How is TOTAL RMS different than AVERAGE RMS

Question

I create spoken-word recordings.  One company that I work with specifys that the recording should be at a level that measures between -23 RMS and -18RMS.  It appears that audio editing programs label their RMS readings in various ways.  I had been using Average RMS but I think maybe this company is using a standard that comes closer to what ADOBE calls TOTAL RMS.  What attributes cause Audition to give one value to AVERAGE RMS and another value as AVERAGE RMS?

Answer 1

Total RMS used to be what happens if you add together all of the signals in your measuring time window (in Audition normally 50ms), and perform a RMS calculation on them, and that was a pretty standard way of measuring it. Average RMS in Audition is a little different, though. Reading through the somewhat cryptic help files about this reveals that to all intents and purposes, the RMS measurements will vary at different levels, because some attention is being paid to either Noise Criteria or Noise Rating curves. The lower the level gets with these, the more they favour lower frequencies over higher ones; in other words the figures reflect human annoyance factors rather than a specifically defined measurement criterion. And no, we can't tell you exactly how this is going to work, because the actual curves and weighting figures have never been made public, and it would take me several days to second-guess it from comparative readings. I simply don't have time...
It's been like this for ever. The original design of this was done by David Johnston, the program's original creator when it was Cool Edit, and I belive he's no longer associated even indirectly with Audition these days. Whether information about this is actually documented anywhere at all, I don't know. Questions about it come up periodically, and we can never answer them satisfactorily.
The original Cool Edit manual was just as useless about this. The exact words it uses to describe the Average RMS measurement are "Average RMS Power represents the average power of the entire selection. This is a good measure of the overall loudness of the waveform selection." And it's these words I've bolded that make the difference.
What happens with Total RMS measurements now seems to be a little different, and you end up with a figure much closer to a LUFS reading (you get this acccurately at the bottom of the statistics panel in CC). I know (because this is relatively recent) that the way the LUFS figure is calculated is according to the BS1770-2 standard - and that takes the whole program dynamics into account. What appears to happen with Total RMS in Audition is that it appears that an average of all of the windowed measurements, and that this in general turns out to be pretty close to the LUFS value as a rule.
Now, judging by the numbers they've quoted, the chances are that the company you are working for is using the recent standard for what you're doing - so if you are using CC, then it's pretty easy - select the whole waveform and quote them the value shown at the bottom of the screen. If you aren't using CC then there are several free LUFS meters available as VST plugins.
My personal view on this new standard, FWIW, is that it's a complete waste of time. It certainly doesn't achieve what it sets out to, or failing that all the broadcasters that use it have no idea what to do with it. And that very much includes the BBC, I'm afraid. The standard is supposed to keep overall programme dynamics within a certain loudness range, and save viewers/listeners having to adjust volume levels between programmes. And, it fails at that pretty comprehensively. Quite frankly, when it comes to broadcast dynamics standards, they'd have been better off forcing everybody to use the same settings on an Orban Optimod; that would almost certainly have worked better!