Volume unequal from file to file

Song 1 is great.

I used the same plugs and instruments on Song 2 and the Master meters show about -12 for both songs - but the volume is very different (lower) on Song 2.

Cakewalk Analyst shows nothing untoward, the song appears to have a decent showing in all frequencies with no horrible spikes that might account for the difference.

How can I identify the offending instrument or frequency?

Thanks - rev

Perceived volume is hard to measure objectively and has nothing to do with peaks or spikes.

The closest measurement for perceived loudness is RMS, and that has severe limitations. Think of RMS is an average over some period of time. For RMS meters, that period might be short, perhaps 40 or 50 milliseconds - too short to help you match two songs. What you need is an average of averages over a longer period of time, a measurement called Average RMS.

SONAR does not have the capability of taking such a measurement (and that's really too bad as it would be a nice feature), so you'll need to turn to third-party tools for this.

I use Adobe Audition, but that product's now defunct. There are, however, other wave editors that will report Average RMS for any portion of a song or even the entire song. Unfortunately, the free Audacity does not, AFAIK. Sound Forge might be one answer, if you can afford it.

The trick is in selecting which portions of the song to take a reading from. You generally want to skip intros and fadeouts that would skew the measurement. I like to look at three measurements: a typical verse, the final chorus, and the bulk of the whole song (minus fades).

Let's say you determine that song 1 has an average RMS of -8db and song 2 comes in at -12db, but both already peak at -1db. How do you bring them closer together? You can't turn the quieter one up because it's already peaked out. The answer is in your final stage of compression, the limiter settings on each song's master bus.

If you're using Ozone, for example, the threshold control on the Volume Maximizer is your magic volume-matching knob. You can even automate this, which I often do, especially on more aggressive and louder pieces.

Pick a target RMS value and tweak the threshold while watching the meters.  Find a threshold setting that causes the RMS meter to hover around your target value, and automate it if no single threshold works for the whole song. Use broad strokes with automation, though, and avoid sudden or dramatic changes, as you want the manipulation to be transparent to the listener.

Note that all of the above assumes that you have already achieved a consistent spectral balance in all your songs. If one song is especially bright it might be perceived as being louder.

You also want to establish a consistent style for vocal, drum and bass compression. A highly-compressed vocal can make the whole song seem louder. Having consistent production policies is the key to giving all your songs a coherent, identifiable "sound", and will make volume-matching easier.

Also accept that not every song on an album necessarily needs to match every other one exactly. Making a quiet ballad louder to match a rocker doesn't make sense.

Volume differences can also be creatively useful, especially when sequencing a CD compilation. Use volume differences to your advantage, for dramatic impact or so the compilation gets louder as it plays through from start to end. This may actually be necessary to maintain a consistent perceived volume, since the listener's ears will adjust such that song 2 doesn't sound as loud as song 1 just because it followed a loud song - even if it's the same measured loudness.