I have WAV files from different sources and I want to balance them so that they sound relatively equal in volume when played back. The files differ in length and spectral properties, so average and peak levels vary a great deal. The Normalize function does not seem appropriate. Is there another method?
Normalize has good name recognition and great press, but all it does is turn the volume up or down…once. It doesn’t reassign volume values on the fly, which is what you were hoping it did.
relatively equal in volume when played back.
That’s what it says on the tin for Chris’s Compressor.
Chris designed it to even out volume variations so he could listen to opera in the car. I use it once a week to even out the volume variations in a podcast download…so I can listen in the car. I don’t use the factory values. I change the Compression value (top value from 0.5 to 0.77. When I do that, Chris produces the same volume evening-out as the NPR radio station which carries the same show on the air.
If you don’t do that all-in-one solution, then you can use Effect > Compressor, Effect > Limiter, and then Effect > Normalize, which you’ve already met. The limiter is a special effect and I don’t remember where the instructions are.
You can also use the Envelope Tool to graphically click and push the volumes up and down manually through the show.
If you’re going to do that, you should probably get a lot better at time line management, Zooming and turn off Timeline Auto Update.
I am conducting research on nurses’ perception of auditory signals emitted by medical devices–alarms and such–and I want to minimize differences in Loudness/Volume across individual WAV files as a factor. I am not trying to balance (compress?) loudness within a given signal/WAV file. Rather, I want different signals to sound equally loud by some standard. My naive assumption is that this would best be done by balancing peak dB levels, or average dB levels across the duration of individual signals. But, I don’t see a way to equate different WAV files.
There is no actual “standard” because "loudness is subjective (consider an elderly person that has reduced hearing for high frequencies - a fairly “quiet” low frequency sound may sound “louder” than a “loud” high frequency sound). “Loudness” is actually very complicated because there are many factors affecting how we perceive it (Loudness - Wikipedia). There is however EBU R128 which is a “recommendation” rather than a “standard”. EBU R128 is available here (PDF) https://tech.ebu.ch/docs/r/r128.pdf
The simplest approach is to use the same RMS level. This is a kind of “average” level (Root mean square - Wikipedia)
The RMS level is a much better approximation of loudness, though it takes no account of hearing being more sensitive to some frequencies than others (Equal-loudness contour - Wikipedia). it is however a measurable and verifiable scientific measurement. There is a plug-in available for Audacity that can “normalize” to a specified RMS level: RMS Normalize
A more elaborate approach is this plug-in, based on the ReplayGain algorithm. ReplayGain plug-in
The “New Version” allows you to amplify to a specified “loudness” level.
There is also a 3rd party application called “Wave Gain” RareWares - Other audio tools and formats
The second link is a version that is optimised for recent versions of Windows and is marked “Download (135kB)”.
I’ve not tried this myself.
There is also a very good free audio player called Foobar2000.https://www.foobar2000.org/
This has “equal loudness normalizing”, though I’m not sure if the normalizing can be applied to WAV files.
Thanks, Steve. I am a cognitive psychologist, so I understand the subtleties of perception. That’s why I’m thinking about this question. I’m not an acoustician, though, so I need help with the ‘how to’ part of it. It does sound like (no pun) balancing RMS is a good way to go.
as I wrote, the advantage of normalizing RMS is that it is easily verifiable, but the disadvantage is that it ignores the effect of frequency, and that can make a big difference to perceived loudness. On the other hand, the ReplayGain algorithm does take frequency into account, and generally gives a better measure of perceived loudness.
I want to minimize differences in Loudness/Volume across individual WAV files as a factor.
The step after that is most important. What are you going to do with these files? The instant you play one to an audience you involve the speaker system and its characteristics.
Do you have a Sound Pressure Level meter? It seems you would not be able to get past “hello” without that. Radio Shack used to carry one, but I see they don’t any more. It had the advantage of being cheap.
A-weighting simulates the lumpy characteristics of human hearing for low level noise (which is what it was intended for). As can be seen from the equal loudness contour, the “lumpiness” is much more pronounced at low SPL.
Often the case yes, but that’s due to the proliferation of A-weighted meters rather than because A-weighted is the best measure. C-weighting is actually a better measure for loud noise, and D weighting for extreme noise. There’s a bit about this on Wikipedia: Sound level meter - Wikipedia