Tweet
Since I have been using the Replay Gain DSP set to use -18 LUFS, i continue to be surprised at the number of tracks that the DSP reports are not able to be adjusted to -18LUFS, as it would drive the peaks into clipping. There are several issues:
First, the report of the replay gain DSP shows up in the information log at the end of the rip, but appears not to be saved in the filed copy I save of the rip report. Furthermore, I am using the multi-encoder to FLAC and AAC, in different directories, and the rip report is only saved in the FLAC directories (the first encoder). I suspect that the information from other DSP's is not being saved in the report as well. It would be highly desirable to save this information with the rip report, and to have the option to save it with each encoding when using the multi-encoder. Perhaps this option should be another DSP for the encoder rather than or in addition to being in the secure rip setup options.
Second, even in the on-screen report, all it says is that it wasn't able to fully apply the DSP. It would be highly desirable if it reported, both on-screen and in the saved log, what the calculated loudness of the track (with the replay gain applied) is. If it misses by a dB or two, not to worry. 10 dB, well not good.
Third, how is the DSP determining what it calls "clipping"? Two or more consecutive samples at 0dBFS? Some number based on the possibility of oversampling or filtering generating a "between sample" result above 0dBFS or the analog equivalent possibly clipping? Or some more conservative maximum allowed peak level? This is still a very controversial subject in the pro audio community. At the major broadcast network I designed audio systems for, our delivery spec failed any peaks above -6dBFS (with a target max loudness of -23 LUFS +- 2), but that is for the different audio level standards we used in the professional world, and is not appropriate for the consumer world as it now exists.
I am particularly surprised at seeing so many tracks like this, assuming the clipping message is triggered by some peak level very near 0dBFS, as the EBU (I recall) is now recommending that consumer loudness normalization aim at -14LUFS, 4 dB hotter than the -18 setting I (and I suspect many other dBPoweramp users) am using. -18 is a quite conservative setting.
Now, the second part of the wish list: to provide an option to provide/apply (as chosen) replay gain normalization adjustment that fully applies, but soft limits (not clips) the peaks that would otherwise clip. It would be really nice if the report said at how many minutes/seconds into the track each limiting episode occurred (as the silence detector DSP now does). Now, I know some purists won't want to modify their ripped FLAC file in any way, you could keep two copies, one with no limiting and one with limiting. For me, having my played back music at a constant loudness trumps that concern. That is particularly true for the m4a copy I listen to in the car.
I've worked in this business for decades, and recorded and mastered a lot of live recordings myself. The vast majority of times, these peaks 18 or more dB above the average program level are small instantaneous spikes, totally inaudible, that can safely be limited without degrading the quality of the end product. Some of them may be artifacts of the digitization process. In pop music, many of them are momentary spikes from cymbal crashes or such. When I've mastered CD's, I've always chopped (soft limited) them off to make a reasonably loud playing CD, and believe me, I'm no supporter of the pop music "loudness wars". I've never had anyone comment/complain about what I've chopped off, and I get a lot of compliments abiut the quality of my live recordings.
Once in a great while such limiting may result in "pumping" where limiting a peak in one frequency band causes audible loudness "dips" in the rest of the spectrum. The solution when mastering is to use a "multi-band" limiter, but that is for use in an audio editing program, not a ripper.
The only music I can think of where automatic limiting may not be appropriate is wide dynamic range classical music and possibly some jazz.
First, the report of the replay gain DSP shows up in the information log at the end of the rip, but appears not to be saved in the filed copy I save of the rip report. Furthermore, I am using the multi-encoder to FLAC and AAC, in different directories, and the rip report is only saved in the FLAC directories (the first encoder). I suspect that the information from other DSP's is not being saved in the report as well. It would be highly desirable to save this information with the rip report, and to have the option to save it with each encoding when using the multi-encoder. Perhaps this option should be another DSP for the encoder rather than or in addition to being in the secure rip setup options.
Second, even in the on-screen report, all it says is that it wasn't able to fully apply the DSP. It would be highly desirable if it reported, both on-screen and in the saved log, what the calculated loudness of the track (with the replay gain applied) is. If it misses by a dB or two, not to worry. 10 dB, well not good.
Third, how is the DSP determining what it calls "clipping"? Two or more consecutive samples at 0dBFS? Some number based on the possibility of oversampling or filtering generating a "between sample" result above 0dBFS or the analog equivalent possibly clipping? Or some more conservative maximum allowed peak level? This is still a very controversial subject in the pro audio community. At the major broadcast network I designed audio systems for, our delivery spec failed any peaks above -6dBFS (with a target max loudness of -23 LUFS +- 2), but that is for the different audio level standards we used in the professional world, and is not appropriate for the consumer world as it now exists.
I am particularly surprised at seeing so many tracks like this, assuming the clipping message is triggered by some peak level very near 0dBFS, as the EBU (I recall) is now recommending that consumer loudness normalization aim at -14LUFS, 4 dB hotter than the -18 setting I (and I suspect many other dBPoweramp users) am using. -18 is a quite conservative setting.
Now, the second part of the wish list: to provide an option to provide/apply (as chosen) replay gain normalization adjustment that fully applies, but soft limits (not clips) the peaks that would otherwise clip. It would be really nice if the report said at how many minutes/seconds into the track each limiting episode occurred (as the silence detector DSP now does). Now, I know some purists won't want to modify their ripped FLAC file in any way, you could keep two copies, one with no limiting and one with limiting. For me, having my played back music at a constant loudness trumps that concern. That is particularly true for the m4a copy I listen to in the car.
I've worked in this business for decades, and recorded and mastered a lot of live recordings myself. The vast majority of times, these peaks 18 or more dB above the average program level are small instantaneous spikes, totally inaudible, that can safely be limited without degrading the quality of the end product. Some of them may be artifacts of the digitization process. In pop music, many of them are momentary spikes from cymbal crashes or such. When I've mastered CD's, I've always chopped (soft limited) them off to make a reasonably loud playing CD, and believe me, I'm no supporter of the pop music "loudness wars". I've never had anyone comment/complain about what I've chopped off, and I get a lot of compliments abiut the quality of my live recordings.
Once in a great while such limiting may result in "pumping" where limiting a peak in one frequency band causes audible loudness "dips" in the rest of the spectrum. The solution when mastering is to use a "multi-band" limiter, but that is for use in an audio editing program, not a ripper.
The only music I can think of where automatic limiting may not be appropriate is wide dynamic range classical music and possibly some jazz.
Comment