ReplayGain LUFS Target Volume in CD Ripper, is not the same as in Batch Converter

[oeps]

In the ReplayGain Advanced dialog box of dBpoweramp Batch Converter (Windows R17.1), I set the EBU R128 LUFS Target Volume to -23. After that, when using the batch converter, adding ReplayGain tags to audio files works as expected.


But when I rip a music CD, the added ReplayGain tags are for EBU R128 LUFS Target Volume -18! Is there a way to set the LUFS Target Volume to be used when ripping a music CD? I would like to set this to -23.

[schmidj]

Re: ReplayGain LUFS Target Volume in CD Ripper, is not the same as in Batch Converter

Why did you choose -23? Unless you are going to apply the resulting tags to change the actual LUFS levels of the stored files and then use the flies in professional digital audio equipment at a TV station or network, -23 is the wrong value. Professional TV audio standards developed by SMPTE are very different than those used by the manufacturers of consumer audio equipment. If you set the loudness to -23 LUFS you will find your tracks will play very softly in most any consumer player. Settings from -14 (aggressive) to -18 (conservative) work best for consumer players such as Itunes or Fubar2000 on a PC or most portable players.

[Spoon]

Re: ReplayGain LUFS Target Volume in CD Ripper, is not the same as in Batch Converter

It is the same DSP effect, you should also have an advanced page in cd ripper.

[oeps]

Re: ReplayGain LUFS Target Volume in CD Ripper, is not the same as in Batch Converter

@Spoon:
You are completely right. I just didn't look in the DSP tab when I had to. My bad.

@schmidj:
I want to choose just one LUFS Target Volume for all my music, including classical music. The dynamic range of classical music, is larger than the dynamic range of most other music styles. A LUFS Target Volume of -18, just doesn't offer the headroom needed for classical music.

I tested all players I will be using for LUFS -23 audio files, and although the volume of the sound is less than with LUFS -18 audio files, the sound is never too quiet when I adjust the volume setting upwards. Perhaps, as a legacy of the loudness wars and the earlier customary ReplayGain setting of 89 dB, people are not used to a higher volume setting when playing music.

See "Example: Mary Black's album »No Frontiers« (1989)" in this page: https://github.com/Moonbase59/loudgain. Quote from that page: "At -23 LUFS, we have ample headroom, no gain shifts are needed to prevent clipping, a good thing. Apparently the EBU knew what they were doing."

When I watch TV, I get -23 LUFS. Why wouldn't I want the same, when listening to music? To me, the idea that when you use ReplayGain tags and you are not using professional digital audio equipment at a TV station or network -- that you then need to lower the quality of the music you listen to, is rather silly. When I get a disease, I hope I will get the same treatment as my doctor -- when this doctor gets this disease.

[oeps]

Re: ReplayGain LUFS Target Volume in CD Ripper, is not the same as in Batch Converter

@Spoon:
You are completely right. I just didn't look in the DSP tab when I had to. My bad.

@schmidj:
I want to choose just one LUFS Target Volume for all my music, including classical music. The dynamic range of classical music, is larger than the dynamic range of most other music styles. A LUFS Target Volume of -18, just doesn't offer the headroom needed for classical music.

I tested all players I will be using for LUFS -23 audio files, and although the volume of the sound is less than with LUFS -18 audio files, the sound is never too quiet when I adjust the volume setting upwards. Perhaps, as a legacy of the loudness wars and the earlier customary ReplayGain setting of 89 dB, people are not used to a higher volume setting when playing music.

See "Example: Mary Black's album »No Frontiers« (1989)" in this page: "https://github.com/Moonbase59/loudgain". Quote from that page: "At -23 LUFS, we have ample headroom, no gain shifts are needed to prevent clipping, a good thing. Apparently the EBU knew what they were doing."

When I watch TV, I get -23 LUFS. Why wouldn't I want the same, when listening to music? To me, the idea that when you use ReplayGain tags and you are not using professional digital audio equipment at a TV station or network -- that you then need to lower the quality of the music you listen to, is rather silly. When I get a disease, I hope I will get the same treatment as my doctor -- when this doctor gets this disease.

[schmidj]

Re: ReplayGain LUFS Target Volume in CD Ripper, is not the same as in Batch Converter

As I tried to explain earlier, and posted in an extensive post a couple of years ago, the audio standards used in the TV industries adopted by SMPTE and various international agencies set a "reference level" (long before loudness) of -20 dBFS (in the USA, -18 in most of the rest of the world). This means in the digital audio world, the level of the "setup tone" is recorded at a level of 20 (or 18) dB below the 0 dBFS peak level clip point. If you are using a properly calibrated VU meter on a professional mixing console the tone will be at the "zero" level near full scale. The intent was that one would mix the program material such that the peaks on the meter of the loudest part of the program would occasionally to regularly (subject to a great deal of personal interpretation) have the needle on the meter hitting "zero", but not too much above.

Since the VU meter is an average reading meter, not a peak reading meter, the resulting digital audio peak levels would range from -10 dBFS upwards to possibly -6 dBFS, depending on the program material and its dynamics processing. In fact TV network "delivery specifications" typically would reject a program if the audio peaks exceeded -6 dBFS, as there was concern that in analog parts of the network facility, anything above -6 in the digital world might put the analog side into clipping because of insufficient analog "headroom". And much of the professional music recording business adopted similar standards once digital desks became common partly because the same manufacturers made many of the same mixing desks for TV and recording studios, and many studios gained clients from both the TV and music businesses.

But many of the people who started mastering CDs in the 1980s copied what they had done on vinyl, make it as loud as possible. This made sense on vinyl, with its limited signal to noise ratio, but in hindsight was a mistake with quieter digital CDs. They did this by "normalizing" the loudest point of the recording to 0 dBFS, the loudest you can put in a digital file without clipping, or slightly below. And to make it worse, producers, particularly of "pop" music wanted their music to be louder than the competition. They did this by compressing, limiting, and clipping the source recording to reduce the dynamic range. This meant when the now reduced peaks were normalized to 0 dBFS, the average level would be higher, hence louder to the listener, who supposedly would more likely listen to the song if it was louder.

In the TV world, a somewhat similar story was evolving, at least while the broadcast was still analog. In order to comply with then FCC rules, the peaks of the broadcast had to be between 80 and 100% modulation of the transmitter carrier. That is a range of two dB. Furthermore, the modulation FM sound carrier on TV is pre-emphasized at sending and de-emphasized in the receiver in an attempt to decrease the perceived noise level. The side effect of this is to greatly reduce the peak legal audio level of material containing high frequencies. In order for stations to comply with the FCC and still sound reasonably "loud", they installed audio processing involving compression, limiting, and clipping similar to that used in mastering CDs at the audio input to the transmitters.

The advertising agencies, wanting their commercials to be as loud as possible (thinking that more people would probably notice a loud commercial) also processed their audio, often more extremely that the surrounding program material, greatly reducing the dynamic range. Because of the way the TV transmitter audio processing worked, maintaining the peak level in compliance with the FCC rules, the average levels of the commercials (and therefore the perceived loudness) was often louder than the surrounding program.

Members of the TV audience world-wide took offense to "loud commercials" and complained loud and clear to the regulatory agencies like the FCC and to their (often elected) representatives who found it necessary to take legislative action. But they needed some means of determining how "loud" the audio was. They bounced the ball to SMPTE and the European Broadcast Union (EBU) who did some research and came up with the Loudness Unit (LU) similar to the Volume Unit (VU) but supposedly adjusted to conform to the human ear and brain perception of loudness. The different agencies came up with simple measurements (look at the details, pretty basic) to measure "loudness" and determined (very arbitrarily from some listening tests) that to match an average program recorded properly with an audio reference level of -20 or -18 dBFS, the level as measured on a standard "loudness meter" should be -23 (or -24, depends on what country you're in) LUFS or whatever that country called its loudness units, and that if all TV broadcasts (and the standard and resulting legislation only apply to TV) met the standard when measured with said loudness meter, that supposedly the "loud commercial" problem would disappear, and if stations didn't meet the standard, they could be fined (at least in the USA).

But there are some caveats that one should be aware of: The actual measurement is very basic and simplified, it only takes a small part of knowledge of human hearing perceived loudness in the measurement technique. It was mostly tested with speech, remember this was for TV. It seems to work well enough there that it quieted the complaints to the politicians about loud commercials, and therefore to them was a "great success". The people who developed this gave little if any thought about its application to the recorded (or streamed) music business. To my far from perfect (at my advanced age) ears, it works "so so" for music, probably not much better if any better than a simple audio "normalizer" which sets average VU levels (not LU) to some predetermined output level. It can be "fooled" by some music, and does not work particularly well when matching music with little dynamic range with music with large dynamic range. And if applied on a "track by track" basis to an album, as opposed to on the album as a whole it will remove the dynamics between tracks that the producer and/or musicians intended.

Obviously, you can set the target volume wherever you want, it is your music and your listening. And the loudness normalizer in the ripper/converter will never actually modify the audio. It only sets a tag, so you can change your mind and run it again if you don't like the original results. It also in the default setting will never let the music "clip", so if as an extreme case you set it to run it with an LUFS of let us say 8, the heavily compressed pop music might actually end up with an LUFS of 8, but the wide dynamic range classical and some jazz will only be set to the loudest it could be without clipping, so it might be an LUFS of -20 or whatever. The dBpoweramp normalizer does not include any dynamics processing so it will never decrease the dynamic range of the source, no matter how great that might be.

In my personal preference, I'd actually like to make a second copy of my library and apply the -18 LUFS tags, modifying the files in the copy, then run them through a (quite) transparent loudness limiter to lightly limit the material which had too much dynamics to set the LUFS to -18 so it hits the target. I own a copy of Izotope's loudness limiter and to my not-so-great old ears, the limiting is pretty transparent. Much of the "loudest" part of live recordings is very brief transients that are pretty inaudible. If they are stomped down a few dB, I think most people would never hear the difference. The only reason I haven't done this is first, I have to read how to use the Izotope limiter on a batch basis, and then how long it would take to convert my 60,000+ files.

The one issue you may run into when using -23 LUFS, if you are listening on a consumer receiver and switch to FM or a CD or any other source without first turning the volume down, you'll probably get blasted by some pretty loud audio. The designers of most consumer gear assume that the audio from a file (or the "aux" input if it isn't a "network" receiver is at the equivalent of something like -14 LUFS.

And I'll be frank, excessively wide dynamic range doesn't always equal "quality". It just as often is "laziness" on the part of the mastering engineer, mixing engineer or producer. This is particularly the case on so much material produced in someone's "home studio" and provided for download with no professional mastering. You really have to determine if those loud peaks are really part of the music or just some kind of inaudible artifact in the recording.

[garym]

Re: ReplayGain LUFS Target Volume in CD Ripper, is not the same as in Batch Converter

Thanks @schmidj. As always, very informative.

[oeps]

Re: ReplayGain LUFS Target Volume in CD Ripper, is not the same as in Batch Converter

Sorry for my double post.

@schmidj: Thank you for your impressive post.

I agree that excessively wide dynamic range doesn't always equal "quality". But I do consider a lot of clipping prevention, to be a reduction of quality -- if clipping prevention does something to make the loudest part of the music less loud without changing its average loudness level. However, if clipping prevention does nothing else than reducing the music's average loudness level [or keeping it low] without changing its dynamic range (i.e. it does some kind of peak normalization), than clipping prevention does not reduce quality. Now I understand that the latter is exactly what happens with ReplayGain.

It also in the default setting will never let the music "clip", so if as an extreme case you set it to run it with an LUFS of let us say 8, the heavily compressed pop music might actually end up with an LUFS of 8, but the wide dynamic range classical and some jazz will only be set to the loudest it could be without clipping, so it might be an LUFS of -20 or whatever. The dBpoweramp normalizer does not include any dynamics processing so it will never decrease the dynamic range of the source, no matter how great that might be.

If I understand what you are saying, this is what happens when using ReplayGain with a relatively low asolute LUFS value like -8:

• The more compressed a piece music is (i.e. the lower the dynamic range), the higher the perceived average loudness of this piece of music will be.
• The less compressed a piece music is (i.e. the higher the dynamic range), the lower the perceived average loudness of this piece of music will be.

So, with a relatively low absolute LUFS value, the "ReplayGain action" does not do what one would expect: to make the perceived average loudness of different pieces of music about the same. (The higher the absolute LUFS value is, the smaller this effect will be.)

So, if you want to get an equal perceived average loudness for all your different pieces of music, you need to choose a relatively high absolute LUFS value. And if you really want to get a perceived average loudness that is equal to your piece of music with the highest dynamic range, you have to choose a very high absolute LUFS value.

So, when clipping prevention is a kind of peak normalization, then the choice of the LUFS value is not about the perceived average loudness you want for all your different pieces of music -- but about what difference in perceived average loudness you want (or accept) for your different pieces of music. (The less dynamic range there is in your pieces of music, the lower absolute value for LUFS you might tend to choose.)

I thought that clipping prevention was about making the loudest part of music less loud without changing its average loudness level -- and if this would happen, it would reduce the quality of the music. Now that I understand that I was wrong about this, and that there will be no reduction in quality when choosing a lower absolute LUFS value, it makes sense to reconsider my earlier preference for LUFS -23. If I would choose something like LUFS -18, I will have to be prepared to have to adjust the volume setting when listening to some pieces of music -- but perhaps this will not happen very much.

I supposed that if -23 LUFS is enforced at the TV station, that I then would get -23 LUFS at my TV set. Is this assumption wrong, and do I get less dynamic range at my TV set?

[oeps]

Re: ReplayGain LUFS Target Volume in CD Ripper, is not the same as in Batch Converter

After thinking it over, I decided to standardize on a EBU R128 LUFS Target Volume of -20. This is a compromise between a value of -23 and -18. With this, when playing music, there will be less perceived average loudness differences than with a value of -18, and more perceived average loudness differences than with a value of -23.

In addition, I decided not to use any ReplayGain tags with classical music and other high dynamic range music, because ReplayGain with clipping prevention, is just a kind of peak normalization that will not do very much anyway with high dynamic range music.