Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Mikkel]

Dear Matt and the rest of the JRiver team,

First of all I again (and again) have to thank you for implementing the loudness function in JRiver.
Second, I think I stumbled upon a slight misinterpretation of the loudness correction when it comes to how JRiver corrects high frequency content...

Looking at the Fletcher-Munson curve it seems that there is no change in relative loudness from 4khz and above. It is my impression from looking at the analyzer-window and the loudness function enabled, that JRiver corrects for high frequency loudness when reducing volume below the reference level. Am I correct?

If so, I think JRiver loudness needs to be revised. As per the 80phon there is indeed approx. a 12db increase in perceived loudness. Following the philosophy from movie industry the audio spectrum of the soundtrack will be reproduced correctly at 83db (or whatever the reference level). Given that the high frequency curvature remains the same across phon-levels, it is already hard-printed in the soundtrack and hence needs no correction.

Same goes for music production I suppose, although in this case, we don't really know what the reference level is supposed to be.

So to some up: JRiver corrects high-frequency content with the loudness-function when it really shouldn't.

If I got it all wrong (or just a part of it) don't hesitate to tell me. I AM and audiofool on a learning curve :-).


Best regards,
Mikkel

[Mikkel]

One correction: I said Flecher-Munson when JRiver probably follows the ISO 226:2003 specifications. The issue remains the same, though.

[Mikkel]

*bump* Are there plans on investigating/changing the loudness-correction in JRiver?


Best regards,
Mikkel

[Matt]

If you look at ISO 226:2003 curves, a similar thing happens above ~10kHz as happens below ~800Hz.  The curves diverge, just to a lesser extent at the high frequencies than the low frequencies.

Source:
http://en.wikipedia.org/wiki/Equal-loudness_contour

In other words, I think JRiver is correct to adjust high frequencies as part of Loudness.

[Mikkel]

Hi Matt,

Thank you for your response. With all respect I still think the loudness-function is wrong. I will argue my case here and then let it rest if you still disagree.

1. The "U-shaped" eq-curve created by the loudness function incorrectly goes all the way up to 20khz instead of approx 10khz
2. The "U-shaped" eq-curve shows no or only microscopic relative changes from around 6khz to 10khz
3. If there is a band that needs correction above 800hz it seems to be the band between 1.5khz and 6khz.


Best regards,
Mikkel

[AndyU]

What strikes me about the Fletcher-Munson curves, or their more recent variant the ISO 226 equal loudness contours is that there is no level at which a flat response is actually perceved to be flat, and by quite some margin. If you look at those curves, even at 100db you need to boost treble and bass levels by 10db or so for them to be perceived as equally loud as the level at 1k. So .. if you buy into using a loudness contour to adjust for low level listening, shouldn't you for the same reason buy into using one for high level listening too? ie boost treble and bass even at high levels. And does that make sense?

[Matt]

Here's the ISO 226:2003 curve, with the 80 phon curve (in green) layed over the 60 and 40 phon curve.  See how the green and red lines diverge the most at low frequencies and a little at high frequencies?

[Mikkel]

Hi Matt,

I agree that there is indeed a difference of a few db. I guess it makes sense to model. However, more importantly in comparrison to the current model in JRiver: notice how the curve tops at 10khz and not at 20khz as in the JRiver-model. In fact, above 11-12 khz it seems to decrease slightly.

@AndyU: Indeed, but imagine a symphony orchestra during a concert. The tonal balance is adjusted both through signatures like forte and piano (and all their variants) as well as the conductors interpretation. So at whatever level a concert is played at, the tonal balance will be right. Only when one departs from the reproduced level is some sort of loudness correction called for. In the movie industry soundtracks are mixed at a fixed reference level of 83dbc. So if you play back a movie at 83 dbc at home, the tonal balance will be correct. Only if you decrease the volume below reference level will you need loudness corrections.


Best regards,
Mikkel

[mwillems]

Hi Matt,

I agree that there is indeed a difference of a few db. I guess it makes sense to model. However, more importantly in comparrison to the current model in JRiver: notice how the curve tops at 10khz and not at 20khz as in the JRiver-model. In fact, above 11-12 khz it seems to decrease slightly.

The shape of the chart generally tends to decrease above 10 KHz, but the difference between the two lines is actually increasing.  For example, you can see in Matt's overlay that the 60 phon line is a few dB higher than the 80 phon line at 20 KHz.  So the compensation needs to follow it all the way up.  

[JustinM]

I'm just reading this post out of interest.
Thanks Matt; Your post clarifies how to interpete the graph / how the loudness function works.
 ((eg:  I see that when a  80 db signal is turned down to 40 db: 100hz would be eq'd aprox +10db.))

ps: love the new function  .. (tuned by ear)
  ***edited ***

[Mikkel]

The shape of the chart generally tends to decrease above 10 KHz, but the difference between the two lines is actually increasing.  For example, you can see in Matt's overlay that the 60 phon line is a few dB higher than the 80 phon line at 20 KHz.  So the compensation needs to follow it all the way up.  

I stand corrected, and apologize to Matt for incorrectly challenging the model. Sorry  .


Best regards,
Mikkel