There does seem to be something is a little whacky with the fade rate at the end of a fade.

I did a quickie test with a 25-second fade on a pink noise signal starting at about -15dB RMS, and got the same result whether using clip fades or volume automation on the Master bus:

Using a "fast" curve, which should have a decreasing rate of attenuation at the end, I found that the signal was dropping 8-10dB/second in the last few seconds, and that the signal went from -50dB to -INF in the last 2.5 seconds (these are all RMS values).

Using a conservative value of -90dB for the noise floor of a typical analog output, this means there is at least as much attenuation (in dB) in the last 10% of the fade period as there is in the first 90%. 

Maybe I'm wrong to think that the rate of change in dB RMS is what should be "slow", "fast" or "linear", but given that what I measured agrees with the consensus impression that the rate is too high at the end, it seems a reasonable way to look at it.

I haven't looked at the other curves, but I would expect them to have an even higher rate at the end if they behave according to their shape.

This seems to be the case for fade-INs as well.
 
Steve

Sonar's faders do drop off a cliff of sorts. I don't remember the exact number, but at somewhere between about ~-75dB and -80dB they jump to -infinity (digital silence).

Note that this is a "the way Sonar works" issue and not anything particular to do with digital audio in general. Why they did it that way at that point, or whether it's a "problem" or a "beneficial feature" or whatever, is another question. 

Sonar's meters and peak displays don't display any signal below something like -96dBFS either (don't remember the exact number) - if you have a signal that peaks at -110dBFS, Sonar's meters and peak displays seem to indicate there is no signal at all even though there is. Of course in the real world, this is irrelevant, as it's extremely unlikely that things at those levels would ever be audible.


Now the loaded question: When you are hearing this on a fade, are you listening at a normal volume? Or are you cranking the volume way up during the fade? If it's the latter, you may be wasting your time worrying about something that is unlikely to be a problem under real world conditions (how often do normal people crank up the volume drastically during a fade?).

But if not (or you choose to worry about it anyway), you can probably find a plugin with a level control that doesn't have this limitation that you can automate instead.

somewhere between about ~-75dB and -80dB they jump to -infinity (digital silence).

I wouldn't say it "jumps", exactly. It's a slope, but it's steep enough below -75dB, that it might as well be a jump.


Also, I'm not seeing the limitation on what the peak indicators will show - at least not that high as you suggested. In the course of this testing, I saw values down to -140dB or less on the peak indicators, though obviously the normal meters won't show that, and because of the steep fade-in/out rate, you have to be quick on the peak reset to catch it unless you're looking at a signal that has a constant low amplitude as opposed to a fading one.


Also, I just checked, and the indicator values go from white to black at -96dB, probably just to tell you you're at the dynamic range limit of 16-bit audio.

Lynn is correct on this. Fading the actual session is a waste of time for various reasons. I put fades in after the final mastering stage. After mastering I open the track up in a program like Adobe Audition and clean up the start do some final checks and put the fades in at the end very last thing. 

The reason why this is good is because any noise that might be associated with the music gets faded as well and you end up with a track that goes to total silence. You have got the most control over the fade shape and sound. You can also do several fades over the fades. eg the first fade might be over say 20 seconds but you can do another one over say the last 5 seconds at the end. You can really control then how the music sounds as it is going away.

Fading is a mastering process not a mixing one. All you have to do with mixing is just make sure you print the music right out past the end of the fade so you have got all of it to work with. The reason why you should also fade after the mastering process is because if you do it before and then master, the various processes in mastering eg compression and limiting will screw up the fades that you have already done. Sound reasonable?

brundlefly

somewhere between about ~-75dB and -80dB they jump to -infinity (digital silence).

I wouldn't say it "jumps", exactly. It's a slope, but it's steep enough below -75dB, that it might as well be a jump.

No, it's not a slope. Click on the fader's displayed number and try to type in a value of -80dB in directly. You can't! Anything you try to enter below a certain number becomes -infinity.

Also, I'm not seeing the limitation on what the peak indicators will show - at least not that high as you suggested. In the course of this testing, I saw values down to -140dB or less on the peak indicators, though obviously the normal meters won't show that, and because of the steep fade-in/out rate, you have to be quick on the peak reset to catch it unless you're looking at a signal that has a constant low amplitude as opposed to a fading one.

I'm talking about the peak level indicators associated with the meters. Feed a track that peaks at say -40dBFS to a bus. Then reduce it by 70dB and reset the peak and you'll get no indication. 

Also, I just checked, and the indicator values go from white to black at -96dB, probably just to tell you you're at the dynamic range limit of 16-bit audio.

I'm thinking it may have just been "let's pick a number" kind of deal, and the 96dB seemed a reasonable choice.

On the grounds that it's practically inaudible

If you enter a # lower than -75 it goes to -INF.
 
If you then 2xClick it and click the + you get -143.0.
 
After cranking up all the output volumes to hear what's happening I found this:
 
 
-75 is still audible but barely.
 
-76 is dead silence.
 
 
We have a range of -76 thru -143 that's unuseable and unaudible by the user.
 
 
Steve
 

drewfx1

brundlefly

somewhere between about ~-75dB and -80dB they jump to -infinity (digital silence).

I wouldn't say it "jumps", exactly. It's a slope, but it's steep enough below -75dB, that it might as well be a jump.

No, it's not a slope. Click on the fader's displayed number and try to type in a value of -80dB in directly. You can't! Anything you try to enter below a certain number becomes -infinity.

Okay. You're right. I thought that only applied to fader positions, but it applies to automation as well. What I meant was that you get level indications in the peak meters below -75 if you reset them. But it turns out those indications are just a placebo because, as you say, the actual output signal is at digital zero below an indicated -75dB. This doesn't seem good. I'd be curious to know what other DAWs do.

Also, I'm not seeing the limitation on what the peak indicators will show - at least not that high as you suggested. In the course of this testing, I saw values down to -140dB or less on the peak indicators, though obviously the normal meters won't show that, and because of the steep fade-in/out rate, you have to be quick on the peak reset to catch it unless you're looking at a signal that has a constant low amplitude as opposed to a fading one. 

I'm talking about the peak level indicators associated with the meters. Feed a track that peaks at say -40dBFS to a bus. Then reduce it by 70dB and reset the peak and you'll get no indication.

Not sure which indicators you mean. I'm talking about the peak indicators at the top of track header above the FX bin. I guess if you use vertical meters, they would appear near the top of the meter.  On my DAW, if you play back a recorded signal that is actually below -75 on its own, not because of fader position, that peak indicator will show values below -75dB. I double-checked that, too.

Also, I just checked, and the indicator values go from white to black at -96dB, probably just to tell you you're at the dynamic range limit of 16-bit audio.

I'm thinking it may have just been "let's pick a number" kind of deal, and the 96dB seemed a reasonable choice.

I'm sticking with my contention that it's because that's the dynamic range limit of 16-bit audio - not an arbitrary choice.