How can I objectively measure relative volumes?

dhjdhjdhj's icon

I have a collection of patchers, each one representing the core of the setup (MIDI and audio routing) for a different song from a playlist. The sounds for each patcher are typically completely different.

The biggest difficulty is figuring out how to match the overall volume (as perceived by listeners) of every patcher so that when I switch from one song to the next, the rest of the band, most of who are using in-ear monitoring systems (I am too) don't have to contend with widely varying output.

I am not in a position to pay close attention to meters while I'm playing.

I have seperate "outs" for pretty much every sound I'm using (these go to FoH) as well as a mixed stereo out for my own in-ear system.

What I'm wondering is if there is a decent way to capture the max volume values for each song, from each "out"(and perhaps the stereo pair) in a meaningful way so that I could play multiple songs and then go back and look at the data (somehow) and determine
A) which songs are out of whack with some reference song and
B) which "outs" contributed the most to the problem

I am using brick wall limiters to protect against real disasters but I'm really after a more fine grain solution)

Suggestions would really be appreciated (if bribes are allowed, I'll throw in a pair of tickets to one of our shows in Feb for someone who is in Chicago, Milwaukee, Phiadelphia or NYC that has a really feasible solution -:)

Peter Castine's icon

Have you looked at Tristan Jehan's analyzer~ external?

The thing is, although amplitude is simple to define and measure, volume isn't. It's not even entirely objective. If you dig through the psycho-acoustic literature, you'll find lots of curves indicating how different amplitudes at different frequencies are perceived as volume. You'll also find some formulae for approximating perceptual response to amplitude levels at different frequencies. I've probably got some documented somewhere on my hard drive. I don't know how much math you're comfortable with, though.

dtr's icon

Sounds like the easiest way would be to just audition each part and manually set the master gain so they match up.

dhjdhjdhj's icon

Thanks Peter. Unfortunately I've done zero advanced math in 35 years so these days, it would be beyond my capabilities (sigh). I do know that volume is very subjective, frequency dependent but I thought that some of the stuff they do for MP3 compression that takes psychoacoustic info into consideration might have helped.

But I certainly was hoping for a tool that would analyze an entire song and then give me some useful numbers (min, average, max) in some comparable way that could tell me by how much I should adjust the gain (or trim) for each song to match some kind of standard reference. One can always hope!

@dtr until yesterday, there was no such thing as a master gain in my system. I do now have a master trim whose value will be recalled automatically when a patcher is loaded and I'm probably going to be doing exactly what you suggest

Thank you both for the responses and happy new year.

D

dtr's icon

Oh there surely are analyzing tools like that, for music mastering for example. The iZotope VST range comes to mind, though I don't know how this could be integrated in a live Max environment. Maybe FTM / Gabor have something useful? http://ftm.ircam.fr/index.php/Examples

I'd still do it manually though, unless the realtime aspect is really essential to the project.

dhjdhjdhj's icon

Peter, thanks for the reference to analyser --- I'll take a look at that just to satisfy my curiosity.

Roman Thilenius's icon

while it is all true what peter is saying about percieved loudness, i dont think that should
matter too much for a show unless you are mixing rock, classics, and kl!kkl4k music.

have you tried using root mean square 60,000 seconds or something like that?
(which will also only work for you when it does not happen that one track starts
with a smooth intro and another one directly with a bassdrum)

-110

Roman Thilenius's icon

eventually i would preanalyze the tracks for "highest rms 4000 ms of whole track", this seem to be the most relevant information.

LiamCormacGould's icon

Have you looked at the Alex Harker externals? there are a pair of very nice analysis objects in there (one realtime one offline). These analyse loudness along with a bunch of other stuff.

dhjdhjdhj's icon

@Liam Looking for them now.