A Learning Experience? Peak/RMS Levels

https://www.dropbox.com/s...ln3uu37jn/EFT%20CV.jpg


Console View screen shot
Sorry Photobucket was messing up and dont know how to get Dropbox to post pic

I use the console view to compare relative audio levels from one track to the next.... working mainly in the track view.

Another angle for looking at this, Mike, is the crest factor. That's simply the ratio of peak to RMS, but it's the closest thing we have to an objective measure of what's referred to as "punch" or "sparkle".

In your example values, your kick has a crest factor of 4.5 (18 / 4). That kick is probably going to sound punchier than one with a crest factor of, say, 2 (which is what you might get from compressing it heavily). 

I include the caveat "probably" for two reasons. First, we don't know if the -4db maximum peak is typical or an anomaly, and second, we don't know the frequency content of the peaks. You could, for example, have a "ticky" beater sound typical of metal music that's not very punchy but has a high crest factor.

Vocals, OTOH, sound best with a low crest factor in most genres. A high crest factor might indicate a vocal with too-prominent "T"s and "K"s. Your lead vox crest factor of 2.5 is fine, but might suggest that more aggressive compression at least warrants a test.

You can lower the crest factor by either shaving peaks or by raising the average level while keeping peaks the same (depending on whether you want to adjust overall perceived volume or not). You can raise the crest factor by decreasing compression ratios. Tape sims lower crest factor. A transient shaper can be used to go either way (I like the one in Superior Drummer a lot for kick, toms and snare).

In my own projects, I look at the crest factor of the finished (mastered) product as an indicator of possible over- or under-limiting. I don't usually measure it, but can tell by looking at the waveform whether it needs a little more or a little less. My favorite productions tend to have a high crest factor, although what's acceptable depends on genre and style - a quiet ballad will have a lower crest factor than an aggressive jazz-funk piece.

Here's a slappy bass guitar with light compression and a modestly high crest factor. Peak is -8db, average RMS is -21db, for a crest factor of 2.6. Note how you can "see" the crest factor by comparing the thick area where most of the samples live to the spikes that represent each note attack. Depending on the source, the peaks may take on the appearance of grass or hair, and I've actually heard the term "hair" used in this context.


Here's the same clip with heavier compression to trim the peaks. The highest peak is now -11db, but average RMS has only dropped to -22db. The perceived volume didn't change much, but the crest factor as dropped to 2 and the track sounds less punchy.

You sure your two columns weren't mis-labeled? How can you have RMS levels higher than peaks?

Anyhoo, generally you should go by ear, as others have said. This is where a calibrated listening environment can help, so that you know that a certain master bus dBFS level (with a test signal) corresponds to a certain SPL in the room. That dBFS to SPL mapping is set up so that anything with too high an RMS will sound too loud and anything with too low an RMS will sound too low. It'll naturally push you towards an appropriate level of compression. So you just arrange for the right dBFS levels on the master bus, and then use your ears to get the appropriate level of compression.

yea, those numbers don't look right

Firstly I think the OP should do some research and reading up on rms and peak values. A good simple start might be here: 

http://myhometheater.homestead.com/rms.html 

The OP values are wrong as peak values are higher than rms values. But you have to realise there is quite a difference between peak/rms relationships of a continuous sine wave compared to a musical event such as an acoustic guitar pluck. With the guitar pluck, new strings, heavy pick etc there are two parts to the sound. The first is the attack transient part of the sound and it usually happens first followed by some sort of sustained part. 

Another thing to add to the confusion is that in audio engineering we are using a standard where the very peaks of the continuous part of the waveform are actually considered as the rms value not the point 3 db down from that.

So now start thinking with the pluck sound the highest part of the transient at the start may reach a value and it may be -4dB FS but then after that the sustained potion of the sound may settle around -14 dB FS for a much longer period. In musical terms the peak value could be said to be -4 but the rms part of the sound may be considered to be around -14. (In DAW terms that is)

In musical terms you cannot think about the rms being .707 x peak as it is in electrical engineering because it does not apply so much that way. In music it is more about the acoutsic guitar pluck I have just mentioned.

Start thinking about whether any part of a sound over time is either peak or rms orientated. Very percussive sounds have a high peak value but are very quick over time and they have little or no sustained portion so we can think of them has having a low or no rms value. But someone singing AAHH however will have a relative low peak value that might only equal the same value as the longer sustained value eg rms part of the sound.  And a sound can have quite a high peak transient level followed by a lower average or rms level or sustained part level.

Most DAW meters are useless at effectively measuring the sustained part of the signal. And Sonars meters show true rms value and are 3 dB lower than the standard making it harder.

You need VU meters (either real or VST) to effectively measure rms levels. The ballistics of the meter also play an important part in terms of what they respond to and show. You also need a VU meters to respond and show you a 0dBu reading on the VU meter itself while working at the chosen digital reference. Something most DAW's (except Studio One) cannot do.

A system should be used such as the K System to keep  rms levels on tracks and busses even and consistent. You do not worry about the peaks so much. They take care of themselves and vary depending on the sound as to how much higher they are than the rms or sustained values. But as the headroom is built in you don't have to worry about them. This is far superior to the approach of not taking notice at all of what your meters are telling you or by reading too much into peak metering. 

VU (rms) is not a perfect loudness indicator but it the best one around and far superior to peak readings. And overall it is very effective. I have mastered many CD's and used rms readings as an indicator regularly and they have turned out very even and consistent so yes VU and rms readings do work. (the only thing you have to do in mastering is change the overall rms reading for a track that might only have voice and one instrument) 

Our ears are much more rms trained and as we did in the analog days we worked with rms all the time. Most DAW's favour the peak indicators and approach but it is possible to incorporate both very effectively. It is just that most DAW's leave out all forms of effective rms or VU indicators but you can easily put them back in.

BTW  Sorry I was out all day and took so long to get back to read all this  

Digi Mike (gotta address the right Mike lol)

I think it's great you're getting into this stuff. Between bit and Jeff, I even picked up a few things here. As you may or may not know, I usually stay away from the science oriented parts of audio. Anything that has to do with intense understanding just isn't for me. Some of it is a downfall for me I'm sure, but I'm sort of like that country Dr. that can pretty much do anything due to the experience of doing it without the scientific end. My dr's office is dirty, my tools are primative, I can't tell you the history of the bleeding patient but I'll stop you from dying and save your @ss every time. LOL!

That's been my approach to meters, the K system, RMS etc. I'm in the Herb camp on this. I look at a meter to record....I like to keep things at -6dB going in...I look at my meters while mixing and keep them out of the red with my master bus showing an LED peak of -3 dB and I export when I like what I hear.

When I master something....it's funny...by the time I get done doing my entire procedure by hand manually, my numbers end up right where they need to be without trying. I can master a song, look at the number read-out (min/max samples, RMS min/max/average etc) and tell you what the threshold the limiter I'm about to use will end up at before I add it. LOL!! I don't even need to listen to it after (yet I always do hahaha) I can just put it on and I know that my RMS levels on a final will end up at -10 to -7 max with an average of -11 to - 14 dB at all times...which is my opinion, is beautiful for rock/metal and pop. 

But during a recording, I never think about this stuff for a second. As a matter of fact, I don't even think about it at the mastering stage. I master things until the sound good, but have a clue on what good and bad numbers will do to the sound. There are times when a client expects to hear a certain loudness...so I'll worry about stuff like that when asked, but most times I stay away from those types of jobs because I really want no parts of degrading music.

Anyway, the reason for my post here....like I said, it's great you're getting into this stuff and have the patience to learn. Just don't allow this stuff to dictate your decisions unless you see a series of numbers that set off an alarm that the audio may be degraded because of them. That's when you need to worry.

Also, keep in mind some programs may be inaccurate in how they read these numbers so make sure you have a few different sources to try. For example, I use Wave Lab 6 here....and the analysis is a little extreme where if I use other editors that do the same thing, the RMS numbers look way different. Wave Lab always seems to make things more extreme than they really are. It will show me a -4 RMS reading as a max value where Studio 1, Adobe, Sound Forge and others will show me -7 or -8 dB max RMS.

To me, Wave Lab would be the more trusted source here coming out of the gate....but then I learned that some of the read-outs as well as the meters are a little goofy in that version. So make sure you get a few different takes and try not to make too many decisions based on any type of number or graph read-out. The information is always an absolute plus to have under your belt, but nothing tops what your ears tell you...and I know you know that already. :)

-Danny

WOW I think Im getting a hold of this.

Thanks Danny and Dean

I messed with a song Im working on today and have really grasped the concept or thought of what I think my confusion/lack of knowledge is on exactly what Dave and Jeff explained.  Ive gotten better control of the crests, the initial transient attacks and the sustained part of the transients. Im noticing Im getting the perceived volumes up while still maintaining a good dynamic/attacks, separation and presence on each piece without squashing and making everything sound similar for a lack of words. 

Its an entire different/new learning curve compared to mixing live sound and getting it to where Im happy with what I hear on the monitors or cans.

Thanks again everyone!!!


I forgot to mention this is all part of my OCD issues of needing to understand what Im seeing along with what Im doing/hearing  I like to know that hen I hear something not quite right I can find and understand with my eyes as well