEDAR, NoNoise, the denoising /dehissing plugins in Audition , Sound Forge etc.  I only have experience with Audition, btw (and that's as an amateur - I am not an audio/sound professional.)  There's a debate going on that has come down to this challenge:

"Find me a link [to a prosound page] where someone claims that they can apply NR without any audible artifacts."


The underlying claim there is that audible increase in the S/N via digital NR is *always* accompanied by some unwanted , degradative artifacting affecting the music.  

Is that the experience of the pros here --  or can digital NR ever be 'harmless' with respect to the music -- i.e.,  leaving the music sounding the same (or better), while appreciably reducing the hiss?

FWIW it has been my experience that some light passes of digital NR with Audition can leave the music subjectively intact while reducing hiss.  But i haven't ever tested the difference rigorously.


- Steven Sullivan

[zetterstroem]

in small amounts noise reduction can be useable.... but it will ALWAYS leave artifacts...

then it's just a matter of judgement...

[ammitsboel]

Yes, It's a big compromise.

I have a CD here with a ton of hum noise on it, but i wouldn't trade it for anything else.

[maxim]

there are always trade-offs, but, fwiw, i've always used arboretum's ioniser for this purpose, and found it very powerful

of course, in the long run, the best noise reduction process has been to improve my recording technique

[Bobro]

"Noise", non-periodic movement, non-integer partials... all this is part of musical sound, just ask anybody who has worked with Fourier synthesis. There is a reason why the (theoretically) most powerful means of synthesis is called "additive plus noise" or other similar names.

And these noises get mixed in with recording noise.

So how do you avoid throwing out the baby with the bathwater? Even a noise reduction system without audible motorboating or whatever other byproducts will probably never have the ear and logic to know which of two otherwise identical noises is part of the musical sound or part of the unwanted noise.

In practice it's a musical decision more than a technical one. There's very audible noise-gating, to my ears, on a certain famous recording that was presented to us as the Holy Grail in recording class: the tails are, AFAIC, as much "synthesized" with an EMT as recorded (maybe part of the 3dness of the sound?). But that is a landmark recording... so, while we can measure many things, and "rate" our noise reduction techniques with a good deal of fairness, in the end it's still going to be a matter of opinion!

-Bobro

[PP]

[bobkatz]

krabapple wrote on Thu, 31 March 2005 11:45

I hope this is the right forum for this.

I'm sure it is! Welcome, krabapple...

Quote:

The underlying claim there is that audible increase in the S/N via digital NR is *always* accompanied by some unwanted , degradative artifacting affecting the music.

Unfortunately, it is true. The art of noise reduction is an art of compromise. The very best systems (Cedar, Algorithmix, Backdrop (by TC, Sonic Solutions) have the least artifacts. It is a matter of knowing not to go too far, remembering that the ear itself is the world's best noise reduction system (it knows what to ignore and what to pay attention to). Learn the art of employing contrast and "sameness" to mask noise. What I mean is that if a noisy segment comes out of black, it will call attention to itself, but if you run a low level of noise prior to the noisy segment, it will not seem so noisy!

Can some of these systems be acoustically invisible (without artifacts)? Well, I prefer to call it a "cure versus disease" decision. In many many cases, and in skilled hands, the cure can sound much better than the disease, and the artifacts can sound, shall we say, "negligible, or inaudible to all but the most skilled observer, or masked sufficiently to be inaudible in most places".

You can't get something for nothing, there may be losses in transparency, noise modulation artifacts, "space monkeys", and so on. Once you educate yourself to the artifacts, you soon learn how to be gentle!

How's that?

BK

[krabapple]

bobkatz wrote on Sat, 02 April 2005 14:28

krabapple wrote on Thu, 31 March 2005 11:45 I hope this is the right forum for this. I'm sure it is! Welcome, krabapple... Quote: The underlying claim there is that audible increase in the S/N via digital NR is *always* accompanied by some unwanted , degradative artifacting affecting the music.   Unfortunately, it is true. The art of noise reduction is an art of compromise. The very best systems (Cedar, Algorithmix, Backdrop (by TC, Sonic Solutions) have the least artifacts. It is a matter of knowing not to go too far, remembering that the ear itself is the world's best noise reduction system (it knows what to ignore and what to pay attention to). Learn the art of employing contrast and "sameness" to mask noise. What I mean is that if a noisy segment comes out of black, it will call attention to itself, but if you run a low level of noise prior to the noisy segment, it will not seem so noisy! Can some of these systems be acoustically invisible (without artifacts)? Well, I prefer to call it a "cure versus disease" decision. In many many cases, and in skilled hands, the cure can sound much better than the disease, and the artifacts can sound, shall we say, "negligible, or inaudible to all but the most skilled observer, or masked sufficiently to be inaudible in most places". You can't get something for nothing, there may be losses in transparency, noise modulation artifacts, "space monkeys", and so on. Once you educate yourself to the artifacts, you soon learn how to be gentle! How's that? BK

Pretty good, thanks (and also to the others on the thread).  But I want to dissect this a bit more.  Let's say the user is educated about the artifacts, and is the soul of gentleness itself.  You seem to be saying that *even then* there will be an audible hit to the music (though many might not hear it).  That you are always 'losing' a bit of the music.

Yet the whole reason for NR is that it's supposed to make the recording sound *better*, right?

If one one is of the opinion that the musical content (the 'signal') *always* takes a hit when applying digital broadband NR, why would one ever use it? I'm looking for reasons that aren't client-driven -- for the reasons behind *aesthetic* decisions, in other words. In what sense is the cure ever better than the disease?

I am not anti NR , btw.  But I have encountered 'purist' positions, which claim that using no digital hiss reduction is ALWAYS preferable to using it.  (Some in the biz seem to advocate this stance -- e.g. Steve Hoffman).  For these people the cure is NEVER better than the disease.  What do you make of such a stance?  Is digital broadband NR a compromise that the pros here would completely avoid if they had the choice?  Or does it actually ever make the track sound *better* (like it's supposed to, really)?

[blairl]

krabapple wrote on Mon, 04 April 2005 12:52

If one one is of the opinion that the musical content (the 'signal') *always* takes a hit when applying digital broadband NR, why would one ever use it? I'm looking for reasons that aren't client-driven -- for the reasons behind *aesthetic* decisions, in other words. In what sense is the cure ever better than the disease?

If the hiss is more distracting to the listening experience than the NR, then choose the NR over the hiss and visa versa.

[bobkatz]

krabapple wrote on Mon, 04 April 2005 14:52

\ Pretty good, thanks (and also to the others on the thread).  But I want to dissect this a bit more.  Let's say the user is educated about the artifacts, and is the soul of gentleness itself.  You seem to be saying that *even then* there will be an audible hit to the music (though many might not hear it).  That you are always 'losing' a bit of the music.

Not exactly... basically that it's the art of compromise. Can you make it so that there are no audible artifacts, even for an educated ear?  Fortunately, yes. And then would that mean that you are or are not losing a bit of the music?  Well, maybe yes, but it's probably not audible  .  I'm not trying to play games with that statement, but simply to have us remember that all of these noise-reduction devices take advantage of the ear's ability to mask noise when signal of a high enough level and the right frequency content is occurring. So if you are playing the right music and it is masking the problems sufficiently, then by definition, you won't hear the problem... I guess that means that if you don't hear the problem and you have educated ears, then there is no problem  

Quote:

If one one is of the opinion that the musical content (the 'signal') *always* takes a hit when applying digital broadband NR, why would one ever use it? I'm looking for reasons that aren't client-driven -- for the reasons behind *aesthetic* decisions, in other words. In what sense is the cure ever better than the disease?

A simple answer: The cure is better than the disease when the noise is no longer annoying or bothersome or obvious AND neither are the artifacts.

Quote:

I am not anti NR , btw.  But I have encountered 'purist' positions, which claim that using no digital hiss reduction is ALWAYS preferable to using it.  (Some in the biz

I'm very familiar with that position. It's the "less processing is less invasive" approach, and I certainly subscribe to that adage. But "less is more" has to always be weighed against "cure versus disease". My position is that "a little bit of noise reduction usually goes a long way." He may just be extremely sensitive to the artifacts and/or extremely tolerant of the noise.

My position is that when reducing broadband or specular noise, I always take a cure versus disease approach, and I'd say, out of the last 100 candidates for noise reduction, I found that, probably 20 of them just did not work (the disease was worse than the cure), and that 80 of them benefitted from a tetch of it. I'm not afraid to advise the client that no noise reduction proved better than any.

How's that for a track record  

Pick another engineer, the same sources, and the same toolset, and depending on their skill and/or tolerance for the artifacts, you'll get a different ratio. The engineer you cited is at the extreme of "I can't stand any noise reduction system". I can't fall into that category, because when a source has an annoying noise in it, we have to try to deal with it, as invisibly as possible, of course.

BK

[krabapple]

Very interesting. Thanks again to all.

What do you think of this summary?

Tape hiss is itself an artifact that can audibly 'interfere' with the music; 'artifact free' digital hiss reduction is only sometimes achievable; so it often comes down to which artifact one can more easily bear listening to.