[TheSonicTruth]

Well the issue with a truncated noise floor is that it modulates with the signal, so it depends entirely on what the signal is down there. If you've dithered, it sounds like hiss, but for other things it can sound like a weird robot kind of thing. You can always gain something down 90dB or so without dither then gain it back up to test. Or you can take a fade-out and gain it up. I can put up an example later.

I think I have heard what you're talking about. While embarking on one of my intentional destructions of a track to demonstrate remastering, with 6:1 compression applied at -50FS, I pulled up a particularly long fadeout and heard it fading into a tinny metallic oblivion. On others, I have heard the fade slowly replaced with a loud hissing. I guess the latter was the applied dither on that track off the CD.

 

[RRod]

I think I have heard what you're talking about. While embarking on one of my intentional destructions of a track to demonstrate remastering, with 6:1 compression applied at -50FS, I pulled up a particularly long fadeout and heard it fading into a tinny metallic oblivion. On others, I have heard the fade slowly replaced with a loud hissing. I guess the latter was the applied dither on that track off the CD.

Dither is easy to see on a spectrogram. If it's shaped, you'll see fuzz at the very bottom and top of the spectrum. If it's flat, you'll just see white-noise fuzz. Note that any real life noise that is loud enough can also dither the signal if it's independent of the music. 'Metallic oblivion' is a good way to describe truncation distortion.

 

[Steve999]

I've heard the weird robot kind of thing. I didn't understand what it was (though I think I do now. . .thanks), but I figured out how to avoid it. I was running digital recordings of LPs through audio software to clean them up, particularly the noise floor and some of the louder clicks and pops (you can just point and click them away!). And if I did certain things I wound up with the robotic swirling noise floor. At first I was miserable because I thought it was from my turntable, but the unedited recordings from my turntable didn't have it at all. I don't remember the details but it was pretty easy to figure out how to work around it. I think it had to do with what kind of processing I used to get rid of the noise floor--there were two or three choices.

I think I have heard what you're talking about. While embarking on one of my intentional destructions of a track to demonstrate remastering, with 6:1 compression applied at -50FS, I pulled up a particularly long fadeout and heard it fading into a tinny metallic oblivion. On others, I have heard the fade slowly replaced with a loud hissing. I guess the latter was the applied dither on that track off the CD.

Well the issue with a truncated noise floor is that it modulates with the signal, so it depends entirely on what the signal is down there. If you've dithered, it sounds like hiss, but for other things it can sound like a weird robot kind of thing. You can always gain something down 90dB or so without dither then gain it back up to test. Or you can take a fade-out and gain it up. I can put up an example later.

 

[TheSonicTruth]

All those corrections can or will involve changes in the dynamics (levels) and therefore dynamics processing is virtually always involved. Secondly is "achieving a mix that approximates the original release as millions of listeners remembered it 20, 40 years ago." How people remember it years ago is relative to how they heard all music years ago, how they hear it today is relative to how they hear music today, which generally is louder! How much louder a remaster should be, comes down to a judgement call, exactly what the original master is, it's genre, what it is trying to achieve musically and what is an appropriate amount of compression.



1. If we're talking about a "final 2-channel here" then a limiter cannot make "a back-up vocalist subjectively half as loud as the lead vocalist, almost as loud as that lead" and neither can a compressor! A compressor cannot do this for exactly the same reason as a limiter can't, because essentially they are exactly the same thing!

2. That's all true BUT it's all true of a compressor as well because they are the same thing. If you're going to say that a compressor can act on the vocals, that would be true but it would be equally of true of a limiter. A limiter is the same as a compressor, the only difference is that a limiter compresses to a limit whereas a compressor doesn't, unless it's ratio is set to infintity:1 with a very short attack time, in which case our compressor is now a limiter!

3. Yes it is entering into limiting territory but it depends on where the threshold is set and the input signal it is fed. I'm not sure you understood what I stated earlier about compressor settings? Let's use your example song again, ff the input signal goes up to say 0dBFS then with a 10:1 ratio and limit set to -40dB the output signal would be -36dB, it won't have actually limited to -40dB and therefore it's not actually a limiter! If however we set it much more sensibly, say -6dB then with a 10:1 ratio our output will indeed be -6dB (or a decimal point or so above), it is now a Limiter! And obviously (hopefully!), if we fed that same -6dB signal into a compressor (with a suitably short attack time and a 10:1 ratio) our compressor would also limit to -6dB and therefore is now a Limiter! A ratio of 10:1 under most usual conditions is enough to be a limiter but to be certain it's ALWAYS a limiter then the ratio has to be infinity:1.

G

"sor cannot do this for exactly the same
reason as a limiter can't, because essentially
they are exactly the same thing "

With all due respect, I will never believe they are exactly the same thing. While they serve the same purpose(constraining a range of signal) they accomplish that in different ways. It's like saying a Gulfstream and a Huey are "exactly the same thing": Sure, they both fly from point A to B, but they do so using entirely different propulsion and lift systems. And believe me, that Gulfstream and Huey chopper bear nothing in common, except for the space between them and the ground when they are both flying.

 

[TheSonicTruth]

Dither is easy to see on a spectrogram. If it's shaped, you'll see fuzz at the very bottom and top of the spectrum. If it's flat, you'll just see white-noise fuzz. Note that any real life noise that is loud enough can also dither the signal if it's independent of the music. 'Metallic oblivion' is a good way to describe truncation distortion.

Yes, I saw that demonstration in the Montgomery Hi-Res video.

 

[TheSonicTruth]

I've heard the weird robot kind of thing. I didn't understand what it was (though I think I do now. . .thanks), but I figured out how to avoid it. I was running digital recordings of LPs through audio software to clean them up, particularly the noise floor and some of the louder clicks and pops (you can just point and click them away!). And if I did certain things I wound up with the robotic swirling noise floor. At first I was miserable because I thought it was from my turntable, but the unedited recordings from my turntable didn't have it at all. I don't remember the details but it was pretty easy to figure out how to work around it. I think it had to do with what kind of processing I used to get rid of the noise floor--there were two or three choices.

If the click removal processing was an out of box plugin with a maximum bit depth LESS than that of your DAW, then you would be introducing some truncation at and around the removal points.

 

[gregorio]

With all due respect, I will never believe they are exactly the same thing.

I can't do anything about what you "will never believe", all I can do is state the facts, back them up where practical and then it's up to you to either believe what I'm stating, do some further research yourself or just ignore the facts and carry on believing whatever you like!

"A limiter is a compressor with a high ratio and, generally, a fast attack time. Compression with ratio of 10:1 or more is generally considered limiting." - Wikipedia
"The difference between a compressor and a limiter is only in the compression ratio used." - Hugh Robjohns, Technical Editor SOS. "Dynamics processors: what's the difference?"

G

 

[bigshot]

Most digital click removal takes the samples around the click being removed and creates pink noise to fill the gap. In music, if the clicks are short, it's imperceptible. The swirling noise floor sound comes from something entirely different. That is an artifact of broadband noise reduction, not impulse noise reduction. Usually it's from pattern matched NR where you take a sample of the lead in groove to establish the noise floor and use that to cancel out the noise through the rest of the side. It doesn't always work well because the sort of surface noise on the lead in groove can be quite different than the surface noise further in towards the center. When the pattern doesn't match, you get those weird phasey swirling sounds. There's also brute force high end attenuation NR applied dynamically with stronger NR in the quiet parts and weaker in the loud parts.

In order to run into the noise floor of 16 bit, you'd have to some pretty massive compression... I don't think that would be an issue most of the time... especially since most music is recorded and mixed 24 bit and just mastered to 16. You're much more likely to run into just general room tone- air conditioning, traffic outside, etc.

 

[Steve999]

Yes! That's exactly what happened with me. That's really interesting. I learned to avoid it by experimentation but I did not have a clue as to the reasons for it. Next time I know what to do and the reason why I should do it. : )

The swirling noise floor sound comes from something entirely different. That is an artifact of broadband noise reduction, not impulse noise reduction. Usually it's from pattern matched NR where you take a sample of the lead in groove to establish the noise floor and use that to cancel out the noise through the rest of the side. It doesn't always work well because the sort of surface noise on the lead in groove can be quite different than the surface noise further in towards the center. When the pattern doesn't match, you get those weird phasey swirling sounds.

 

[bigshot]

Pattern matched noise reduction works well. But it works best when you split all the tracks on an album side and use the lead in for each song rather than using the lead in from the first song for all.

 

[danadam]

The DR rating depends on the 2nd highest peak. If I lop-off 6dB from that peak, and the RMS stays (roughly) the same, then the DR will go down by 6.

How so? Won't another peak become 2nd highest? So the DR will go down only by the difference between 2nd and 3rd highest peak of the original waveform?

 

[RRod]

How so? Won't another peak become 2nd highest? So the DR will go down only by the difference between 2nd and 3rd highest peak of the original waveform?

I should more properly have said 'if the new 2nd highest peak ends up 6dB below the previous one'. I was considering my example, where all the original 0dB peaks were limited to -6dB.

 

[danadam]

I should more properly have said 'if the new 2nd highest peak ends up 6dB below the previous one'. I was considering my example, where all the original 0dB peaks were limited to -6dB.

Ah, I understand.

Out of curiosity I tried that too, and yes, with -6dB it's hard to tell the difference. I then tried with -12dB, to see what I should expect:

sox *.flac out.wav compand 0,0 -12,-12,0,-12

With "Mosaic" from "Guitar Noir" (from AIX Records) it reduced DR from 13 to 5 and it sounded the same, only in the parts where it had to squash the most (e.g. like in the last minute) you can here some additional distortion:

 

[tansand]

Should probably do some abx testing before we conclude it sounds the same...right?

 

[tansand]

I was just listening to Nevermind, I've had the 2011 remaster for a while, which I learned had a bad reputation. so I got the original mastering. Cobain's vocals on smells like teen spirit sound like a howling roaring buzzsaw. The 2011 sounds like his (dead) balls are in a jar somewhere in comparison.