Most people will naturally tend to adjust their volume control such that a loud but comfortable level is 77 dBSPL. We just seem to gravitate toward that level, as described by Mastering Engineer Bob Katz.
And in my own lab, I have standardized on the EBU R128 levels, where 0 dBVU = -23 LUFS. And I run an EBU R128 Loudness meter on my DAW to keep an eye on this.
0 dBVU refers conventionally to the average recording level (VU) over the course of a track or album, in self-relative units. However you choose to measure that overall average level, in whatever units, we refer to it as the 0 dBVU level. It’s just a shorthand way of stating “average recording level over the track or album”.
-23 LUFS refers to where we would like that same average level in loudness units (LU) relative to full scale (FS), where the scale has 0 dBFS at the maximum, full-scale, clipping level. This specifies how much headroom you want to leave available for peak excursions above that average level.
(…are you confused yet? The problem for communication here, is that dB measure is always relative to something. And so you need to indicate what that something is. Just like when you talk about percentages… percent of what? That’s the purpose of the XX in dBXX.)
And so this is stating that your overall average recording level on the track or album should register 23 dB below full scale.
Now, these initial paragraphs have been describing two completely different things. The first paragraph, stating that 77 dBSPL is a good listening level, is talking about how loudly things should sound in your room or headphones. The remaining paragraphs were describing the recording levels encoded in the audio stream on the CD or audio track.
It is the job of your output audio amplifier and speaker system to convert the internal recording levels to audible listening levels for your ears. Listening level describes pressure waves in the room air, while recording level describes bits on a computer medium. The amplifier converts bits to electrical waves which feed the loudspeaker, which then produces the pressure waves.
The implication here is that the average playback level, at 0 dBVU, should correspond to a loud but comfortable listening level. In your own listening, however, you can always choose to listen at reduced levels from these. Just back off the input volume control till you are comfortable. Nobody can demand that you listen at these loud levels. We just use these level standards in order to have a consistent calibration, so that when you say you use X LUFS, anyone else will understand exactly what you mean.
I think EBU R128 is a good standard to adopt (meaning, 0 dBVU placed at -23 LUFS, versus 0 dBVU placed at some other LUFS), because it leaves Crescendo with a decent amount of headroom for its processing. And this level also accommodates orchestral recordings, where crest factors can be as high as 20 dB, compared to more squashed rock music sounds.
So I set my system such that a 1 kHz -20 dBTP peak sinewave produces 77 dBSPL by adjusting the volume level of my output audio amplifier. This joins the two concepts, listening level and recording level, together.
to make things all the more confusing for us Physicists, the EBU R128 standard specifies that a 1 kHz -23 dBFS peak sinewave signal shall measure -23 LUFS, even though its RMS average level is really -26 dBFS. But I guess this convention makes things easier for everyone else.
Sorry, my mistake.. that was totally incorrect. The EBU R128 spec will produce RMS-like levels around 1 kHz on calibration tones. So a -20 dBTP Sinewave at 1 kHz will produce -23 dBLU in any mono channel that it plays through, when properly calibrated. That 3 dB error on my part came from monitoring the loudness meters with a stereo signal. The left and right channels are added together to produce the loudness estimate. And for a sinewave signal that sum will produce a 3 dB louder reading than in a mono channel. ]
[… and the EBU R128 spec really calls for calibrating against a -15 dBTP 1 kHz sinewave signal, 5 dB louder than the already loud but comfortable level. So that should measure 82 dBSPL on the sound level meter. Presumably, this is intended to minimize the effects of extraneous environmental noises on your calibrating measurements.]
Sources of Audio for Playback
When a track begins to play back from some outside source, like a Web page, or a track in iTunes, I just naturally back off their output levels if they are too loud. And they are almost always too loud.
Even Apple’s SoundCheck levels seem too loud, by anywhere from 6-8 dB, compared to the EBU R128 standard. But that’s okay. That’s why they offer volume sliders.
And so rather than use SoundCheck on iTunes playback, I have a bunch of flat EQ’s set up in iTunes with varying amounts of output attenuation. I’ll set up each track in an album to play back through one of those Equalizer settings for the whole album.
I find the right one to use with an EBU R128 album scanner of my own making. It scans each track and reports the overall album LUFS level. I use as much attenuation as necessary to get that level down to -23 LUFS +/- 1 dB. But you could also find out the right level to use by watching your Loudness meters.
Crescendo Processing and Peak Excursions
Crescendo will produce an expansion of your peak range. If you have followed the EBU R128 recommendations, and your hearing impairment is not too severe, you may get along just fine with 0 dB Headroom. There may be plenty of headroom in the -23 LUFS setting already.
In my case, with my rather severe levels of corrections needed, I give Crescendo an additional 6 dB of Headroom in its control panel. That attenuates my sound levels by 6 dB of course, so I need to make up for it in the output amplifier. And that’s okay, because it avoids clipping distortion and gives me maximum clear dynamic range.
Today’s systems are almost all 24-bit capable. But you are highly unlikely to ever fully realize 24 bits of precision in any output audio chain. The bottom bit, in a 24-bit EBU R128 system calibrated at 77 dBSPL, would represent sound levels of -37 dBSPL (!!). That’s 37 dB below the absolute threshold of unimpaired hearing. Even SuperMan would be unlikely to be able to hear that.
However, 24 bit-width gives you plenty of bit-space to accommodate something like a Crescendo that only wants another bit, or even 2 bits in extreme cases with 12 dB headroom settings. (1 bit = 6 dB)
Crescendo has a dynamic correction range of 70 dB, meaning that it will make corrections to music channels that have anywhere from 30 to 100 dBHL. So you won’t need anything more than 12 bits of precision in those channels. But your other uncorrected channels down in the bass region might allow you to hear even more.
I rather doubt that you’ll need more though. Once you reach down to the 30 dBSPL level, you are listening to air conditioner noise in the recording hall.
You should always first calibrate your audio system as described at the beginning of this post. And when you do that you should ensure that Crescendo remains running, but that Process and HdphX are not engaged. That leaves the Headroom parameter in Crescendo operational, but otherwise disables all sonic modification from both of these modifiers.
Depending on how severe your hearing impairment, you might need some Headroom to avoid clipping during Crescendo processing. When you specify Headroom in the setup parameters, Crescendo diminishes the output levels going into your amplifier by that much. And you need to recover that diminution by increasing your amplifier output gain a bit more. That will happen during your audio system calibration.
Note that disengaging Crescendo Process is NOT the same thing as disabling the Crescendo plugin. When you disable the plugin, it no longer has any control over the audio processing, and you will experience an immediate increase in volume by the amount of Headroom you had specified in the Crescendo setup parameters. You don’t want to bypass Crescendo. Leave it running, and just make sure the checkboxes for HdphX and Process are unchecked.
Once your audio system has been calibrated, we come back and calibrate your Crescendo. Go ahead and re-engage with the Process and HdphX. Crescendo also needs to know which recording level corresponds to what listening level. All it ever sees are recording levels, measured in bits. But since it deals with shaping sounds for your hearing, it needs to know what you will be hearing at each recording level.
That’s the purpose of the Cal dBFS and Cal dBSPL controls tucked away in the setup parameters. Cal dBFS is where you tell it what your calibration recording level was, and Cal dBSPL is where you tell it what that calibration recording level produced in the audible listening level.
So, if you follow the EBU R128 standard calibration, you would set Cal dBFS = -20 dB, and Cal dBSPL = 77 dB.
If you need some Headroom, depending on your audio hardware, it is possible that you might not be able to reach 77 dBSPL for the calibration level. In that case, just enter in Cal dBSPL whatever level you did achieve. It just means you will be doing your ears a favor by listening at softer levels. Crescendo will also know this, and will make different hearing corrections at these softer listening levels than it would have made at the nominal 77 dBSPL level.
This is important to understand: Crescendo must be kept fully informed of any and all volume adjustments, once your system has been calibrated. That can only happen if you adjust the playback level ahead of the Crescendo processing. It can figure out for itself that you have changed the volume setting.
But if you change the audio amplifier gain, or your post-Crescendo output level control, anytime after calibration of your audio system, Crescendo can’t see these changes, and it will never know that you have done this. And all of its hearing corrections will be incorrect at the new loudness levels.
[Whew! Much ado… dB,dBVU, dBFS, dBLU, dBLUFS, dBSPL, dBHL, dBthis, dBthat…. eiyiyiyi… makes you go cross-eyed, doesn’ it? But once you’ve calibrated your system, you will achieve audio nirvana.]