RMS or Just BS?

**aj113** · 06-22-2012, 04:59 PM

OK so I got the TT meter that everyone bangs on about, all it's done is confuse me.

I see Brandon going on about mastering at -6db, -4bd etc - is he talking about the average RMS for the recording? Or is he talking about just what his eyes are checking out on the realtime meters at any given point during the recording?

I just checked a really meaty section of one of my recordings - Sound Forge told me the average RMS is -11.3 and TT told me it was -10.4. I mean, that doesn't sound too loud does it, but when I play the section live and watch the meters they rarely dip below -0.6 in SoundForge and -0.3 on the TT meters.

If I made it any louder the waveform would turn into a brick.

What's that all about?

**acidfrost** · 06-22-2012, 05:17 PM

I think he's talking about realtime RMS. So looking at the moment of your loudest part it should show around -6 -7 db rms.

**aj113** · 06-22-2012, 05:25 PM

Wait what? My loudest parts should show 0db surely? Even if all you did was normalize the wav without any limiting or compression the loudest parts would be 0db. Do you mean the quietest it should go to is -6db to -7db?

**acidfrost** · 06-22-2012, 05:44 PM

It should be 0 db Peak. but -7db RMS.

RMS is an average of the volume at a given time, something like that. peak is volume at a given time.
So your loudest moment should have an RMS of about -7db which is still super loud.
If you had an rms of 0 db it would sound like distorting white noise lol.

1.jpg

**rook2c4** · 06-22-2012, 06:47 PM

Originally Posted by aj113

but when I play the section live and watch the meters they rarely dip below -0.6 in SoundForge and -0.3 on the TT meters.

Come on, think about it for a moment:
If the level meter would ride with every peak and valley within the waveform during playback, it would be moving so fast that it would be little more than a blur... which would make it practically useless as a visual tool. The level meter is more of a rough appoximation of what is actually going on, level-wise.

**aj113** · 06-23-2012, 02:35 AM

Originally Posted by rook2c4

Come on, think about it for a moment:
If the level meter would ride with every peak and valley within the waveform during playback, it would be moving so fast that it would be little more than a blur... which would make it practically useless as a visual tool. The level meter is more of a rough appoximation of what is actually going on, level-wise.

Come on think about it for a moment. If the meters are showing a rough appoximation of what is actually going on, level-wise, and my eyes are seeing -3 to -6, why does a static scan of the same section show an average of -11.3?

**shackman** · 06-23-2012, 02:58 AM

AJ. I can't explain it. But I can say you're the first I've come across who finds the TT meter hard to understand.

**aj113** · 06-23-2012, 03:14 AM

I see, so......if you understand it, why can't you explain it?

**fHumble fHingaz** · 06-23-2012, 04:32 AM

Ok, I'll try...
If you take a finished stereo file, import it into your DAW & zoom right in as far as you can go, you'll see the big peaks/downbeats in the song peaking right at the top of the waveform, but then the waveform tapers off quite a lot before the next downbeat. During that tapering off, some of the smaller sections of waveform are pretty much half the height of the peaks.

Maybe the waveform peaks at say -0.3dBFS, but then it might go down to much less before the next downbeat. The TT meter calculates it's RMS average across that whole section of waveform, so although the peak meter might looks like it is only bouncing between, say -3dBFS & -6dBFS, it is actually dropping much lower for certain moments (we're talking samples here).

So here's a 90's song waveform zoomed right in Mixcraft 6 with the TT meter. It is averaging around -12dbRMS, but for most of the time the meter appears to jump between -3 & -6dBFS:
TT.jpg

Now, here's that same file zoomed pretty close (but not right in) in Sonar X1. Sonar has a (very helpful) metering scale on the LHS of the waveform. You'll notice that the highest peaks show above -3dBFS, almost right up to 0dBFS, whereas the lowest visible "troughs" appear around between -18 & -15dBFS. If I zoomed in even closer, the higher resolution would reveal that the deepest "troughs" are even lower:
Waveform.jpg

I don't know the exactly over what timing period the RMS is averaged in the TT meter, but I do know that it can vary from plugin to plugin... So depending on which meter you use, you may get slightly different results... For example, on Slate Digital's FGX plugin, you can vary the time over which the meter measures the RMS average... but unfortunately it only describes them as "Fast", "Medium" & "Slow", not in milliseconds or whatever.

Nonetheless, hopefully that explains why -6dB RMS is DAMN LOUD, -4dB RMS is pretty much insanity!

Clear as mud?

**aj113** · 06-23-2012, 04:47 AM

Yes, thank you. I actually understand that, but it is not what I am asking, so I need to explain myself in more detail.

If you now take that 90's wav and run it through the standalone version of TT, it will not show an average rms of -12db, it will be much lower.

Now, I know the reply to that will be that there are usually some really quiet sections of a song which invariably knock the average rms rating down even further, but I am saying that this low scan rate happens even if you do a scan just of the loudest sectionof the wav.

Similarly if you do a scan in Soundforge, it will show a much lower rating than -12dbrms.

So my questions are:

a)Why?
b)Which rating is right?
C)If the standalone/scan is right, what is the point of using the realtime meter in the first place?

**fHumble fHingaz** · 06-23-2012, 05:11 AM

If you now take that 90's wav and run it through the standalone version of TT, it will not show an average rms of -12db, it will be much lower.

Ok, I did that, but I got a different result than what you predicted:
Offline.jpg

So my questions are:

a)Why?

I can only assume that the answer has to do with the time periods that the rating is taken over... here's a thought - is there silence at the beginning & end of the file you are measuring? - That would make a BIG difference, particularly if it was for a few bars.

b)Which rating is right?

I'd guess that they are both right, if you take into consideration the time period being measured.

C)If the standalone/scan is right, what is the point of using the realtime meter in the first place?

Because you get a realistic idea of the overall loudness at that moment in the song. By my reckoning, if you had a song with Death Magnetic-style "square wave" clipped madness for most of the time, interspersed with a quite a few sections of really quiet stuff every now & then, it might actually appear to be a sensible RMS level on the standalone meter, even when it actually isn't for most of the time....

...Not certain about that, but it seems logical.

**acidfrost** · 06-23-2012, 05:29 AM

As per your first question, for the Slate Cup you use realtime, not average. So you press play and look at the TT while it's playing the loudest section.

**shackman** · 06-23-2012, 05:30 AM

Originally Posted by aj113

I see, so......if you understand it, why can't you explain it?

I read the mixing level as RMS. That's "HOw Loud" the mix is. My max is usually -6db and often closer to -8db.
I watch the peak meters to ensure no clipping.
I watch the dynamic range meters (in the centre) to see I don't squeeze that below 6 - usually 8.

That gives me chunky, loud dynamic mixes.

I may be using the thing totally incorrectly, but that's how they told me to use it and that works for me.

Whhops just seen Fumble's explanation. Nice work FF

**acidfrost** · 06-23-2012, 05:49 AM

Yeah Shackman, you're doing it right! lol

**aj113** · 06-23-2012, 07:07 AM

Originally Posted by fHumble fHingaz

Ok, I did that, but I got a different result than what you predicted:
Offline.jpg

So my questions are:

I can only assume that the answer has to do with the time periods that the rating is taken over... here's a thought - is there silence at the beginning & end of the file you are measuring? - That would make a BIG difference, particularly if it was for a few bars.

I'd guess that they are both right, if you take into consideration the time period being measured.

Because you get a realistic idea of the overall loudness at that moment in the song. By my reckoning, if you had a song with Death Magnetic-style "square wave" clipped madness for most of the time, interspersed with a quite a few sections of really quiet stuff every now & then, it might actually appear to be a sensible RMS level on the standalone meter, even when it actually isn't for most of the time....

...Not certain about that, but it seems logical.

Did you miss this bit?

Now, I know the reply to that will be that there are usually some really quiet sections of a song which invariably knock the average rms rating down even further, but I am saying that this low scan rate happens even if you do a scan just of the loudest section of the wav.

Suggestion: - upload the file in question so that anyone who is interested can do their own measurements.

**fHumble fHingaz** · 06-23-2012, 07:22 AM

I surrender

.

I don't know the answer to your question - I'm just trying to use logic to work it out ... I did read what you said about the file appearing to have a much lower rms even on a short section with the off-line TT.

I swear I haven't used any trickery to manufacture the results I got, I just dumped the file into TT offline, & that's what came out. I figured that might be proof enough. I'd be happy to share the file if it were my own composition, but it's not, so I won't.

Maybe we should try an experiment with one of our own files.

**venuestudio** · 06-23-2012, 08:11 AM

Let's see if this is helpful.

The two measurements we are talking about are peak and RMS. One is derived from the other and are not different depending on the company writing the code for the plugin. These are absolute values. RMS is equal to peak voltage divided by the square root of 2. It does not ever change from that number. What does change is how often that measurement is taken by the software, as in number of samples that go by before the calculation is made. If you sample too quickly you would not be able to see the changes. If you sample to slowly you run the risk of missing the peaks. That can lead to inter-sample peaks which is a whole 'nuther problem I will get to later. Regardless of how you do it the peak value and RMS value are taken at the same moment in time. It has to be this way or the measurement is meaningless.

The RMS value comes from the peak value just as peak to peak value comes from the peak measurement. You can get the RMS value by dividing peak by 1.414. That also means you can get the peak value by multiplying the RMS value by 1.414. That is significant because it answers the question why would the RMS value not be 0dB? Because if it is then the peak value would be 1.414dB ABOVE 0db. Distortion! The beauty in having realtime metering is that you can track when a signal gets too hot (overage) and use it to identify were you need to adjust the signals to remove the peaks.

I talked about inter-sample peaks before. This happens when when a sine wave peaks between samples taken by the software. SSL has a free meter that you can get that identifies this problem and helps you to fix it. They have a nice write up on the website HERE were you can read about it and download it. I highly recommend it. I use it along with the TT on the master bus.

**aj113** · 06-23-2012, 08:13 AM

Originally Posted by fHumble fHingaz

.....I swear I haven't used any trickery to manufacture the results I got,....

I'm not suggesting you did. Your answers to my questions were based around the possibility of there being quiet sections in the wav that I scanned. I had already stated that I was scanning a small, loud section. No quiet sections, no silence.

OK I'll post something that illustrates what I mean.

**acidfrost** · 06-23-2012, 08:17 AM

I'm not suggesting you did. Your answers to my questions were based around the possibility of there being quiet sections in the wav that I scanned. I had already stated that I was scanning a small, loud section. No quiet sections, no silence.

OK I'll post something that illustrates what I mean.

I don't understand where you're going to. Your original question was about brandon asking for db rms being around -6db/-7db. What's your question now?

**aj113** · 06-23-2012, 10:29 AM

Originally Posted by acidfrost

I don't understand where you're going to. Your original question was about brandon asking for db rms being around -6db/-7db. What's your question now?

I'm not going anywhere other than my original question, which was this:

I see Brandon going on about mastering at -6db, -4bd etc - is he talking about the average RMS for the recording? Or is he talking about just what his eyes are checking out on the realtime meters at any given point during the recording?

followed by this

I just checked a really meaty section of one of my recordings - Sound Forge told me the average RMS is -11.3 and TT told me it was -10.4. I mean, that doesn't sound too loud does it, but when I play the section live and watch the meters they rarely dip below -0.6 in SoundForge and -0.3 on the TT meters.

I'm not sure I can explain it any simpler but I'll try:

1. Play aloud section of a wav. TT realtime RMS meters show -7db
2. Save that section of wav as a new file, then use the standalone version of TT to scan it. RMS is significantly lower.

Same section of music, same tool, different reading. Why?

**aj113** · 06-23-2012, 10:33 AM

Originally Posted by venuestudio

Let's see if this is helpful.
RMS
The two measurements we are talking about are peak and RMS. ......

No. The two measurements I'm talking about are both RMS. Same section of music, same tool, one reading is realtime, the other is a scan of the entire section, but the readings are diffeernt. Why aren't they the same? Or at least similar?

**acidfrost** · 06-23-2012, 10:43 AM

I don't know why it's lower but it's different because one is realtime so it takes in account the RMS at that precise moment. It doesn't take in account what was before or after where as the other takes the whole part.

Weird that it's lower though.

Brandon wants realtime.

**aj113** · 06-23-2012, 11:09 AM

OK just to clarify - I don't really care what Brandon wants because I am not involved with the competiton, I just want to know which is the 'real' average RMS.

Originally Posted by acidfrost

I don't know why it's lower but it's different because one is realtime so it takes in account the RMS at that precise moment. It doesn't take in account what was before or after where as the other takes the whole part.

Weird that it's lower though.

Brandon wants realtime.

As you say, weird that it is lower. I want to know why, and I want to know which one is 'right' .

**acidfrost** · 06-23-2012, 11:13 AM

Personnaly I use the realtime one. I can really hear that 7db realtime is really loud where 10db realtime is ok loud. So I use realtime because it seems the correct one.

**venuestudio** · 06-23-2012, 11:30 AM

Originally Posted by aj113

No. The two measurements I'm talking about are both RMS. Same section of music, same tool, one reading is realtime, the other is a scan of the entire section, but the readings are diffeernt. Why aren't they the same? Or at least similar?

They are different because the realtime version is taking only a short slice of time, a snapshot of the signal. The scan of the entire section is a look at a much larger portion, possibly the entire song or clip. They should be different. The standalone version may be averaging in a peak hold, which is kind of like taking the worst case for the entire song.

If that is the case, which it probably is, they are both correct. It is much like fuel economy right now verses over the span of a whole trip.

Another thing to look at is gain. I suspect you have a gain setting difference between the two clips. That would certainly account for the difference. I have noticed in Logic that I can see lower metering once the song is bounced vice when I am listening to the tracks together.