Volume normalizing

[sokd623]

Re: Volume normalizing

Hello schmidj and thank you for your detailed answer.

Just to clarify, I am not confused about ReplayGain, if I'm confused, it is about how to apply it using dBpoweramp.
I used Foobar2000 to apply ReplayGain metadata on my library in the past, but that didn't seem to make much difference on the very loud files, lossy or lossless.
So, I checked dBpoweramp and I saw other options, so I decided to try them out, but this program works differently and I just want to be clear before I do anything, of course, I work on a copy.
I do use a player that reads ReplayGain data, I use Pulsar for Android, also supported by Android Auto! But as I said, didn't seem to make much of a difference on the loud files.

Last night I run the ReplayGain SDP on my library. Still need to test it. I will return after the test is complete. If it does what I want it to do, then I'm good.
Since you and garym never used the ReplayGain (Apply) SDP, I shouldn't need to either, thus, not making and permanent change to my files.

Thank you for both of you,

[garym]

Re: Volume normalizing

Hello schmidj and thank you for your detailed answer.

Just to clarify, I am not confused about ReplayGain, if I'm confused, it is about how to apply it using dBpoweramp.
I used Foobar2000 to apply ReplayGain metadata on my library in the past, but that didn't seem to make much difference on the very loud files, lossy or lossless.
So, I checked dBpoweramp and I saw other options, so I decided to try them out, but this program works differently and I just want to be clear before I do anything, of course, I work on a copy.
I do use a player that reads ReplayGain data, I use Pulsar for Android, also supported by Android Auto! But as I said, didn't seem to make much of a difference on the loud files.

Last night I run the ReplayGain SDP on my library. Still need to test it. I will return after the test is complete. If it does what I want it to do, then I'm good.
Since you and garym never used the ReplayGain (Apply) SDP, I shouldn't need to either, thus, not making and permanent change to my files.

Thank you for both of you,

Aha. Got it. (if we'd known you only wanted ReplayGain tags and didn't want to permanently modify your files' audio content, we would have initially said ignore the ReplayGain(apply) or Volume Normalization DSPs, as they are not relevant.)

Perhaps you thought that ReplayGain(apply) somehow "turned on" the RG tags so that players can use them. This is not the case of course. But your comment raises a few other issues. I use both foobar2000 and dbpoweramp to add ReplayGain tags to my files. foobar2000 uses the same algorithm and adds the same tags. The default in foobar2000 is to use EBU R128 method and -18 LUFS, and to not override clipping prevention. So both programs handle RG tagging the same way.

If you're not hearing any effect from having the ReplayGain tags in your files I have a couple of thoughts. Of course first confirm you see those tags in your metadata with a tag editor. Next, even if you have RG tags, you must tell your player to USE those tags in playback. For example in foobar2000, you go to FILE > PREFERENCES > PLAYBACK, then in the top section on ReplayGain, select an option. I prefer "by playback order". This automatically uses ALBUM RG tags if playing a complete album and TRACK RG tags if playing a mixture of tracks from different albums. On this page you can also set the preamp, to make everything louder if you want. I set this to 0db for files with RG tags, and make it -5db for files with no RG tags. My other players are similar. LMS requires you to turn on use of RG tags (and also has something called "SmartGain" which is equivalent to the "by playback order" in foobar. Roon labs has something similar. And foobar2000 mobile on my iphone has the ability to turn on or off use of RG tags.

So bottom line, one must have RG tags in the metadata *and* tell the player to use those tags. I hope this helps.

[sokd623]

Re: Volume normalizing

ReplayGain setting on my phone was turned on the day I received it. It is set to ReplayGain Mode: Auto, which is equivalent to SmartGain in Foobar2000, Preamp is at 0DB and there is a setting for files which do not have RG tags (not applicable in my case) called Default Gain, and is also set at 0DB. I'll try the loud files and see how they sound. I'm pretty sure they were lowered compared to the whole library.
In Foobar, I could see the values applied to my files, is there a way to see them with dBpoweramp? if not, I'll load the files on Foobar, not a big deal.
I may need to edit tags manually, to lower the values even more, or, if nothing else works, I'll have to use ReplayGain (Apply) or Volume Normalizer, at least on the lossless files. They seem to be easier to work with since the only thing that could happen to them is lower the loudness too much and I know is not going to happen. I already played with some of my loud files, used Volume Normalizer to lower them by -2db, there was a significant difference in how they sounded.
I'll rewrite in a few days.

Thank you,

[garym]

Re: Volume normalizing

just remember, even on lossless files, if you use replaygain(apply) they are no longer bitperfect copies of the CD. I use mp3tag for tag editing work and can easily see the RG tag info of all my files. But foobar2000 works too. and if I right click a file in file explorer, dbpoweramp shows up as an option for looking at file properties and tags (if one set that up when installing dbpa).

[GBrown]

Re: Volume normalizing

... if nothing else works, I'll have to use ReplayGain (Apply) or Volume Normalizer, at least on the lossless files. ...
Thank you,

There should be no difference using RG versus RG Apply when played - except with Apply there is no need to have a player that supports RG. BUT this will destroy the integrity of the original files permanently. Volume Normalizer works in a different way, so this may be more what you are looking for, but again the files are no longer 1:1 copies of the original.

[schmidj]

Re: Volume normalizing

I think it is time to talk a little more about loudness control and replay gain. I’m sorry this is so long, but there is a lot of meat to chew on here.

The matter of controlling loudness when playing back recordings and in broadcasting has been an issue since recordings began to be made. The motion picture industry has its standards, the music recording industry has discussed it ad infinitum but has no standard, and the broadcast industry has had standards (which differ depending on the national jurisdiction) shoved down their throats in response to complaints about “loud commercials”.

The basic problem is that no two people will consistently agree on what is louder than what. There is a lot of psychoacoustics involved on what is conceived as loud, including to a good degree whether the listener is enjoying the audio or annoyed by it. To that, add the effects of dynamic range and audio processing. If I listen to most pop music, it has little dynamic range. From beginning to end, the volume is unchanged. But, for instance, most classical music has a wide dynamic range. Part are “loud”, parts are ”soft”. If I listen to the pop music and then immediately to some part of the classical music, the classical piece may sound louder or softer depending on what part of it you are listening to. If it is of any length, it is unlikely that most listeners would be unable to indicate whether the entire piece is louder or softer than the pop music.

Back in the 1960’s CBS Labs, studied the “loudness” issue and developed a highly regarded (among researchers) loudness meter and controller. These devices took into account the sensitivity of the average human ear to different frequencies. The ear is relatively insensitive to low frequencies. It gets more sensitive to higher frequencies up to a few kilohertz and then becomes less sensitive to the highest frequencies, particularly as people age or are excessively exposed to loud sounds. The devices were not commercial successes, however, as the FCC had broadcast modulation requirements which did not take loudness into consideration, and each record company had its own policies (often violated by producers and engineers) on how loud each recording should be. In particular, pop recordings were made as loud as possible and most pop radio stations processed their audio to be, again, as loud as they could make it while complying with FCC regulations. Why? Because the general opinion was that if it was loud, it was likely to get more listeners and sell more.

But advertisers also would make their commercials as loud as possible, thinking that consumers would be more likely to remember the commercials and buy the products. Some of this loudness was technical processing, heavy compression and limiting, and some of it was psychoacoustics. (Think the screaming Crazy Eddy commercials on many New York stations. They sounded loud even if they weren’t, because he was screaming (and annoying.) And the listening/viewing public complained more and more to the FCC and then their representatives in Congress. Congress eventually directed the FCC to take action. Similar complaints in Europe and elsewhere began to force broadcast regulators to take action.

The various broadcast regulators got the International Telecommunication Union to establish a committee to investigate and propose a solution to the television loudness issues. For those who may be interested, the ITU was founded in 1865 to standardize telegraph communications between nations and is now claimed to be the oldest intergovernmental organization. It is now part of the United Nations.

After some study and testing, the ITU came up with a standard for measuring loudness called BS.1770. It is now on ( I think) its fourth revision, BS.1770-4. Note that this is a broadcast standard, not a commercial recording or streaming standard. It is based on SMPTE and EBU standard broadcast production audio levels, which are very different from the audio levels used on most commercial audio recordings, files and streams. More later.

But in typical bureaucratic style, each affected jurisdiction (the USA via the FCC, in the European Union via the EBU and various other broadcast regulators added their own tweaks to the standard before making them requirements to be followed by TV broadcasters. The EBU adopted R128. The FCC tasked the ATSC (which recommends TV broadcast standards in the USA) with developing a standard, they responded with A/85, the USA standard. The Chinese, the Japanese, the Australians, etc. each developed their own particular variant. These variants work similarly for the most part, but terminology, reference levels, display standards and some of the details of the calculations differ.

Some may ask, what about using the old CBS Labs loudness meter? Why re-invent the wheel? Unfortunately, CBS Labs has been out of business for many years. Also, that device is a purely analog two channel device, unsuited for today’s digital, multi-channel world. But Orban Labs, a well known manufacturer of radio and television audio processing devices has written an interesting paper comparing the CBS Labs readings (for which Orban has developed a digital approximation of, free for download) with BS.1770 and some of its offshoots. If you are deeply interested, the paper is worthwhile reading.

https://static1.squarespace.com/static/58f8d954b8a79b4ccf726c3b/t/5996dc75a5790ac3a352590a/1503059062973/Using+the+ITU+BS.1770+and+CBS+Loudness+Meters+To+M easure+Loudness+Controller+Performance.pdf

I mentioned earlier, the television industry uses different digital level standards than the recording, radio and streaming worlds. When digital audio became widely used in TV production in the 1980’s there was a need to develop a standard digital level to correspond to the TV production “set up” level. But, even between the TV broadcasters, their internal standard levels varied, and in the case of the ABC New York network facilities, in the early 1980’s we converted from a +8 dBm 150 ohm matched network to a +8 dBu bridging network. Meanwhile, almost all recording facilities standardized on +4 dBu bridging plants. It is beyond this report to dig deeper into this, but it is important to realize that these are “setup” or “reference” levels, not peak audio levels. Historically, the peak audio level would be about 10 dB above the setup level and an additional 6dB safety margin to a clip point of +24 dBu. This is all in the analog realm. But with the onset of digital recording and transmission, there was pressure to increase the headroom/safety margin to a full 20 dB. The SMPTE chose a digital audio setup level of -20 dBFS which gives the digital system the 20 dB of margin In Europe, the EBU chose a setup level of -18 dBFS, two dB hotter, hence two dB less margin.

But the recording industry was used to recording music as loud as possible, as stated before (and for some other reasons beyond this report) adopted digital setup levels nearer -15 dBFS, varying between facility. And released CDs mostly followed the practice of discs, as loud as possible, which meant the peak track (not the setup) level would be set as close to 0 dBFS as they dared. This meant that digital audio levels on commercial CDs were typically 6 or more dB “hotter” than the levels in TV plants.

And radio stations, at least in the USA mostly wanted to be louder than the competition; furthermore FCC regulations for modulation (still) want the modulation on peaks of “frequent recurrence” to be within 2 dB of the maximum allowed

This did not matter until the various TV loudness standards were developed, and at about the same time, people started ripping CDs to play on computer systems and realized the CD levels were all over the place, annoying to listen to. Various algorithms were developed to adjust the levels, and the “Replay Gain” tag (and its Apple equivalent) were developed to allow the volume to be adjusted at the playing software, leaving the ripped audio unchanged.

People soon realized that the TV broadcast algorithms tended to do a better (but by no means perfect) job of estimating the loudness of a track than most of the proprietary algorithms (such as the dBpoweramp original “replay gain” algorithm). But when they applied them using the TV standard level of -23 or -24 (depending on which standard one applied), the audio was way too soft. This was because of the previously discussed difference between the typical CD audio levels and the SMPTE or EBU standard TV levels. Some experimenting resulted in choices between -18 and -14 instead of the TV -23 or -24. dBpa has a default value of -18, on the conservative side. That works fine for me on my home system but is on the “soft” side for my phone/Bluetooth/car playback. So, I recalculate the replay gain numbers to -16 for the m4a tracks on my phone.

Because many of the Internet services and streamers use BS.1770 to adjust the levels of their audio, you might find this link interesting showing the different levels they use. I can’t vouch for the accuracy of the chart.

https://www.rtw.com/en/blog/worldwide-loudness-delivery-standards.html

There is one more setting it is important to discuss here. There is a check box on the replay gain algorithm titled “disable clip prevention”. What does that do and why would I ever want to disable it?

The issue here is tracks with particularly loud parts on mostly soft music. Sometimes it is the dynamic range of the music, and sometimes (actually quite often in some genres) it is poor mixing where some not particularly audible part of the mix, like a cymbal crash or even some almost inaudible brief clicks are much “louder” than the “important” parts, like the singer or melody instruments. If the replay gain tag value were set to make the playback of the average (presumably the “important” part) of the music to the desired level, those loud parts might be above 0 dBFS at the player, i.e. clipped. Oh awful, distortion! Well, possibly, possibly not. It depends on what you clip, and how much. I’ve had field recording sessions where due to an error, either in level setting or hardware configuration, there were literally thousands of minor clips in the recording that were totally inaudible, but I’ve had other cases where a single burst of clipping was all too audible.

Leaving the box unchecked will tell the program to never let the value of the replay gain tag be so high for a track that the audio will be clipped in the player. This avoids any chance of playback distortion, but may make the track softer (sometimes much softer) than the adjacent tracks.

But there is another solution. First remember that no matter what the number stored in the replay gain tag is, the actual audio as ripped is not changed. If it wasn’t clipped on the CD, it isn’t clipped on the digital file. Once again, the replay gain tag does nothing until a player plays the track, it does not change the stored audio of the track file.

And several of the better players, including Foobar 2000 and Android’s “Poweramp” (no connection AFAIK to the dBpoweramp folk) have selectable limiters as playback options. That option will essentially always momentarily turn the playback audio down enough to prevent clipping if the replay gain adjustment would otherwise cause the audio output to clip. Recording engineers and broadcasters use limiters routinely to avoid momentary peaks in the audio that might result in clipping, and a good limiter properly applied is typically inaudible.

So, if your player has the option, you should consider tuning the limiter in the player on and checking the box in the replay gain settings to disable the clip prevention. And remember, because the Replay Gain setting is just a tag which can be recalculated as many time as you want, you can try this setting by running the Replay Gain DSP by itself as a “conversion” with the box checked, and if you want to change it back, just run the DSP again with the box unchecked.

This is probably far too much to absorb in one sitting, but if you dig through it you’ll have a better idea of where the settings came from.

[garym]

Re: Volume normalizing

Thanks @schmidj, I always learn something.