Do you know how to read your meters? -18dBfs theory?

[Citizen]

http://therecordingrevolution.com/2013/ ... ur-meters/

I just read this article, and wanted to get some perspectives on it - as I've heard contrary information in the past.

The article suggests that you should be aiming to get your levels averaging at -18dBfs to maintain optimal sound quality in your mix. This, apparently, is the 'digital sweet spot'

(I had no idea such a thing existed)

I've got a few questions about this - chiefly, if everything is sitting at that approximate level, how are you supposed to adjust the balance between your different tracks without deviating from that level? (Have I missed something here?)

Would really appreciate some feedback here, as it's an area that I could do to improve my knowledge on.

[re:dream]

Hmmmm... I am no expert but this seems very specious to me.

I make sure that I mix quite softly - more to save my ears than anything else, and I don't drive my audio close to clipping. But I have never heard of any reason to specifically average at -18 dB FS & the reasoning in the article seems odd. I don't know whether you can even argue by analogy from analogue in this way...

Let's see what the experts say...

[re:dream]

It's been posted many times before, but this http://tarekith.com/assets/mixdowns.html is good advice.

[Angstrom]

Because Live uses floating point math to internally handle the audio streams they won't clip at digital zero, so all this talk of digital zero on each channel is misleading. I want you to try an experiment to show yourself something

Put a sinewave, or an Operator playing a low sine wave onto a track [Track One], and route that output to a second track [Track Two] set Track Two monitor to "on".
On the second track place a Spectrum device, set its range to 0 to -180, and look at the trace of the sinewave, there is a big peak at your fundamental and a bit of very quiet noise up in the high frequencies.
Now, on the originating track, put a Utility on there and crank it up to +35db. The meters go RED!

On the second track, put a Utility before your Spectrum and set it to -35. Look at the trace, it's the same as before. It was +35db (over "digital zero") and has now been brought back down without harm. The miracle of floating point math!




In addition you can see how individual tracks are able to save the day all on their own. Try turning off the Utility on Track Two (which is set to -35db). The meter leaps back into the red, and your master output meter is red too! Now move the fader of Track Two down until the master fader goes green. Copy your Spectrum device and place it onto the Master channel, take a look at it, you will see that there is no distortion.


The short version is: Don't slavishly fear the idea of "Digital Zero" on each channel because it wont really help you understand the issues. Instead seek to understand what is going on : there's lots of headroom in floatingpoint right until you hit the output stage at THAT point it becomes a fixed point file, at THAT point you should make sure everything is below zero.

[stevemac]

A lot of the -18 theory is based on the vintage plugin emulations I think. A lot of them also model the way they react to input levels. Obviously a vintage piece of gear will react much different to a really hot signal vs a low one.

So I think this is mostly relating to that.

[clydesdale]

I think the article is implying that the visual feedback you get from different DAW software is different and it's based on the scale of the meter. If you look at his examples you can see that each meter has a different scale-- one is linear, another piecewise linear, and still another looks pseudo logarithmic. It just means that you need to pay attention to the meter value (objective) and not just the meter's relative height, movement or lighting cues (subjective).

I don't want to commit any heresy but based on Angstrom's proven point it seems to me that as long as you can guarantee that you're not clipping in the integer realm (i.e. losing waveform information) that you really want your master signal and therefore your rendered file to have as a high a peak as possible-- always. The reason being that upon re-importation of the file, into Live for instance, if you were to place a utility at -6dB you would be accomplishing the exact same thing as having rendered the file with master at peak -6dB in the first place. You wouldn't need to leave any headroom at all since you can always make headroom. I'm sure this has been discussed to death on the internets so I'm just thinking out loud here, feel free to ignore this rather than needlessly chastise me.

[Angstrom]

There's nothing really wrong with following this guy's advice, but I think sometimes people take the wrong message from these sort of articles. These articles need to be quite careful when talking about things like a "digital sweet spot" because it sounds almost mystical. Like there are some nice-sounding floating point exponents and some bad-sounding ones (or some nonsense like that).

Of course it's good practice to keep your meters from peaking, and to make sure you aren't overloading any plugins in the signal chain - but talking about digital sweet spots in the context of effects without exploring it in detail can give rise to some odd misconceptions.

Lets look at the phrase

So the point is that the way converters work, -18dBfs will be pretty darn close to the analog equivalent of 0dBVU. And since many DAWs and plugins are built to emulate analog gear, that sweet spot is still smart to shoot for to get the best sound possible.

So if our meters read an average of -18dB, then it will sound different? Better?
really?

Try something : When recording
In Live, when you record audio into a track (in from your converters) ... What does moving the fader up and down do? Not much. It's not actually affecting the file being written at all. The meters to pay attention at this stage are those on your converter, not on your DAW. Its quite possible to overload your converter meter, and have a DAW track meter reading -18dB . An absurd example, but very common, and it points out the flaw in the article : don't focus on a methodology, focus on understanding the issues.

Try something : When Mixing
There is talk of plugins, which model analog equipment. The implication being that by having the meter reading an average of -18 the plugin will work best. I invite you to try something : make Operator play a sine-wave sound in a clip in session. Slap a saturator on there (a volume dependant effect), and turn the Saturator->drive right up to make it sound distorted. Now ... Turn down that track's fader so the meter reads -18dB

Did the saturation change?
Nope.

Why? Because the track fader and the meter is post-effects chain, the track effects are inserts. The fader and the meter reading have no relation to the waveshaping effects of the insert effects.



conclusion
So where does that put the information in the article? That guy told us if we turn our fader down so the meter reads -18 it would make special volume dependent effects (like the waveshaping saturator) sound different, "sweeter". But we just demonstrated that is an erroneous conclusion, because insert effects are pre-fader.

Hmmm, eh?

[Citizen]

Hmmm....lots of food for thought. I'm glad I posted, as it didn't sound quite right to me.

That said, there isn't even a single person in the comments section of that article debating against his point. Seems to be an area where there is a lot of misconception.

Is Live the exception, purely because of its floating point math, or do these principles hold true for all DAW?

[Angstrom]

Most DAWs use floating point and some also allow the saving and importing of the files as floating point, meaning you could render out a wave all clipped to fuck and still pull the gain back into clean audio. Magical !

Example:
https://www.youtube.com/watch?v=Qt-EJhDDHUI

in this article DAWs and floating point is discussed. I only skimmed it but I'm sure it's more informed than I am about these things
http://www.soundonsound.com/sos/jan08/a ... 0108_3.htm

[jlgrimes]

I've heard about -18 DBfs being the 0 equivalent on an analog mixer but I wouldn't get to anal about it. The main thing with digital is not to clip but also with 24 bit digital unlike 16 bit, there is no need to record anywhere near zero.

Usually anywhere between -20 and -10 would be a reasonable level. If you go to -9 I wouldn't worry but just remember to lower it a little next time.

The main thing is keeping ample headroom when it is time to mix.

[lunabass]

If you’re using analogue hardware such as EQ, Mixers, Compressors etc with your DAW then -18dbFS is excellent advice, if you aren’t, then I believe it to be a non-issue. As Angstrom said, there is nothing wrong with following this guys advice, it’s actually OK, but people rarely take the time to understand or test it for themselves.

Take analogue emulation plug-ins for example. Do you really think Waves will emulate their plug-ins so as to distort at 0dBFS just like a piece of Analogue Hardware would with that much level? Try running that much level through a hardware Neve EQ and tell me how it sounds.

Run a simple test.
1. Take a loud drum loop with peaks close to 0dBFS.
2. Place an audio effect rack onto the track.
3. Place an analogue emulating EQ into it and make a few cuts or boosts. Make it noticeable and ensure if boosting that it doesn’t clip. Name this chain “dBFS” or something that indicates it’s high level.
4. Duplicate this chain (and therefore all of the EQ settings) and name it “VU”.
5. On the VU chain, place a Utility set to -18dB before the EQ plug-in. This will ensure it receives 0VU “the sweet spot”
6. Place another Utility after the EQ set to +18. This enables you to do an A-B comparison between the 2 chains without any difference in level.

Now just solo each one and listen. Do you hear a difference? If VU sounds “better” to you then it’s obvious what you need to do.

For what it’s worth, I’ve done this test (although I did them at -12dBFS as the studios I’ve worked in, 0VU was calibrated to -12dBFS) on a bunch of my plug-ins without hearing any difference BUT that doesn’t mean you won’t get a different answer with your plug-ins, DAW and ears.

[Stromkraft]

In addition you can see how individual tracks are able to save the day all on their own. Try turning off the Utility on Track Two (which is set to -35db). The meter leaps back into the red, and your master output meter is red too! Now move the fader of Track Two down until the master fader goes green. Copy your Spectrum device and place it onto the Master channel, take a look at it, you will see that there is no distortion.


The short version is: Don't slavishly fear the idea of "Digital Zero" on each channel because it wont really help you understand the issues. Instead seek to understand what is going on : there's lots of headroom in floatingpoint right until you hit the output stage at THAT point it becomes a fixed point file, at THAT point you should make sure everything is below zero.

Excuse my late response, here. I contemplated quoting and starting a new thread, but all your source data is here so.

I think you gave examples that serves your argument well here and it's a good reminder to consider floating point as well as the fact that faders doesn't have much to do with the end result really.

However, in the spirit of your suggestion in bold in your quote that I certainly agree with I set up a little experiment seeking to investigate if feeding Saturator with a different signal compensated later in the chain would render a different result. After all, that's actually what the article is suggesting it would. It's not focused on clipping, if I recall correctly. This is what I did in Live 9.2b4 and feel free to repeat this or a similar experiment with Delay Compensation active:

• 1. One track with Operator default sine playing in my case a pattern of E1, G#1, F1 at a suitable master volume.
  2. Add an effects chain with Utility, Saturator set to driven quite hard but inactive at this stage, Utility and another Utility.
  3. Duplicate the effects chain and rename to "Reversed". At this point you should still hear audio.
  4. set the last utility to Reverse polarity. At this point you should hear only silence as the chains are cancelling out each other
  5. In the Reversed chain lower the first Utility to -8dB and raise the second to 8dB. The chains should still cancel out so no audio should be heard.
  6. Now activate Saturator on both chains. You should now have something like these chains (ignore the Limiter for now):

Original Chain


Reversed Chain


The Results Are In!
Now, does the audio still cancel out with these identical Saturator settings for you? Because it sure doesn't for me.

• 7. As a final confirming test turn off the first and second Utility of the Reversed chain. There should be only silence again.

What can be concluded here?
We can observe that

• 2 Saturator plugins fed the same signal and one being polarity reversed cancel out
• If these are fed different input levels they don't cancel out

It would seem that feeding at least some audio effects with different input levels renders different results. It's possible here that something is wrong on execution or set up because I'm surprised of the strength of the signal left at step 6 above. But this experiment seems coherent and logical at least at this early hour.

Do you have any suggestions for a better test? What are your thoughts on this?

[lunabass]

Your results make perfect sense for Saturator.
The saturation effect is enhanced by an increasing/decreasing input signal. That is what the drive knob is for, think of it as a Volume knob. Changing the input using a Utility beforehand is going to give you the same effect as changing the drive knob. Basically, both chains effectively have different drive amounts hence the difference in sound.

Try a simple test to prove it:
Take a sine wave close to 0dBFS.
Make 2 chains with each containing a Utility and Saturator.
Chain 1, turn the drive knob on Saturator up to 25dB
Chain 2, turn the Utility volume up to 25dB instead of using the Saturator Drive knob.

Do you get a similar typr of saturation sound from both chains?

[Angstrom]

Yes, Saturation alters the waveshape of the sound based on "drive" (input gain). If you add or remove 8 db into a waveshaper it will alter the output wave utterly. That's how a saturator / waveshaper is supposed to function. More input gain = different output due to increased wave shaping.

There is no way two saturators with different drives should ever phase cancel. That would be absurd.

[Stromkraft]

There is no way two saturators with different drives should ever phase cancel. That would be absurd.

Exactly. That there are devices that react dynamically was the main point I made. If even one device of the audio effects chain reacts dynamically to the signal level, then you cannot assume that a higher volume is as good for the sound you want as lower volume and vice versa. It's the result you want that should guide you. This means you need to acknowledge the actual devices.

When this subject comes up the facts of floating point is sometimes presented in a way that can be understood as if that's all there is to it (not saying that is the intention on your part). I just want to point out the fact that this isn't all there is to it.

When you have wrecked the sound with dynamic devices in your signal chain— which is probably very subjective what that is — there is seldom a way to magically turn that back, as may be the case with floating point and clipping.

For his reason using some RMS level and peaks well below 0dBFS makes sense to me simply because usually you can add compression and makeup volume in later stages without bad side-effects (certain current releases to the contrary). It also makes room for fader rides upwards and increasing transients to body ratio.