Audacity Forum

stevethefiddle:

The .wav file I originally attached with my initial post on this topic was recorded without any record playing at all. The turntable was not "running" at all. As a matter of fact, with the turntable unplugged from the 110v, Audacity still sees this "noise". That's why I referred to it as "line noise", knowing it was not the noise generated naturally by imperfections in the vinyl LP record.

It may be that I simply can't eliminate this noise from a "hardware" standpoint. Your comment that it "seemed a bit loud" spured me to try to do something about it from a software direction. But, this too, may be beyond me for the time being.

The noise generated by the LP records themselves (general imperfections, dirty records, scratches, badly worn, etc.) is a different matter. I'm trying to learn how to use Audacity's NR tool properly, to minimize these unwanted sounds, as well.

And, thanks to the folks on this Forum, I am making some progress. I appreciate all the help you and others are offering.

Regards, Pat

patcurci wrote:Your comment that it "seemed a bit loud" spured me to try to do something about it from a software direction. But, this too, may be beyond me for the time being.

It may be worth just checking in the Windows Control Panel that the on-board sound card is disabled for recording. No guarantees that it will make any difference, but it would eliminate one possible source of noise.

I think this topic may have taken an unintended, yet important, change in direction: a discussion of the source and kinds of noise experienced in recording. Of course, Audacity doesn't know what noise is (for the Noise Removal Tool) until you tell it to "Get Noise Profile". There is even some talk now that Audacity's NR tool may not be behaving as intended.

(When I make changes, from one extreme to the other, in the settings for "Frequency smoothing" and "Attack/decay time", I can't hear any differences when applied to a sample piece of music. But I don't know how the tool is supposed to behave.)

But, if possible, I would like to bring the topic back to a more practical direction.

Stevethefiddle described "vinyl noise" as the noise you hear at a lead-in or between music tracks on an LP record. And my LPs certainly have a lot of that kind of noise.

I have a 3 minute jazz music track with piano, bass & guitar (Oscar Peterson). When all are playing it's difficult for me to hear any noise, but during the solos and particularly at the end when the piano is supposed to fade out to quiet, the piano fades in to noise. The NR tool is, I think, the proper one to use to help fix this. I know that some "music" is removed whenever the NR tool is used, so all the noise can't be eliminated without damaging the music.

Ignatz has pointed out "trust your ears" and that's great advice. And stevethefiddle suggested geting a noise profile at the beginning or between tracks of the LP, also great advice.

But can some of the folks here offer some suggestions as to what "settings" should I start with (Noise reduction, Frequency smoothing, Attack/decay time). [I've used Noise reduction as low as 6 with some success.]

Should the NR tool be used on the whole track?
Only the solos or end fade out?
Do you try to fade out the noise along with the piano at the end?
Do you start the NR tool at a loud place on the music track or when noise is most evident?
What experiences have you folks had that yield the best results?

I will certainly appreciate any suggestions and considerations.

Regards, Pat

patcurci wrote:I have a 3 minute jazz music track with piano, bass & guitar (Oscar Peterson). When all are playing it's difficult for me to hear any noise, but during the solos and particularly at the end when the piano is supposed to fade out to quiet, the piano fades in to noise. The NR tool is, I think, the proper one to use to help fix this. I know that some "music" is removed whenever the NR tool is used, so all the noise can't be eliminated without damaging the music.

But can some of the folks here offer some suggestions as to what "settings" should I start with (Noise reduction, Frequency smoothing, Attack/decay time). [I've used Noise reduction as low as 6 with some success.]

Should the NR tool be used on the whole track?
Only the solos or end fade out?
Do you try to fade out the noise along with the piano at the end?
Do you start the NR tool at a loud place on the music track or when noise is most evident?
What experiences have you folks had that yield the best results?

From your description, I would consider doing a fade out at the end of any troublesome track. I do it all the time. Many of my tracks have surface wear and unavoidable vinyl/shellac noise. As the song fades, that stuff is going to jump out at you. So, I evaluate each track by listening to the final 3 or 4 seconds. If I don't like what I hear, I do a fade out from the effects menu. You just sweep with the mouse over the area you want to fade and then hit the button. I usually fade out over maybe 1/2 of a second or less. Occasionally, I might use 1 second or so, but the noisier the tail is, the quicker I fade.

On a particularly noisy ending, I will end the fade still within the music---there's some waveform beyond the fade ending on the right side. But it's too nasty to keep, even if you attempted to fade it.

You might then use generate/silence to append a second or 2 of silence to the end of the fade before the next songs comes in.

You have to decide which is more bothersome: the noise that rises when the music falls or NOT getting the last millisecond of Oscar's piano. If you want the last millisecond, you can't fade. Personally, I ALWAYS choose to fade rather than hear something I don't want to hear.

I think you would drive yourself crazy in short order if you tried to use NR on just portions of a track---maybe a few rare cases excepted. You'd end up chasing your own tail. You have to draw the line at how much time you are going to spend per track---on average. Otherwise, you'd be in there manually repairing 50 ticks per song.

I just get a noise sample, select all, apply 12 NR, and then listen to the opening and maybe a couple of random spots in the track to evaluate the NR and any artifacts. Then maybe undo and adjust the top slider. I use 12 as a starting point nearly all the time, with the others at default.

I agree with you about not noticing any effect from frequency smoothing and attack/decay. Frankly, I am wondering how they both got into Audacity 1.3.12 since Steve has said one of them is broken--I've forgotten which one.

Thanks Ignatz for your suggestions on using a fast fadout as a solution to the end of a noisey music track. I'll give it a try.

I have tried using the NR tool witin portions of a music track and even if the "Noies reduction" slider is set low (around 6 or 8) I hear a noticable drop in the loudness of the music. Maybe not such a great solution.

It's my understanding the the "Attack/decay time" is the slider that has no effect in Audacity 1.3.12. But I still don't know what the "Frequency smoothing" is supposed to do, since I can't hear differences when I change this particular slider.

I'm very pleased with how Audacity works and the tools it offers. Some folks have suggested that Brian Davies offers 2 software programs (declick and denoise) for a fee, that are superior to Audacity's. I'm a bit reluctant to go in this direction for an incremental improvment, but................?

Regards, Pat

patcurci wrote:even if the "Noies reduction" slider is set low (around 6 or 8) I hear a noticable drop in the loudness of the music. Maybe not such a great solution.

I don't follow your inference. After noise reduction, you can apply amplfication to overcome the "drop in the loudness of the music". Noise is part of the original signal, so naturally when you remove it, the apparent volume falls. Likewise, if you bring a crying baby into the room as you are doing a live recording, the "loudness" will rise, meters go up, and you may clip.

patcurci wrote:But I still don't know what the "Frequency smoothing" is supposed to do, since I can't hear differences when I change this particular slider.

I have yet to detect any effect, but:

Frequency Smoothing: The larger this value the more that the effect considers different frequencies as "the same". If your noise is a single frequency such as mains hum or a high-pitched whistle, then keep this value small. If your noise is more like a hiss, then a larger value will generally be better.

patcurci wrote:Brian Davies offers 2 software programs (declick and denoise) for a fee, that are superior to Audacity's.

ClickRepair works very well, but I have not compared it to Audacity's Click Removal effect. All my click repairing in Audacity has been done manually using "repair". Based on other user's reports, I suspect the Davies' app does a better job. It can batch process and is laid out very well without an overload of buttons and sliders.

DeNoise and DeNoiseLF also work very well, but whether your ear finds them superior to Audacity is another matter. I won't be buying DeNoise because of the difficulty in selecting a small noise sample without a zoom control and because I am satisfied with Audacity denoising.

I haven't yet bought ClickRepair, but probably will. But I first have to try Audacity Click Removal and see how unhappy I am with it.

After you get Audacity NR's capabilities sorted out in your mind to your own satisfaction, you ought to download the Davies apps for a 21 day trial.

Ignatz:

I don't follow your inference. After noise reduction, you can apply amplfication to overcome the "drop in the loudness of the music". Noise is part of the original signal, so naturally when you remove it, the apparent volume falls

.

I'm no expert, not even close, and certainly don't know the proper terms, but I seem to have read in the help files (?), Wiki pages (?) or in some forum posts that whenever the NR tool is used, it can't help but remove a little bit of the music too. Music frequencies and noise frequencies may overlap or something like that. So some music is lost. Whether this shows up as a change in volume or something else, I don't know.

I selected a small portion of a music track, near the end where the music was not loud. I looked at the "waveform" and made the selection start in the middle of a "loud" part of that waveform, and applied the NR tool. (I did a "Get Noise Profile" very close by, but with no music.) After the NR tool did its magic, I heard a noticeable drop in the loudness of the music where the NR tool started, compared to immediately before. My ears told me so .

In any case, the Noise Removal Tool is a complex, but very useful tool. Learning to use it to improve my recordings is not an easy task. For many of us, hearing the improvment, is the only practical yardstick balanced against the time necessary to make those improvments.

Many thanks.

Regards, Pat

patcurci wrote:For many of us, hearing the improvment, is the only practical yardstick

Hearing the improvement (or not) is the yardstick for the experts too.

patcurci wrote:I seem to have read ....that whenever the NR tool is used, it can't help but remove a little bit of the music too. So some music is lost. Whether this shows up as a change in volume or something else, I don't know.

I'll stipulate that some music is lost.

So what? Plenty of music was lost in various stages of the production of the LPs you are now trying to quiet. You lose even more music every time a stylus hits the vinyl--additional wear and frequency response declining with every play. A diamond stylus doesn't have any trouble carving up vinyl.

I can't hear what is lost. You don't move the slider far enough to where you can hear a decline in the quality of the music.

Which is worse:

1: Knowing you are losing some music when you can't actually hear any decline in quality.

2: Living with noise you can hear many times in an hour of listening.

For me, 2 is far far worse.

patcurci wrote:Learning to use it to improve my recordings is not an easy task. For many of us, hearing the improvment, is the only practical yardstick balanced against the time necessary to make those improvments.

Hearing the improvement is the only practical yardstick for everyone, not just you.

If you do a quick A/B comparison on a recording (with NR and without NR) and you prefer without NR, then don't use it. In that sense, it's about as easy as something can be. No learning curve at all. You prefer the NR version or you don't.

As for time, NR typically takes me less than 30 seconds per recording. Get a noise sample, apply, confirm results. Perhaps one time in five I might readjust the slider and re-evaluate once--which might take another 30 seconds. I always apply NR to the entire track if I use it at all and I've never taken more than 2 minutes to apply NR. Controlling intros and fades is more time consuming.

That's my experience. Your mileage may differ.

Ignatz:

I appreciate your comments. I think one of the things I may have been doing wrong(?), is trying to apply the NR tool to parts of the overall recording (the lead-in on an LP, the spaces between music tracks, the ends of music tracks where I hear noise) rather than a complete side of the LP or even the entire recording of the LP.

At what point in the "process" do you apply the NR Tool? Is it one of the very first steps after making the complete recording. After "Click Removal"?

I'll certainly give your technique a try. Thanks.

Regards, Pat