Low frequency noise - is this normal?

Audacity 3 Help Forum GNU/Linux
1

I get quite a lot of low-frequency noise. The famous mastering suite Filter Curve effect gets rid of most of it … but then I sometimes fail the noise part of ACX Check.

My set up is a Rode NT-1A cardioid condenser, with a pop screen, and a Focusrite Scarlett 2i2 interface. The particular sample I’m attaching wasn’t actually recorded directly into Audacity but using arecord on a Raspberry Pi … but that’s a matter for another post!

Any advice from you experts on the level of noise I’m getting? Normal - or should I be doing more to control it at source?

I haven’t reached any conclusions about whether I should speak directly into the mic, or at an angle, and at what distance.

I feel like I need more volume - however the Focusrite gain is at about 80% of max and I’m getting plenty of yellow on the light.

The only edit I’ve done on this is to trim it to 10 seconds.

2

Definitely should filter-out that infrasound before applying threshold-dependent effects,
like compression, limiter, DeClick, DeEss.
Otherwise those effects will be applied in an inconsistent manner across the track.

3

Thanks Trebor, I’ll check through those effects and threads.

In your opinion is the source quality good enough for cleaning up in Audacity, or are you suggesting I should change the setup to reduce the infrasound - and if so how?

4

In your opinion is the source quality good enough for cleaning up in Audacity

No. I think there’s something in your room doing that. Are you using the microphone parked on the desk? You didn’t mention the spider vibration mount, boom stand, or anything else in that collection.

Screen Shot 2022-01-29 at 6.01.02 PM.png
How are you doing that, exactly?

There’s no shortage of tricks to this, but we need to know how you’re working right now.

The particular sample I’m attaching wasn’t actually recorded directly into Audacity

Should we be worried about that?

Koz

5

The “bass cut” or “100Hz rumble” filter curves in Audacity should remove all the infrasound.

In the raw recording, the worst audible problem is the clicking, but fixable with the DeClicker plugin …

suggested DeClicker settings.png
suggested DeClicker settings.png577×541 10 KB

6

The “bass cut” or “100Hz rumble” filter curves in Audacity should remove all the infrasound.

Yes, but. The low pitch noise associated with “home microphones” has a characteristic appearance in Analyze > Plot Spectrum. It’s different from this noise. This noise is more likely (in my opinion) to be air conditioning, computer fan noise, or building rumble being picked up through the table.

That’s why it’s important to know how the recording is made. Some of these noise makers are easy to fix. The cleaner you can make the original recording the better.

Koz

7

Absolutely.


The mic is mounted in a shock mount on a boom stand. The recording was done in a closet with carpeted floor and three walls dampened with curtains. The interface and computer were in the next room behind a closed door.

The other sources of sound could be:

  • my phone, about a foot from the mic
  • electric lights overhead - recessed low-voltage lights - I suspect these have transformers
  • mains power supply to the computer (no battery in the raspberry pi)
  • I’m pretty sure central heating was not on, but for this recording I didn’t check right at the time
  • ghosts, aliens, etc

Which of these is most likely given what you’re hearing?

8

Your room is very dead, (that’s a good thing).

IMO the clicks are mechanical rather than electrical: from the mic diaphragm.

IMO the infrasound is probably electrical from the audio electronics, rather than interference from other devices.
Provided you filter-out the infrasound with some sort of high-pass filter before any further processing it’s not a problem.

9

ghosts, aliens, etc

I’m going with aliens.

Power problems tend to group themselves around 50Hz or 60Hz depending on country. This doesn’t. See that little tip at 100Hz. It’s not 100Hz it’s 99Hz.

Screen Shot 2022-01-30 at 13.41.07.png
Screen Shot 2022-01-30 at 13.41.07.png1012×515 40.5 KB

Note on the left there are tones at 3Hz and the sound is still going strong. Also, that energy is at -37dB which if you were a cat, would be clearly audible. We have earthquakes at that pitch and volume.

Being an obsessive engineer (and that sort of thing can affect your sleep) I would (in my free time) find out where it’s coming from.

Anyway, yes, the first tool of audiobook mastering, filter curve, takes care of most of it. Do that first. Tones at that volume can throw the other tools off. That’s why this tool leads the pack in Audiobook Mastering.

You do have to deal with ticks and other mouth noises. That’s what killed my audiobook career.

Koz

10

At least there’s one good thing :smiley:

11

Pitching in having encountered a good many 2i2s, paired with various mics for v/o purposes on Mac and Windows (though never Linux). Is this a 3rd gen 2i2, or 2nd gen?

The noise is easy enough to remove, but I don’t think it should be there with that equipment.

It sounds to me like preamp noise, in which case it’s much louder than I’d expect from a 3rd gen 2i2. Or maybe it was introduced twixt 2i2 and arecord. Perhaps the Pi’s USB struggles to power the 2i2, introducing some noise to the signal. The NT1A itself should have all but zero self-noise.

I’ve heard it before from a client recording into Audacity with a 3rd gen 2i2 on a MBAir. In her case, the problem was solved by a replacement usb-c to usb-c cable. Does it occur on computers other than the Pi, recording straight to Audacity?

Maybe arecord was (and Audacity is?) hearing both the 2i2’s inputs, rather than just the desired input 1, and mixing them into a half-volume mono track, meaning you need more gain (and therefore more preamp noise) to achieve suitable peak volume? If you were seeing regular yellow flashes of the gain knob while recording that sample, the waveform doesn’t reflect it. Perhaps try recording two tracks into Audacity, one mono, one stereo, with as near as possible the same delivery. Then split the stereo track to mono and compare its peaks with the other mono. You may find you can dial down the gain a bit by recording in stereo, then splitting to mono.

Is gain knob 2 all the way down? I doubt that’s it, but good practice.

The NT1A ships with an extravagantly long 6m xlr cable - if you’re using that, maybe it encounters some interference en route to the 2i2. And if you’re not using it, try using it, in case it’s the xlr connection :slight_smile: Røde condensers have a longer ground pin than other makes (hence the little rubber washer they include in every pack).

Booth-wise, it does indeed sound quite dead but a touch boxy, as if you’re off-axis. Are you speaking directly into the gold dot? If you’re fully ensconced in the closet, maybe open the door to let some air in (and some of your bass boom out). With these mics, I’ve found it’s best to put some distance between the back of the mic and the surface directly behind it, and concentrate on deadening the space behind you - i.e. the last surface the echo will hit before bouncing directly into the front of the mic.

12

Hi GDepot, great points thanks.

It’s a 3rd gen. Power on R Pis is always an issue. I have some recordings to a laptop which I’m sure is powerful enough, but not with other conditions the same - I’ve seen similar noise on those but what I haven’t done is compared the levels. In the sample in this thread, I was facing at a slight angle to the mic, but the mic was directly aimed at me - if you follow me without me drawing a picture.

It would never have occurred to me that the gain on the other input would make a difference! I strongly suspect the point about a stereo source being mixed into lower-volume mono affects both the R Pi recording and the laptop recordings.

I will try all those variants…

13

Smashing, good luck.

My only other suggestion is that since Focusrite Control is Windows/Mac only, you may need something else to manage the 2i2 ‘properly’ on the Pi. If you’ve installed Focusrite Control on the laptop and encounter less noise, that may give you a clue.

Ha, yes. I think I get the picture - just so long as you don’t mean the mic is literally ‘aiming’ at you end-on, rather than side-address.