Applying Gain to 16-bit Recordings Options

[Engelsstaub]

I have a few questions regarding some 16/44.1 recordings I have that could be a bit louder. I wanted to boost the gain to somewhere around the point where peaks will be around -1dB without clipping. (It seems that typically the peaks are around -3dB and the meat of the recording is around 6dB and often lower.)

The files will first be upsampled to 32-bit(float)/44.1 and edited in iZotope RX2 Advanced. (Editing consists of light declicking, some possible noise removal between tracks, etc. and then resampling/dithering back.) My questions are these:

1. Is there any reason why applying gain would effect the quality of the original recording in the resulting product? (Would it have just been better to get the level set closer in the first place ?) I know that reducing the volume is "losing bit depth," as I've read this from members in older threads while searching earlier. Is there a reduction in bit depth (or perceivable sound quality) when increasing the volume when using a quality program at the upsampled bit rate?

2. Should the gain be applied before or after the editing (or does it even matter) assuming that number one isn't a boneheaded way to approach this with respect to quality?

Basically I believe it would be ideal to make this...



...look about like this. (About 3dB louder.)



(Sorry if these questions are a bit noobish. I've been reading my ass off through the archives and still can't quite get a grasp.)

[Engelsstaub]

I appreciate that honesty in wording and efforts to maintain honest wording and precision. I'm often sloppy in my wording (as I've demonstrated in this thread alone!) but I don't mind correction. I just remind myself that this forum must strive to maintain certain standards that seem to be regrettably absent elsewhere in cyberspace.

...but anyway: I was under the impression that some of my previous vinyl rips were in need of being re-recorded at closer to peak levels to obtain "better" or "optimal" quality. If I (or a younger person with better ears) can not discern a loss in quality then it should be satisfactory and "just as good" for my purposes.

[Porcus]

I was under the impression that some of my previous vinyl rips were in need of being re-recorded at closer to peak levels to obtain "better" or "optimal" quality. If I (or a younger person with better ears) can not discern a loss in quality then it should be satisfactory and "just as good" for my purposes.

You report peaks at -3 dB. Which is about half a bit wasted, and that half a bit is probably filled with noise anyway. http://en.wikipedia.org/wiki/Dynamic_range#cite_ref-18 suggests you won't get more than 70 dB dynamics into (and thus, not out of) an LP groove, and that's about 12 bits. So if they sound like you need to re-record them, then this isn't the reason why.

Now if it wasn't for the noise that makes the following a bit superfluous: The 'right thing to do', would be to record into somewhat more than 16 bits. Like, 24: eight more bits means that you have plus/minus four bits' worth of margin of error if you think you record at an album peak of about -24 dB. Then you do all the processing (crackle/pop removal and whatnot) in the digital domain, and then afterwards scan for peak and apply peak normalization to 0 dB. Probably there is nothing in the lowest 4-ish except noise introduced by your processing. Dither it down to 16 bits, et voilà.

You could very well be within 1 bit of that, and that's not much if that one and three more are basically noise.


Anyway, a ReplayGain scanner like foobar2000, writes ReplayGain tags only (no re-encoding). If you use e.g. dBpoweramp Music Converter, then there is a 'ReplayGain' which scans and tags, and a 'ReplayGain Apply' which applies the volume gain to the file (re-encoding your FLACs ... are you using lossless). If you are unsure how your application of choice does this, then just make a copy of a sample file, scan ReplayGain, and click OK if it asks you to commit the change; then use e.g. foobar2000 with foo_bitcompare to compare the audio of the original and the processed file. If whatever you used is merely a tagger, they will come out as bit-identical.

When fb2k has tagged a file (well, scan an album as an album, not individual tracks), you can view the RG values. What is Album Peak in your case?



(... that fraction-of-1 is digital amplitude, right? So 0.71 is about -3dB? Anyone?)