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: ISO 226:2003 based loudness correction curves (Read 13500 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

ISO 226:2003 based loudness correction curves

Please discuss [a href='index.php?showtopic=83587']here[/a].

ISO 226:2003 based loudness correction curves

Reply #1
[attachment=6066:loudness_eq.png]

Here the frequency response of a 5th-order IIR (non-linear-phase) parametric filter trying to approximate an equal-loudness correction for different input levels. Things to note:

- The 0-dB (bypass) configuration is a bit bumpy. This is a design limitation. However, it shouldn't matter too much since the max. difference of the frequency response is only 1 dB in the audible range.

- The filter was designed for 44.1 kHz sampling rate but will probably also work for 48 kHz.

Chris
If I don't reply to your reply, it means I agree with you.

ISO 226:2003 based loudness correction curves

Reply #2
The following files for Foobar's impulse convolver plugin are probably severe quality overkill for the task in hand, but they sound excellent. They are in 32 bit float format @ 44.1kHz and have been generated using iZotope Ozone's 64 bit parametric equalizer in minimum phase mode with 5-6 bands each. They track the mathematically derived ISO curves by +/- 0.2 dB at max, mostly even better. The impulses are over 6000 samples long (16384 point FFT) and fade out far below -150 dB. A regular Windows 7 capable PC doesn't even notice that the plugin is running; if you anyhow mind the unnecessary processing power, just cut as many samples as you like from the file endings.

Included are  two versions, for -10 dB (moderately quiet listening) and -20 dB (very quiet) attenuation. Each comes in two variants, that differ in how the frequency range, that is not covered by ISO 226, is handled.

The regular version leaves it alone:

[attachment=6071:83_20.png]

The alternative version interprets it as an extension of the 12.5 kHz level:

[attachment=6072:83_20_alternative.png]

Personally, I prefer the alternative version.

Attention:

  • Do not check auto-level adjust or you will get clipping.
  • Do not try to evaluate the quality by listening at regular volume levels. Music will sound boomy and muffled. Only quiet (-10dB) and very quiet (-20dB) listening levels can be much improved.
  • Do not use it on MP3 or AAC compressed files. This plugin is very likely to reveal artifacts not heard under normal listening conditions. Your results with other codecs may vary. At best use lossless files.
  • For best results set your output sample format in the system mixer to 44.1kHz @ 24 bit with a capable DAC (most modern cards and onboard HD codecs are fine).
  • For convenience, I have included two convolution files, that don't do anything but apply a linear attenuation just as a regular volume control. Use them for comparison.
  • Don't nag about the totally over spec'd EQ, it was a fun project.

[attachment=6070:iso226.zip]

Gives this a try and tell me what you think [a href='index.php?showtopic=83587']here[/a].

My reaction was 'oh, quiet bassy', at first. My ears weren't used to low volume music, that doesn't sound like low volume music. But if you compare the corrected and uncorrected low volume versions with normal listenings levels, you will notice that the corrected version sounds much more similar to the normal volume version than the thin and brittle uncorrected version.

ISO 226:2003 based loudness correction curves

Reply #3
Nifty. Thank you.

 

Re: ISO 226:2003 based loudness correction curves

Reply #4
The following files for Foobar's impulse convolver plugin are probably severe quality overkill for the task in hand, but they sound excellent. They are in 32 bit float format @ 44.1kHz and have been generated using iZotope Ozone's 64 bit parametric equalizer in minimum phase mode with 5-6 bands each. They track the mathematically derived ISO curves by +/- 0.2 dB at max, mostly even better. The impulses are over 6000 samples long (16384 point FFT) and fade out far below -150 dB. A regular Windows 7 capable PC doesn't even notice that the plugin is running; if you anyhow mind the unnecessary processing power, just cut as many samples as you like from the file endings.

Included are  two versions, for -10 dB (moderately quiet listening) and -20 dB (very quiet) attenuation. Each comes in two variants, that differ in how the frequency range, that is not covered by ISO 226, is handled.

The regular version leaves it alone:

[attachment=6071:83_20.png]

The alternative version interprets it as an extension of the 12.5 kHz level:

[attachment=6072:83_20_alternative.png]

Personally, I prefer the alternative version.

Attention:

  • Do not check auto-level adjust or you will get clipping.
  • Do not try to evaluate the quality by listening at regular volume levels. Music will sound boomy and muffled. Only quiet (-10dB) and very quiet (-20dB) listening levels can be much improved.
  • Do not use it on MP3 or AAC compressed files. This plugin is very likely to reveal artifacts not heard under normal listening conditions. Your results with other codecs may vary. At best use lossless files.
  • For best results set your output sample format in the system mixer to 44.1kHz @ 24 bit with a capable DAC (most modern cards and onboard HD codecs are fine).
  • For convenience, I have included two convolution files, that don't do anything but apply a linear attenuation just as a regular volume control. Use them for comparison.
  • Don't nag about the totally over spec'd EQ, it was a fun project.
[attachment=6070:iso226.zip]

Gives this a try and tell me what you think [a href='index.php?showtopic=83587']here[/a].

My reaction was 'oh, quiet bassy', at first. My ears weren't used to low volume music, that doesn't sound like low volume music. But if you compare the corrected and uncorrected low volume versions with normal listenings levels, you will notice that the corrected version sounds much more similar to the normal volume version than the thin and brittle uncorrected version.

This is superb. Thanks a lot.
Though I feel like that the range after 12.5 kHz should be interpolated, since I miss it even in the alternative versions.

btw: where can I get the ISO data? I would like to create my own curves. txt files for REW would be perfect