Hi, everybody. I thought I’d share something I’ve been struggling with while using Audacity and see if anyone has some help or insight to share on this.
I use Audacity to put together a music program I then play on a non-commercial FM station. It’s a bit of an eclectic program - I play music in a range of genres, both old and new. One of the things I try to do with Audacity with mixed success is make the songs match as much as possible volume-wise.
What I tend to do is amplify the volumes of older songs to peak while I lower the volume of newer songs to about -1.4 to -1.8 of peak. (I also often use the bass and treble effect on older songs before this to give them a boost and to make them sound more vibrant.)
Yet, even though I do all this and the resulting audio file sounds good at home, when I play it at the station, some of the newer songs often surprise me by being much louder than the surrounding tracks. These tend to be the poppier, more dance-orientated songs.
So . . .
Why does this happen? If I’m boosting the older songs to peak and throttling the newer songs back, why are the newer songs still louder? Is there anything I can change in Audacity so it matches the levels I’m seeing when I’m playing it at the radio station?
New Music is denser from having been processed and compressed. It sounds louder without triggering radio station broadcast processing which is what simple “making the blue waves taller” would do.
even though I do all this and the resulting audio file sounds good at home, when I play it at the station, some of the newer songs often surprise me by being much louder than the surrounding tracks. These tend to be the poppier, more dance-orientated songs.
Those would be the techno/disco tracks where the sound meter goes up to maximum and never moves except for a slight vibration for the duration of the song. Perfectly normal.
What you did was bring the overall volume of the dance tracks, peaks and loudness, down to match the older songs. The station processing saw the change and since you never changed the character of the song, cranked it back up to “where it’s supposed to be.”
The only way around that is to densify (technical term) the older tracks.
I don’t know that there is a straight-line process for that in Audacity.
You might try Chris’s Compressor. Chris wrote a compressor so he could listen to dynamic, light and airy opera in his noisy car. Not usually a good combination.
I use it with the first variable, Compress ratio cranked up from the default 0.5 to a stiffer 0.77.
This is a waveform analysis when I do that. The show gets noticeably louder.
There is one known bug. Leave a little extra tail on the end of each song you process. Chris is a look-ahead compressor and doesn’t like “falling off the end of a song.” Cut it off before you use the tune.
There may be a process to densify with the Audacity natural tools, but I don’t know of one. You can dig yourself a hole with the Audacity tools because most of them work on relative peaks and not overall loudness. So it’s not hard to imagine a six or eight step process to compress your older songs. How close are you to retirement?
In the old days, radio stations used a hardware compressor/limiter* in real time. I assume they still use real-time compression/limiting (although it may be done in software). Of course, anything live with a microphone has to be done in real-time.
A limiter in front of the transmitter is mandatory because it’s illegal to over-modulate. (Of course they are not required to crank-up the volume to the point where the limiter kicks-in).
Audacity doesn’t work in real-time… It’s for processing/editing audio files. So if you want to use Audacity you’d have to pre-process your files before broadcast.
Some broadcast/streaming software has compression and limiting built-in. So, check whatever application you’re using to see if it has compression/limiting.
There is another method which matches volumes without “densifying” your music. ReplayGain, MP3Gain, and WaveGain all work by analyzing the loudness, then reducing the volume of the louder songs and by sometimes boosting some quiet songs if necessary and if it can be done without clipping (distortion), or if you allow clipping.
The ReplayGain method is preferred by listeners who wish to keep the dynamic contrast… The loud parts of a song remain (relatively) loud and the quiet parts remain quiet… The volume is not adjusted in the middle of a song.
The downside is that, overall, the music is quieter. That’s usually OK “at home” but in commercial radio every station is trying to be louder than every other radio station. (If you have enough analog gain you crank-up the volume control and listen as loud if you want.)
ReplayGain has to be supported by your player software (Apple has something similar called Sound Check). The actual audio in the file isn’t “touched”. The file has to be pre scanned to get the loudness and a tag is added to the file that tells the player-software how much to adjust the volume during playback.
MP3Gain and WaveGain are used to “permanently” alter the volume of the file (or a copy of the file) so they work with any player.
Although limiting and compression tend to “push down” the peaks or the louder parts, “make-up gain” is used to bring-up the overall loudness. The end result you get louder (or more intense) sound without clipping/distorting the peaks. (Or at-least less distortion than you’d get if you simply crank-up the volume into clipping.)
I assume they still use real-time compression/limiting (although it may be done in software
Last I checked, nobody was interested in digitizing delays in the broadcast chain, but no, it’s not done with separate devices any more. I don’t remember the name of the company, but someone came up with a way to marry the broadcast processing to the actual FM generator. It did a remarkable job and I can’t imagine it’s gotten worse since then.
they are not required to crank-up the volume to the point where the limiter kicks-in
Although there is a range. I worked for a station that was so terrified of an overmodulation citation that they collected a citation for undermodulation. I think the director of engineering still has that framed over his desk.
Repay Gain is a playback function, right? How does the player “know” what to do?
“Peak level” is a poor measure of “loudness”. A better measure is “RMS level” (a kind of “average level”).
The “Contrast” tool (Contrast - Audacity Manual) measures the RMS level, so this could be useful as a guide. If two tracks have similar RMS levels, then they are likely to sound about the same loudness.
Thanks for your responses. I’ve been trying out kozikowski’s suggestion of using Chris’s Compressor on the songs that sound the weakest and, once I figured out how to trick it into not boosting the volume at the end of the songs, it’s been doing the trick. I think that solving the problem of certain songs that I could never get to sound loud enough.
Now I need to solve the problem of some newer songs that surprise me by sounding much louder than I expected. steve I just saw your response and I’ll try out the Contrast tool to see if it can identify those songs for me ahead of time.
I need to solve the problem of some newer songs that surprise me by sounding much louder than I expected.
There is a tool now to put the loudness somewhere instead of measuring it and bemoaning that it’s in the wrong place. Steve wrote RMS Normalize to help out with audiobook problems. Audiobook presentations have an RMS specification.
It can measure and set RMS which, while not perfect, is enormously better than anything else. RMS is Root Mean Square, an electrical measurement of work/energy. That measurement works out roughly to loudness and has the advantage that it’s a known, stable unit of measurement (even if nobody but the engineers has ever heard of it).
Audiobook readings are almost always too quiet and so the tool is used to boost the loudness. Goodness knows what would happen if it had to bring volume down.