Get a noise profile from a silent recording and use those freqs for filtering

I am using a second hand condenser mic and there’s some noise and a high pitch. Perhaps it’s a cheap pre-amp, is it abused by the previous owner, or both.

Anyway, in Audacity, you can select some ‘silence’ (with noise) and choose Noise Reduction > Get Profile, followed by selecting the rest and do denoising. This is pretty awesome. Makes the mic sound good.

But I’d like to work with Audacity. I was wondering, does anyone know a trick or plugin that does this on the fly? The ‘noise profile’ is always the same anyway.


High pitch noise, try a different combination of Microphone and Pre-Amp to see if it goes away. In particular the preamp and the phantom power circuits on some of the lower end interfaces and consoles I have used tend to display this characteristic with some decent condensers especially.

In as far as ways to fix it, I use the single band parametric LADSPA with a VERY narrow Q(I double click on the boxes to type in .005 - .001 Q) and sweep to find the exact frequency if I don’t know it from a spectral analysis of the region.

In as far as broadband noise reduction, your options on Linux are very limited. There are several AU plugins for OS X, but Linux has dragged behind in this area and I don’t know of any realtime plugins for Linux, and only a few standalone processors that do this(Audacity and GWC both) and none of them all that good at it. Sorry.


If you suspect the noise is being caused by faulty second hand equipment, then you are almost certainly better off trying to either get it fixed or get a microphone that doesn’t exhibit the problem (if you can afford to) rather than fixing it in software. Whatever method you choose to remove the noise in software will almost certainly cause some unwanted side-effects. It’s much better to use software plugins to polish an already good quality recording rather than to try and fix some defect in the recording process (unless that is completely unavoidable). That said, perhaps I should think about doing some kind of de-noising plugin…

Thanks for the advice! Though this is something I do want to fix in software, even though it’s not very nice.

I wanted a condenser mic real bad but I simply did not have money for it. All of the sudden I see this Behringer C-1U for €35 in a pawn shop and I couldn’t resist. If I filter the noise profile the audacity way, it still sounds WAY better then my other dynamic mic.

Even new it is a cheap model so I think the hiss is in there even when it’s brand new. But the (soft) high pitch, haven’t heard that in other instances so I think that one is introduced by wear and tear. Just a guess though.

USB being digital, I think I don’t have to try and buy a new cable.

seablade, I think what you do manually is what the noise filter does by ‘measuring’ the ‘noise profile.’ With both software being GPL, maybe it’s not hard to port audacity noise filter to LADSPA?


No what I described is VERY different from what the noise filter does. What I am doing is using a very tight EQ to notch out a problem frequency, the noise filter acts differently though I can’t tell you the specifics of that particular implementation, many will act as multi-band frequency gates(Though I don’t think that is the particular implementation in Audacity).

Converting Audacity’s offline process to a realtime capable plugin, you may as well start from scratch actually, they are two very fundamentally different things.

That being said…


That is certainly one area that Linux is very behind in, especially if looking for a realtime solution ala WaveARTS or some of the other larger restoration suites(Though I particularly like WaveARTS due to low latency as well as good performance).


By the way, Behringer Condensers, especially the USB variety are not known for good quality. That being said I think I recall a few threads on these forums about them and how to get a decent sound out of them, but my mind might be playing tricks on me again.

I do agree with what LinuxDSP posted though as well.


Behringer equipment gives you good results, as long as it works. That is the trade-off for lower production cost, and while I own a DDX3216 mixer, that has not disappointed me in the past 5 years, I would probably not buy one second hand. There must be a reason why the previous owner got rid of it in the first place.

My advice: If you get Behringer, get it NEW and look after it. It probably would not last forever, but the value for money factor is better than most.

PS: I would not buy any microphones that has integrated USB sound devices, for one of the following reasons

  1. Lower quality - It was made to cater for the lone bedroom warrior
  2. Limited options in Jack - You have to choose one sound device… So you choose to have only one microphone connected?
  3. 0V-5V rails for a converter? That is not nearly enough head-end for the signal. You want lots and lots for the best transient responce.
  4. DIGITAL gain control - You really want to be able to turn analogue pot to control that.

Until these does not change, or VASTLY inprove, I would not get a USB mic.

Yes I agree with most, USB is not a nice way to get phantom power and second hand hardware is not a good idea in most cases.
But this is what it is, maybe I’ll buy a good new one next year or something. I’m not a studio guy, I just do the occasional recording.

If I understand correctly, the Linux world is meager around noise reduction and plugins, so I guess my best tacle is to quit after deciding what recorded tracks to keep, and pull them through Audacity and go back to Ardour.


I must say that the way you described is pretty much the only way I can think of. I assumed that Audacity does just that. It scans for frequencies in the selected silence when getting a profile, and it eliminates those frequencies in the actual noise reduction. It has a very precise Q. That’s why it sounds so good. Maybe it was different in the past? When did you last try it? Because I don’t really know anything about Audacity, I was just downloading random software to check out their filtering capabilities.

BTW I’ve been looking around, the topics I can find are generally about way better mics, behringer condenser specific and only my topic pops up in the search. :wink:

I think a lot of the problems probably stem from the fact that this is a USB microphone:

If the power is provided by the USB not an external supply, then this is notoriously prone to noise since it will be powered straight from the computer power supply, with all the digital noise from the rest of the system that gets back onto the supply rails. This is normally quite low level noise and not normally a problem for USB sound-cards that expect line-level inputs and therefore operate at relatively low gain internally, but the microphone will incorporate some sort of pre-amp, which may or may not reject this noise effectively (and could therefore be only about as good as the built in mic input on a normal sound interface…)

One possible suggestion, if you have a spare USB hub with its own power supply, plug the microphone into the hub and the hub into the PC, see if its any different.

I assumed that Audacity does just that. It scans for frequencies in the selected silence when getting a profile, and it eliminates those frequencies in the actual noise reduction.

Well that is a ten thousand mile overview I believe;)

I believe Audacity builds a noise profile, by averaging energy content in the spectrum over the profile you have picked, and then attempts to remove that down to the 0 point in the rest of the file, a bit different approach honestly. But I haven’t looked at their implementation this is just a guess.

And the last I tried their noise reduction was somewhere in the 1.3 beta series, but don’t know the exact version. I obviously also tried it in the 1.2 series. And in all cases it was barely usable at best IMO. Nothing against the Audacity devs mind you, it at this point just loses out to DSP programs with a bit more financial backing sadly, and I don’t have the knowledge to be able to help improve it there.


@ linuxdsp

Thanks for the idea. I am running the mic over a USB hub at the moment. It came with a power supply but I’ve never used it since I bought it. I will look for it. The DC adapter must be somewhere in my house. :slight_smile:

@ seablade

If it’s like that, it’s even better, right? Energy in the spectrum is like an amplitude of a frequency, but an EQ has only so many channels to set freq and Q on. Sounds like a very effective method.
But even if it’s flawed, it’s the best sounding linux compatible method I’ve found so far. And accidentally also the easiest! :slight_smile:

If it's like that, it's even better, right? Energy in the spectrum is like an amplitude of a frequency, but an EQ has only so many channels to set freq and Q on. Sounds like a very effective method. But even if it's flawed, it's the best sounding linux compatible method I've found so far. And accidentally also the easiest! :)

Not necessarily better no, in fact in several ways it can be worse. All that being said you are getting more into DSP theory than I am all that comfortable with, so it is better I leave a critical analysis of DSP methods to those more suited to it.

But the key words in your statement, best sound linux compatible are the things to take note of. There isn’t any good sounding Linux compatible solution that I know of. The good sounding solutions range from dedicated hardware (CEDAR) to software for other OSes(WaveARTS MR, IzotopeRX, Sonnox Restore, etc.)

A long time ago, even the noise reduction in Cool Edit Pro(Later Audition 1.0/1.5) beat out the noise reduction in Audacity in my opinion back when I used both of those.

I think a lot of the problems probably stem from the fact that this is a USB microphone:

I certainly agree with the basic concept here, though I will say that many USB microphones can sound much better than what was mentioned here, albiet they still won’t sound good. But yes this is the basis of the problems most likely.


I think you people are far too professional minded here. :slight_smile:
I just graduated two weeks ago and I make random video’s, I just want to record the occasional voice over or pumpkin-smash for a head-bust-in sound effect. If I am lucky, I can afford After Effects CS5 this year, but until then I am not going to buy a new mic or professional audio software. I’m that guy that buys discount wood from woodshack to cheat at making that pseudo-expensive camera move. :wink:

(I use Ardour also for occasional amateur mastering of songs but they are recorded in a studio with real equipment.)

I did notice (!!!) that after some random messing with my wiring, the high pitch background noise from the mic was gone! But the white noise was (ofcourse) still there, and is to stay for the reasons you guys already came up with.

Makes me wonder two things though, would an expensive USB cable clean the noise even more, for maybe the hot wires (the ones that carry the (mere) +5) are better shielded? You know, more expensive cables usually have those cilinder formed chunks at both ends. They are for cleaning the signal, right?
And would it be possible to open up the mic and insert a real phantom power inlet, or are all components specially designed to work with inferior USB voltage?

I cannot play with plugins anymore though. After using Audacity, for some reason Ardour does not play back anymore. Or any JACK program. I’ve been trying for (literally) 3,5 hours now. Rebooting, changing options in JACK even though I didn’t mess with them in the first place, and getting a headache. My confusion every time Ardour or JACK stops working is time consuming and makes me a sad panda. When I get back on the wagon, if the problem hasn’t disappeared as magically as it appeared, I might open a dedicated topic to get it working again.

I might open a dedicated topic to get it working again.

Highly recommended. I would bet your problem is having something already trying to use the audio device you are trying to use with Jack.