Why RMS window...?

[Fabian]

Why is RMS calculated in a window?

Not only Reapers master fader calculates a windowed RMS, but also Schwa's audio_statistics JS FX does a windowed RMS calculation.

Wouldn't it be preferred to calculate the RMS over the whole series of samples emitted during playback? I often (think I) need to know the RMS for an entire media item...

Or is there something I'm misunderstanding here?

[Dstruct]

Originally Posted by Fabian

I often (think I) need to know the RMS for an entire media item...

This would be more useful as a offline detection for information. But levelmeters are fine as they are for "live" readout. I'm using 600ms window most of the time.


Why do you need it for entire files? Most of the time those automatic algos don't cover all aspects and all dynamic ranges so you're still better doing loudness correction by hand (and ears).


Here two related feature requests:

http://forum.cockos.com/project.php?issueid=1509 (RMS weighting filters)
http://forum.cockos.com/project.php?issueid=2005 (RMS normalization)

[musicbynumbers]

The usefulness is because of the averaging of our ears, if you average over the whole file this is not very useful for representing the way our ears here (unless you take the readings of smaller windows and then average those)

[Bristol Posse]

Aside from the academic interest of knowing that the average RMS level of a media item over it's, for example, 5 minute duration was say -14 dBFS, I don't know how this would be useful or what it would tell you that could be useful in a mixing scenario.

It's like knowing the overall increase in elevation above sea level going from my house in So Cal to the city center in Denver CO is around 4000 feet. So what? That doesn't tell me anything about severity the four mountain ranges I would have to cross to drive there

Just my opinion of course
YMMV

[Fabian]

Thanks for the replies. Good points, makes sense...

My interest in the RMS for the total media item, is to get similar average levels (RMS) of several songs to be put on a CD. Sort of mastering... But I see now that this is not as useful as I thought. Of course, the ears do not average "globally" but rather "locally", over small chunks of the music. Yes, that makes sense...

[Bristol Posse]

I Know from use, that Soundforge has some statistical tools that can give you info that will allow you to get an idea of what the dynamic range of an item is which could be useful, I guess if you are looking at trying to set a group of songs to sit similarly in a collection

Personally I like VU meters because of their inherent (relatively) slow response which is typically around 300ms

PSP Audio used to have some free VST ones that you can drop on the track or master in question. What's nice about them is you can set the reference level to what ever you want so you could for example have the 0VU calibrated to -12 dBFS and then just use whatever tools you like to try and get the meters to dance around 0VU when you play the finished songs through them

This combined with actually listening as you finalize would probably get you a more cohesive, consistent volume level in a group of songs than messing about with average RMS levels
You could have two songs with the same average RMS level over the entire song that sound completely unrelated in terms of how comparatively loud they are since averaging anything hides most of the relevant facts

If I put Bill Gates in a room with 5 homeless people then the average yearly income per person in the room would be about a hundred million dollars. Clearly that is a completely useless number and nothing like what the actual real world case is
Similarly, I could have songs with radically different dynamic ranges and duration that end up with the same overall Average RMS level that would not sound anywhere close in volume when played back to back

[Fabian]

Originally Posted by Bristol Posse

Personally I like VU meters because of their inherent (relatively) slow response which is typically around 300ms

PSP Audio used to have some free VST ones that you can drop on the track or master in question. What's nice about them is you can set the reference level to what ever you want so you could for example have the 0VU calibrated to -12 dBFS and then just use whatever tools you like to try and get the meters to dance around 0VU when you play the finished songs through them

I'm just now using these http://www.naiant.com/vst/vu.html

[Bristol Posse]

Originally Posted by Fabian

I'm just now using these http://www.naiant.com/vst/vu.html

Yeah that's the same Idea as the PSP ones I've had for a couple of years now that I use on my output tracks to analog gear to make sure I'm getfing a good Line Level RMS signal going out of REAPER into my analog gear

I actually now use real analog VU Meters on the master since I sum OTB and use them between the analog summing bus and converters to make sure I'm hitting my AD converters around their good spot but the idea is the same

[carbon]

Originally Posted by Bristol Posse

I Know from use, that Soundforge has some statistical tools that can give you info that will allow you to get an idea of what the dynamic range of an item is which could be useful, I guess if you are looking at trying to set a group of songs to sit similarly in a collection

Wavosaur has something like that too.
Just press s for statistics.

[dub3000]

Originally Posted by Fabian

Thanks for the replies. Good points, makes sense...

My interest in the RMS for the total media item, is to get similar average levels (RMS) of several songs to be put on a CD. Sort of mastering... But I see now that this is not as useful as I thought. Of course, the ears do not average "globally" but rather "locally", over small chunks of the music. Yes, that makes sense...

use your ears. if you have a song that starts quiet and ends loud, the next song should have an appropriate matching level. average rms isn't going to help.