Thanks, I like your fix better than mine. This is what I came up with using the Pattern tab in Spectral Repair on the left channel. This clip was copied from the master file after the repair. Overall there were 7 spots like this on side 1 and 10 on side 2. This was the worst. BOTT Stitching Repaired
I was working on The Cars Candy-o today. Here are a couple of screenshots that clearly point how and why you need to be careful when running declicking software and with RX don't process the whole file. the screenshots pretty much say it all. [/url][/IMG] closest citibank branch
First, Zoom in, use Interpolate on those 7 or 8 vertical transients, then a light pass of SR (attenuate unwanted) on the left channel. Cleaned up pretty well.
I think this has been answered before, but I couldn't find it. Are there potential problems with converting WAV files to MP3 in regards to introducing errors depending on what the peak levels were? Normalizing to -.3 in WAV might present issues when converting to MP3? Am I nuts?
Indeed it will. It's been awhile since I looked into why this happens, but the encoding/decoding process for MP3s and AAC files results in peaks exceeding 0dBFS unless you normalize to at least -0.7. I usually normalize audio I intend to convert to to -1dBFS just to be safe.
For further processing, conversions, audio chain tweaks, -3 dB headroom is safe. A well recorded needledrop still sounds powerful at LUFS -20 dB (true peaks between -4 to -5 dB). If you want to listen music not distorted noise. Here is a mp3 sample with L/R true peaks at -7.71dB/-6.80dB and -17.9 LUFS. Dropbox - sample.mp3 - Simplify your life
Well for further processing, I'd advise keeping things in 32-bit floating point until all tweaking is done as you don't have to worry about levels as much. You could normalize in 32-bit FP to +20 if you wanted, as long as you get everything below 0dBFS before converting to 24 or 16-bit word length. However, DickLaurentIsDead's question brings up a good point. Some folks read online that normalizing to -0.1 or -0.3dB is all that's required. However. conversion to MP3, AAC (and most likely other compressed formats, I only tested MP3 and AAC) will result in digital overs sometimes. Sometimes they'll be audible, sometimes not. For my personal use, I like -1dB, it's still fairly loud but safe. In reality as long as one keeps some peaks above -6dB, no bits are being fully wasted, but that's way too low for practical listening on most equipment.
Rats. Can't go back now, most of my stuff is all WAV -.3. Ultimately, I want to put these in cloud storage, in some way shape or form. Play Google doesn't allow WAV, so I was going to convert them. Right now I have them on a small hard drive, but I'd hate the hard drive to eventually die. Where do you guys store your recordings? Thanks for the answers.
Don't sweat it. Depending on what you use to convert to MP3, you can normalize the stream going into the encoder to -1dB and then you'll be safe. I typically use Illustrate's excellent dBPowerAmp software most of the time, altough there are lots of other choices out there. This would leave your WAV files intact while ensuring the MP3 conversion process doesn't introduce digital overs.
Same here. Storage is cheap these days! I just wish I had the time to organize my files the way I'd like. Retirement is 5 years away, so I'm looking forward to doing it then (along with a gazillion other things!).
I don't bother with adjusting levels before converting to mp3. Since it is not my official version of my collection, I don't worry if something exceeds digital 0.
And some clipping can be due to quantization errors, minuscule peaks at best which is why you keep the file in 32 bit floating point if planning on future edits. The audio engineer for Flaming Lips' "At War With The Mystics" album won a Grammy for best engineered album and it has quite a bit of clipping throughout some of its tracks, all quantization errors. Couldn't hear any artifacts or distortions from it.
Very true. As I wrote above, some overs will result in obvious audible glitches, others not. I just don't want to take the chance, even if it's MP3 for the car, I prefer not to have it littered with clicks.
Well if I remember correctly, you use a limiter right? If that's the case, you won't have any overs anyway.
32-bit floating point cannot go over, even without a limiter. Some editors won't show peaks over 0dBFS but the peaks are still there once you normalize down. Your limiter *would* be useful for 24- or 16-bit word length because those formats will chop the peaks off.
I'm curious about what the difference is between applying Normalize vs Amplify effect in 32bit Floating point after I apply an extreme EQ in Audacity which will show a lot of clipping. Does Normalize do something better to the sound over the Amplify effect?
Well, Normalize can correct the DC offset if necessary but other than that, no practical difference in single track mode, as far as I know. However, if you process multiple tracks at once - in a multitrack mix - they behave fundamentally differently. Normalize amplifies all tracks towards 0 dB (or whatever the target is set at) so a quieter track might be amplified by 10 dB while a louder track is amplified by, say, 3 dB. Whereas Amplify would raise the levels equally (in previous case probably by 3 dB to avoid the loudest track from clipping). So they serve a different purpose.
MP3 applies multiband filtering, which can increase the peak level. The LAME encoder will let you see the clipping on encode. I answered this in detail 10 pages ago: Your Vinyl Transfer Workflow (sharing best needledrop practices)*; and in further detail 6 pages ago: https://forums.stevehoffman.tv/threads/your-vinyl-transfer-workflow-sharing-best-needledrop-practices.250442/page-78#post-20389379
True normalizing is nothing more than adjusting the amplitude of audio data in a file so that the highest peak in the data reaches whatever limit you set. all other levels in the file will be adjusted by the the same amount. Let's say you had a file with its highest peak at -3dB and you set the target to -1dB. The entire file would be boosted by 2dB so that the highest peak now reaches -1dB. Likewise, if you had a 32-bit floating point file with a peak that was +4dB and you normalized it to -1dB, then the entire file would be attenuated by 5dB. You mentioned the "Amplify" effect so I assume you're using Adobe Audition (or its predescessor, Cool Edit Pro). It also boosts or attenuates, but there's no preset target. Just be sure to do all such processing with the file in 32-bit floating point format so that you're getting as much accuracy as possible in the math that occurs underneath this all.