Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Mastering and equal loudness (Read 6370 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Mastering and equal loudness

While we're on the topic, I wanted to as something that's been bugging me.

Everybody here is looking at WAVs in Audacity and laughing at them. Isn't this meaningless unless the signal is run through an equal loudness filter, like RG's? Otherwise there could be very real dynamics present in a recording by varying the frequency content, yet the peak level could be flatlined at 0db.

Mastering and equal loudness

Reply #1 –
Maybe. But I am not just worried about dymanics. I am worried about clipping. Clipping is a non-linear distortion. It is a distortion. The drum the violin etc all has certain signature and once it is clipped it is distorted and one can find that unpleasant.

Mastering and equal loudness

Reply #2 –
Quote
While we're on the topic, I wanted to as something that's been bugging me.

Everybody here is looking at WAVs in Audacity and laughing at them. Isn't this meaningless unless the signal is run through an equal loudness filter, like RG's? Otherwise there could be very real dynamics present in a recording by varying the frequency content, yet the peak level could be flatlined at 0db.
[a href="index.php?act=findpost&pid=320599"][{POST_SNAPBACK}][/a]

For real dynamics, you need to get away from the 0db line.. eg. a 20 hz sound at -0 db will have a large period, and so will automatically give you a certain amount of time at lower amplitudes. otherwise, say a 20kz sound, with a very short period, will have to be repeated very quickly (thus limiting dynamics) in order to "flatline" a 0db.

Mastering and equal loudness

Reply #3 –
We're talking about peak values, not instantaneous values. s/peak/rms/ if you need to.

Mastering and equal loudness

Reply #4 –
Here's some samples of an original clip, then it's compressed/limited and then the files are normalized.

http://www.loudnessrace.com/victims/examples.htm

It all sounds the worst when you compare the origianl with the normalized/compressed sample. Pretty night and day.

-Joe

Mastering and equal loudness

Reply #5 –
Quote
While we're on the topic, I wanted to as something that's been bugging me.

Everybody here is looking at WAVs in Audacity and laughing at them. Isn't this meaningless unless the signal is run through an equal loudness filter, like RG's? Otherwise there could be very real dynamics present in a recording by varying the frequency content, yet the peak level could be flatlined at 0db.
[a href="index.php?act=findpost&pid=320599"][{POST_SNAPBACK}][/a]
I dont know if this answers your question or not. The wav is from a 1993 CD. 1st is original wav 2nd is original wav with 6.5:1 compression.

Mastering and equal loudness

Reply #6 –
Quote
I dont know if this answers your question or not. The wav is from a 1993 CD. 1st is original wav 2nd is original wav with 6.5:1 compression.

No, he was talking about something different. Amplitude does not equal loudness. To simplify the issue: if you play a 100hz tone and a 800hz tone at the same amplitude, then they will not sound equally loud. This is because the human ear is not "flat" but instead a curve - thus, our ears "amplify" different frequencies by a different amount. The problem is further complicated by the fact, that the "hearing-curve" of each individual is not exacly the same(however, its usually similiar, so it would not be difficult to pick an "average-joe" hearing-curve and use it for representative purposes)
I am arrogant and I can afford it because I deliver.

Mastering and equal loudness

Reply #7 –
Quote
Quote
I dont know if this answers your question or not. The wav is from a 1993 CD. 1st is original wav 2nd is original wav with 6.5:1 compression.

No, he was talking about something different. Amplitude does not equal loudness. To simplify the issue: if you play a 100hz tone and a 800hz tone at the same amplitude, then they will not sound equally loud. This is because the human ear is not "flat" but instead a curve - thus, our ears "amplify" different frequencies by a different amount. The problem is further complicated by the fact, that the "hearing-curve" of each individual is not exacly the same(however, its usually similiar, so it would not be difficult to pick an "average-joe" hearing-curve and use it for representative purposes)
[a href="index.php?act=findpost&pid=320632"][{POST_SNAPBACK}][/a]
I think I know what you're saying. For example an online hearing test where some frequency's require more amplitude or volume to be heard?  Now I'm most definately not nearly as informed as you guys about all the technicalities of audio, BUT is the original poster trying to assert that there is good dynamic music to be found amongst a highly compressed waveform? And is he saying first it must be run through an equal loudness filter in order to alter or detect its "goodness". All I know is highly compressed wavs, like the ones I've displayed sound horrible however theoretically unproven it might be to him.

Mastering and equal loudness

Reply #8 –
I never said that it was good. I was trying to say that there are more dynamics present than what people are wanting to see.

To put some meat on this conversation, here is a db waveform for the highest-RG'd song I currently have, Merzbow's "Victoria Crowned Pigeon", -16.31db:

http://www.hydrogenaudio.org/forums/index....pe=post&id=1690

And this is the waveform after applying an equalization roughly equivalent to the equal loudness filter that ReplayGain uses:

[url=http://www.hydrogenaudio.org/forums/index.php?act=Attach&type=post&id=1691][a href="http://www.hydrogenaudio.org/forums/index....pe=post&id=1691" target="_blank"][a href="http://www.hydrogenaudio.org/forums/index....pe=post&id=1691" target="_blank"]http://www.hydrogenaudio.org/forums/index....pe=post&id=1691[/url][/a][/a]

Not so bad, isn't it?

The equalizer is the default foobar one with the coefficients -20, -18, -13, -10, -8, -8, -8, -8, -8, -8, -6, -2, 0, 0, -10, -15, -20, -20.

Mastering and equal loudness

Reply #9 –
Also, a lot of people couple low RG values to "bad mastering" too quickly. Recall, that RG approximates the loudness of the file, based on the 95% RMS value of the audio. An audio file can be very loud in the top 5%, but still be full of dynamics, sparkling, clarity, punch, kick and a lot of other subjective quantities

A low RG value of -10 dB may be an indication of bad mastering, not a proof.

Mastering and equal loudness

Reply #10 –
Quote
To put some meat on this conversation, here is a db waveform for the highest-RG'd song I currently have, Merzbow's "Victoria Crowned Pigeon", -16.31db:

http://www.hydrogenaudio.org/forums/index....pe=post&id=1690

And this is the waveform after applying an equalization roughly equivalent to the equal loudness filter that ReplayGain uses:

<a href='http://www.hydrogenaudio.org/forums/index.php?act=Attach&type=post&id=1691' target='_blank'>http://www.hydrogenaudio.org/forums/index....pe=post&id=1691</a>

Not so bad, isn't it?


Very interesting. Thanks for doing this, Axon.

Before i write my main reply, some foreword: In this topic, the ha.org-philosophy is taken upside-down. What music sounds good and not is a subjective thing(and to some extend even a matter of (bad) taste). In true spirit of every audiophile, objective observations can only be an indication - but you really only know it by listening. Why does this matter? It should become clear soon......

When we had "fun" looking at waveforms in overcompression-discussions, we had something to which we could relate. I.e. a heavily clipped waveform was a strong indication that when listening to it, it will probably(but not always) not sound good.

However, this is not the case with the experiment which Axon did. Yes, it looks less bad than the "source-waveform", but thats about it. Because we lack experience with this method, we have no indication about the probability how this sounds like. Thus, it could be that with the old method, a waveform has to look really blocky until it sounds awful, while with the new "more correct" one, different "scales" may apply(thus, it may already sound bad even if it just looks "slightly bad").

To put it simply, without further tests, the "look" of axom's example tells us nothing(execpt that it looks less bad, but this wasn't about art AFAIK ;). I'm not trying to devalue Axom's idea - quite the contrary: this method may be a much more reliable indicator in the future than the "lazy"/uncorrect old one. But without further tests and experience with this method, we have no data with which we can compare our findings - and therefore cannot draw any conclusions yet.

One improvement proposal which i would make, is to disregard any content above 12khz. Low volume content above 12khz cannot really be regarded as "improving the dynamic range". So, when visualizing dynamic range then only content between...... lets say 80hz to 12khz should be taken into account. Or maybe even less, not sure about that. Basically, the idea is to only take frequencies into account, which are responsible for making a track feel "loaded" or not.

- Lyx
I am arrogant and I can afford it because I deliver.

Mastering and equal loudness

Reply #11 –
I agree that this is a somewhat artificial method that needs to have qualification. In particular, I would suspect that excessive clipping would actually look like a dynamic range increase in some cases, due to intermodulation into louder frequency ranges.

How I think this will pan out as is that ReplayGain is a somewhat better indicator of loudness than waveform graphing, and that the margins for an overly compressed record are a lot tighter than we originally thought. That is, whereas we have previously thought that 5dB of dynamic range is perhaps too little, it might be more like 8 dbA.

Here's a test for anybody with Adobe Audition: take a highly compressed track, run it through an equal loudness filter, perform another compression, then invert the filter. If I'm right, the resulting wav may be more compressed than any of us may be able to comprehend.

Mastering and equal loudness

Reply #12 –
Quote
How I think this will pan out as is that ReplayGain is a somewhat better indicator of loudness than waveform graphing, and that the margins for an overly compressed record are a lot tighter than we originally thought. That is, whereas we have previously thought that 5dB of dynamic range is perhaps too little, it might be more like 8 dbA.
[a href="index.php?act=findpost&pid=320716"][{POST_SNAPBACK}][/a]

Like I said in the post above, RG is indeed only an indicator. A piece of music can have very loud passages, which will result in low RG values, but the same track may still be full of dynamics and free of clipping.

Mastering and equal loudness

Reply #13 –
Quote
Also, a lot of people couple low RG values to "bad mastering" too quickly.[a href="index.php?act=findpost&pid=320670"][{POST_SNAPBACK}][/a]

Yes, this happens a bit too often lately.
As you correctly state, a high negative Replaygain value is only proof that the track is LOUD.  If a track is too loud, the perceived audio quality might become questionable. But comparing RG values is not an absolute indicator of quality.
A simple example, compress a track to death and lower the volume (amplitude) 6dB. "Good" RG value, but still no dynamics.

But still, the clippression on current pop/rock CD's is a shame.
In theory, there is no difference between theory and practice. In practice there is.

Mastering and equal loudness

Reply #14 –
Quote
Quote
Also, a lot of people couple low RG values to "bad mastering" too quickly.[a href="index.php?act=findpost&pid=320670"][{POST_SNAPBACK}][/a]

Yes, this happens a bit too often lately.
As you correctly state, a high negative Replaygain value is only proof that the track is LOUD.

In fact, it is only a proof that at least 5% of the file is loud.

Quote
But still, the clippression on current pop/rock CD's is a shame.
[a href="index.php?act=findpost&pid=320728"][{POST_SNAPBACK}][/a]

Definitely. I hope it's just a "hype", although some recent singer/songwriter CD's (not mainstream) are compressed too much to my taste also.

Mastering and equal loudness

Reply #15 –
What software are Axon and Hermetically sealed using to generate their waveforms?  Doesn't look like Audition/Cool edit -- what's the signifcance of the 'lighter blue' core versus the darker blue fringe?

Mastering and equal loudness

Reply #16 –
Quote
What software are Axon and Hermetically sealed using to generate their waveforms?  Doesn't look like Audition/Cool edit -- what's the signifcance of the 'lighter blue' core versus the darker blue fringe?
[{POST_SNAPBACK}][/a]

It looks like Audacity, which you can [a href="http://audacity.sourceforge.net/]download[/url] for free.
 

Mastering and equal loudness

Reply #17 –
Ok, after reading this entire thread, I'm not clear on something. Are we arguing about intensity (i.e. SPL, voltage, pressure, whatever) here, or are we arguing about loudness (using Fletcher's definition as amended by modern understanding)?

It seems to me that people are talking past each other.
-----
J. D. (jj) Johnston