11 replies to this topic

[JustDan]

You could use "Normalize" to ensure it is at its maximum [edit: volume]. However, normalization does not leave headroom, which you should always have in a master (see the book "Mastering Audio" by Bob Katz to learn all the reasons why). The number is arbitrary. The Pleasurize Music Foundation suggests -0.3 dB, so that's what I use.

A trick I use is to include a gain plug-in and, then, an adaptive limiter plug-in after it on the master track. Then, I bounce to the audio file and use a program like the offline TT Dynamic Range Meter to calculate the root-mean-square (RMS) volume of the audio. The RMS value is like the average volume, only it tends to be more accurate because the equation tends to discard a lot of the wilder fluctuations in values (it's like minimizing the contribution of stuff of the left and right sides of a bell curve). Set the adaptive limiter out ceiling to -0.3 dB (or wherever you want it to be, so long as you leave at least that). The adaptive limiter will kick-in to take care of any naughty transients that may cause clipping, but do it in a way that is more transparent (and CPU intensive) than a normal limiter.

Bounce the piece and use the TT Dynamic Range Meter to measure the RMS (you can write your own software to do this too). Decide what target RMS you want the piece to be based on the dynamic range you generally need. Car stereos will need less dynamic range, theaters will need more. I've found by analyzing masters that -11 dB is a very common RMS value and seems to work well (I've noticed anything louder than that seems to result in distortion at times). Take the RMS value of the left and right channel and average them together (right + left) / 2. Then, subtract the dB level you want it to be. Go to the gain plug-in and then add or subtract that value from the gain to make it result in that RMS. Check the file again with the TT Dynamic Range Meter and if it is close, you've leveled the audio to where you want. Keep in mind that certain frequencies sound louder than others to the human ear, so this will get you close, but for certain pieces, you may wish to then drop the master level a dB or two to keep it sounding like the rest of the album, while still retaining the same amount of general dynamic range.

Here's an example: You've bounced an audio track and you want to level it. Using the TT Dynamic Range Meter, you find the left dB to be -13.1 and the right dB to be -12.8. Type into a calculator 13.1 + 12.8 = and then / 2. The average RMS of both channels is 12.95 (no minus sign because it is just faster not to deal with it here). Then enter 12.95 - 11 = and you get 1.95, which means you are 1.95 dB away from the RMS you want. Since -12.95 is quieter than you want, you know you have to add 1.95 to the value in your gain plug-in.

[JustDan]

Just an addendum: there is, no doubt, software out there that adjusts the value based on the human ear's ability to perceive the loudness of the frequencies. The graph of this is called the Fletcher-Munson curve and adjusting the value is called A-weighting. However, it isn't perfect and you might find it easier to just get the RMS and then listen to it in relation to your other pieces and then adjust it accordingly.

[Peter]

Normalizing will only bring the loudest PEAK up to what ever you set the threshold to.
I wouldn't suggest normalizing anything. The word sends shivers down my spine.

Instead the second half of Dans post is where it's at. If you want to achieve "loudness" then focusing on RMS is the way to go.
Aside from slamming into a limiter, good compression on individual tracks and then 2bus will bring up RMS. then the remaining loudness can come from limiting.. The -0.3 comes from the possibility of intersample peaks that can cause distortion depending on the how the digital to analog conversion is done on playback. -0.3 gives a "safety" of sorts.

But if this is even the case here.. I'm not sure.. The problem might be something else all together.
Dans gonna have You hittin the books for sure either way.

[JustDan]

Oof, you're right: I shouldn't have said "loudness" there and normalization is definitely not for achieving loudness. Nobody fights the loudness war with normalization! The only use for normalization in a bounce should be to prepare a wave file for importing into a project that, in the end, would not use normalization. For those cases where there is just not enough CPU to mix everything in separate tracks. I got that completely wrong - my head was all about headroom there and wasn't, apparently, I didn't leave my head enough room to think properly.

[JustDan]

I've been avoiding the compressor because I want the maximum dynamic range except for those transients that slam against the wall and, then, some smart handling of those pesky transients. My goal is to ensure maximum dynamic range in the piece with a RMS of -11 dB. If I use a threshold to attenuate the waves above a certain point, don't I wind-up altering the sound a lot and losing dynamic range? You've got me thinking - there is a limiter in the compressor, after all, which could serve that purpose of keeping the headroom of -0.3 dB and, perhaps fine-tuning the ratio, threshold, attack, release, etc. would wind-up in a better sound at the end? Now I have to experiment. The books provide concepts, but generally not much procedure. Such an art! I love it.

[Peter]

I've been avoiding the compressor because I want the maximum dynamic range except for those transients that slam against the wall and, then, some smart handling of those pesky transients.

If you want that then you shouldn't be concerned with "loudness" at all. Very little compression and possibly no limiting at all except for protection on those random peaks. I guess that's the extreme. Lol but your not gonna get loudness from that.

There is no way to get tasteful loudness with out an appropriate amount compression. Limiting in the end should only be shaving off a few dbs. Anything more and your then distorting and destroying dynamics for sure.

I've been "forced" over the years to smash the shit out of stuff and I'm actually kind of sick of it. Smash I mean like -8RMS. As with most of the heavy stuff I do.

Anyhow it's always an evolving process for the mix/mastering guy. Your ears and techniques are always developing.

[JustDan]

Understood. I do occasionally use compression, but as a very last resort and only if I cannot get the instrument to fit right in the mix. If I can create a bounce and balance the levels so that it fits well with an average of -11 dB and rarely go over, and all of the instruments are distinct, and the melody sits in front, I'm happy.

[JustDan]

I suppose it is worth noting that limiting actually is compression with an infinite compression ratio so, in a sense, by using the adaptive limiter on the master bus, I'm still compressing, but the intent is different: I'm looking to control the exception to the rule and keep as much of the range as possible during most of the audio, where compression intends to make it sound louder altogether by reducing the range and bringing the quieter parts of the instrument more to the foreground.

[Peter]

Merry Xmas ! Happy new year soon too.

Compression works differnt than limiting. As you went on in your post you contradicted yourself.

Using a limiter does not work the same as using compression. It's not compression, it's limiting.
Compressing is not limiting and limiting is not compression.
I'm not sure if your trying to convince me or are just figuring this stuff out out loud.

If the latter, I'm very clear on how all this stuff works.

[JustDan]

A compressor is a limiter if it has a ratio of infinity:1, but I wasn't saying that a compressor is always a limiter, which is why I included how they are used. It's a math thing. See this page: http://www.doctorpro...essors_en.shtml

[JustDan]

No matter what, they are used for entirely different purposes: limiting to control peaks so that they do not go over a certain threshold, compressing to boost the lower peaks and attenuate the louder peaks to bring the sound out. I was saying that, if you were to take things very literally, limiting is a form of compression. I guess a part of me is always aware that an Internet forum is a place where a detail gets attacked, so I tend to try to fill-in the gaps to avoid the inevitable and inescapable truths of talking on the Internet. It's also one of the reasons I haven't spoken much on the Internet since the 90s: I tend to focus on the forest, not so much on the trees, and a lot of threads can wind up lost in the forest due to a few trees.

[JustDan]

Well, I said "boost the lower peaks [under a threshold]" there, but that isn't right either: it attenuates the louder peaks over a threshold but does not change anything below the threshold. Just trying to achieve perfection in my posts and make sure I address that one before someone else does. A limiter infinitely compresses above a certain threshold.