Compressor going over threshold rate (-12dB)

Audacity 3 Help Forum Windows
1

Hello there,

I am new to this forum so apologies if this problem has been stated and solved before.

I am having a problem with the compressor effect in Audacity. I have recorded a minute long speech (only vocal, no music) using a Blue Snowball USB microphone. The playback level at the loudest point is roughly -17db, so not particularly loud. I want to use the compressor effect to make the volume of the whole track relatively even and bring the entire recording down to -12db.

I have tried to do this using the compressor, but every time I set the threshold level to -12dB on the compressor, it always seems to go over it. I’ve tried it using by ticking both “compress based on peaks” and “Make-up gain for 0dB after compressing” as well as un-ticking those options to see if it helps, but so far nothing.

What am I doing wrong here? Any help would be greatly appreciated.

Ben.

Specs:

Windows 7 Professional x64 SP1
Blue Yeti Snowball USB Microphone
HP Pavillion m9162 uk-a

2

You mean “UP” to -12 dB? (-12 dB is louder than -17 dB. 0 dB is “full scale” and absolute silence is -infinity).

The Audacity compressor effect is rather quirky in that, by default, after compressing, the processed sound is amplified to bring the peak level up to 0 dB.

Probably the best way to use this compressor is to first amplify the track up to 0 dB peak (use the Amplify effect with default settings), then apply the Compressor effect using the “based on peaks” setting, then use the Amplify effect again, but set the “New peak amplitude” to the peak level that you require (remember that valid signal levels are negative numbers of dB).

3

With peaks at -17dB and the threshold set to -12dB, the signal will never hit the threshold and the compressor shouldn’t do anything. Except, the make-up gain will boost the levels.

The threshold isn’t a limit or ceiling, it’s the point above which you get compression. The ratio is the amount of gain reduction you get above the threshold.

Do you want the peaks to be -12dB? Why? That’s “louder” than -17dB, but still rather low… On the other hand, 12dB RMS or average is rather “loud”.

It’s more common to “normalize” the peaks at (or near) 0dB. I’d do that before compressing. Then after compressing, you can re-normalize (or use makeup gain) if you wish. Or, set the overall volume to whatever you wish after compression.

4

Thank you for the replies, I appreciate it.

Yes, I did mean “up”. Sorry, it’s late here and I’m a bit tired.

Thanks for clarifying about the threshold. I wanted the peaks to be at -12db because I read online that’s the standard for broadcasters and radio presenters.

I’m about to normalize the peaks so they’re at -3dB. Should I remove the DC offset? What impact would that have?

I am recording a speech for a friend’s short game that’s going on the App Store. What would you say is the optimum dB level for the voice recording? It’s going to play over the top of a piano song. I had a search on Google but I couldn’t find out. Does anyone know the industry game standard for voice recording levels?

Thanks again for all the advice.

Ben.

5

I have normalized the track to -3dB and applied noise removal. So far, so good.

A few times in the track (6 or 7), the sound peaks at 2dB. It doesn’t distort or clip, but it does sound a little odd with the rest of the track. Could someone please tell me how I could bring this down using the compression effect? I don’t want to bring the rest of the track up to this level, just bring the high peaking points down in volume.

I am totally new to Audacity and sound editing, so apologies if I am sounding pretty clueless. I have done a lot of recording in my life, but not a lot of editing/mixing. Thanks in advance.

Ben.

6

I shouldn’t worry about that. Broadcasters have their own standards for how they broadcast, but usually the source material (for example, commercial CDs) are generally normalized close to 0 dB. The broadcaster adjusts their playback levels prior to transmission so that it meets their standards. In other words, broadcast standards are for broadcasters and unless you run a radio or TV station those standards are not your concern.

To maximise the quality of a 16 bit recording (such as an audio CD) it is best that the peak level is close to (just under) 0 dB. -1 dB is a good level for CDs or 16 bit WAV files. For MP3s I prefer to go a little lower. The ACX guidelines for audiobook recording state a peak level of -3 dB (mono 192 kbps CBR MP3) http://www.acx.com/help/acx-audio-submission-requirements/201456300

7

“DC offset” is a fault. It is when the wiggly waveform is not centred around the “0.0” vertical position on the track. Enabling “DC offset correction” is generally a good idea when normalizing an entire track as it ensures that the waveform is exactly centred, though when using good sound equipment it is usually not necessary (but harmless).

A couple of links that explain this further:
http://manual.audacityteam.org/o/man/normalize.html
http://manual.audacityteam.org/o/man/dc_offset.html

8

Thank you so much Steve. Incredibly helpful stuff.

I have normalized the track to -3dB and applied noise removal. So far, so good.

A few times in the track (6 or 7), the sound peaks at 2dB. It doesn’t distort or clip, but it does sound a little odd with the rest of the track. Could you please tell me how I could bring this down using the compression effect? I don’t want to bring the rest of the track up to this level, just bring the high peaking points down in volume.

9

In what way does it sound “a little odd”?
Perhaps you could post a couple of short audio samples to demonstrate the issue - say 5 seconds of “normal” audio and a second file with 5 seconds of “problem” audio (preferably in WAV format). See here for how to post audio samples: https://forum.audacityteam.org/t/how-to-post-an-audio-sample/29851/1

10

It just sounded too loud in comparison to the rest of the track. The normalize effect seems to have solved most of the problem, but I still need a little help with the compressor.

Here is a short sample of the track. Could you please provide some feedback in terms of the overall sound, effects, what could be added, what sounds bad, etc.

https://app.box.com/s/u2jyrhhjc35ni7k8qhwyu8ognq2p0rh6

The gain on the mic is at 3/10
Windows mic level is at 26
Using the cardioid setting on the mic
About 6 inches away from the mic with a pop filter in front.

The reverb is the preset vocal 1.

11

That’s coming along pretty good and ticks most of the boxes.
For audiobooks, reverb is generally not used because it quickly becomes tiresome to listen to. If you are using it for special effect then I would recommend using it sparingly, otherwise for an audiobook, not at all.

12

Thanks a lot, I appreciate it. The reverb is only going to be used in small sentences which are set in the past. For thoughts in the present, I won’t use any reverb at all.

As for the compression, if I want the volume to peak at a maximum of -3dB, but occasionally it goes to 0dB, how would I do this with the compression feature on Audacity? I just want to bring the louder parts down. Thanks again for your responses.

13

You can also try the [u]Envelope Tool[/u] to manually fade up and down the volume where needed. Or, you can try the [u]Leveler Effect[/u].

There is also an application called [u]The Levelator[/u] that you can try. It automatically applies compression and automatic volume control. The Levelator is no longer supported but you can still download it.

I wanted the peaks to be at -12db because I read online that’s the standard for broadcasters and radio presenters.

The peaks will usually be at or near 0dB. -12dB peaks will be too quiet. (Compare that to any commercial MP3 you have.) I assume that -12dB recommendation is RMS or average (not peak). -12dB RMS is rather “hot”. Note that without compression, the ratio* between peak & average doesn’t change… If you boost the volume by +6dB, the peak & average will both increase by +6dB. Note that clipping is a kind of compression (the worst kind)… Once you are clipping, you can’t boost the peaks any more but you can still boost the overall/average volume (to a point).

Note that you can check the peaks by running the Amplify effect. Amplify scans your file and defaults to whatever gain is needed to reach 0dB. So for example, if Amplify defaults to +12dB, your current peaks are -12dB. (You can cancel the effect if you just want to check the peaks.) Finding the RMS or average is not so straightforward in Audacity.

I am recording a speech for a friend’s short game that’s going on the App Store. What would you say is the optimum dB level for the voice recording? It’s going to play over the top of a piano song.

The piano & voice together should peak at or near 0dB. Are they going to be mixed internally in the game in real-time, or will they be mixed in advanced? the mix/balance between the two needs to be adjusted ear. Obviously, the voice needs to be louder than the background music.

The gain on the mic is at 3/10
Windows mic level is at 26
Using the cardioid setting on the mic
About 6 inches away from the mic with a pop filter in front.

The settings don’t tell us much. For “home recording” the signal should generally be recorded for -3 to -6dB peaks. That allows headroom for unexpected peaks (your analog-to-digital converter will clip if you “try” to go over 0dB), but it gives you a strong signal for a good signal-to-noise ratio.

Pros often record at -12 to -18dB at 24-bits, with good low-noise preamps. This allows headroom for mixing and effects as well as unexpected peaks. (In reality, you don’t need headroom for mixing & effects because the software doesn’t clip.)

A few times in the track (6 or 7), the sound peaks at 2dB. It doesn’t distort or clip,

Audacity itself uses floating-point so it can go over 0dB internally/temporarily without clipping.

Your analog-to-digital converter and digital-to-analog converter will clip if you go over 0dB. “Normal” WAV files and CDs will clip if you try to go over 0dB (so you should make sure the peaks are at 0dB or less before rendering/exporting.

Nothing bad happens when you get near 0dB.

The reverb is the preset vocal 1.

Of course, that’s an entirely artistic choice.

\

  • Since decibels are logarithmic, the difference (subtraction) is a ratio.
14

The levelator works very well for me, thanks for letting me know about it.

The piano and voice will be mixed in advance. The piano will play at -3dB and will be lowered in volume significantly when the voice recording is playing over the top. This will also be at -3dB.

The settings don’t tell us much. For “home recording” the signal should generally be recorded for -3 to -6dB peaks.

So is recording at near -12dB and then normalizing to -3dB not advised? The microphone I am using is incredibly sensitive and seems to pick up a large amount of background noise. The lower I have the gain on the mic, the closer I have to speak into it, but this also means I get less ambient noise.

Pros often record at -12 to -18dB at 24-bits

I am currently exporting the file as a 16 bit PCM .wav file. Audacity gives me the option to save as a 24 or 32 bit file. I have been told the higher the bits, the better the quality of the audio. With a simple speech recording, is there enough to be gained to constitute the enlarged file size?

15

So is recording at near -12dB and then normalizing to -3dB not advised?

If you are getting good results, it’s fine.

The microphone I am using is incredibly sensitive and seems to pick up a large amount of background noise. The lower I have the gain on the mic, the closer I have to speak into it, but this also means I get less ambient noise.

The gain setting has no effect on the relative acoustical noise. Boosting the gain will boost the signal and ambient noise together. That means the gain setting is NOT CRITICAL (as long as you don’t go into clipping). If you have electrical/preamp noise, some of that noise doesn’t go down when you reduce the volume, so you need a strong signal to overcome it. There’s always some electronic noise, but usually it’s low relative to the ambient noise so it may not be a problem.

If a lower gain setting encourages you to speak in a stronger voice and get closer to the mic, that may help. (As long as you don’t get too close to the mic where you start getting other problems… 6 inches is usually about right.)

I am currently exporting the file as a 16 bit PCM .wav file. Audacity gives me the option to save as a 24 or 32 bit file. I have been told the higher the bits, the better the quality of the audio.

16-bit/44.1kHz (CD quality) has been shown to be better than human hearing by people doing blind ABX tests between a high-resolution original and a downsampled copy.

It’s very-likely that the audio will be compressed for the game anyway. (NOTE - File compression (such as MP3) is totally unrelated to the dynamic compression of the compressor effect.)

On the other hand… The current pro studio standard is 24-bit 96hHz. (Most pros think high resolution audio sounds better, but most pros have not done scientific level-matched blind listening tests. :wink: ) And if the peaks don’t hit 0dB, you are not using all of the bits and you are loosing resolution. (The blind tests are usually done with normalized files utilizing all of the bits.) Each bit represents 6dB of dynamic range so a 16-bit file that peaks at -12dB is only using 14-bits.

If you are not finished (i.e. if the piano is not yet mixed-in, etc.) the “best practice” would be to export to 32-bit floating point. But, that’s probably overkill and 16-bits should be fine.

High bitrate MP3 can also often sound identical to the uncompressed original in a blind listening test, but it’s still “bad practice” to use lossy compression in audio production. If you want MP3 (or another lossy format) the compression should be done once as the final step.

16

I would normally record a little higher than that - usually around -6 dB peak, but as DVDdoug says, if you’re getting good results, then that is what matters.
It is more important that you do not record too loud. Hitting 0 dB is very likely to cause audible damage, so it is definitely recommended to allow a little headroom.