4th Gen. Focusrite Scarlett Series Interfaces – Solo, 2i2 and 4i4
October 19, 2023
2 Comments
Read More »
When I create an MP3 file from a WAV file using iZotope RX7 or Rogue Amoeba’s Fission (each uses the LAME MP3 encoder…ugh) my final MP3 turns out to be about 0.4 LU lower than the original WAV file. Meaning, the MP3 file is not quite as loud as the WAV file.
To offset this, I set my final WAV files to -15.6 LUFS (stereo) so that the final MP3 file is exactly -16.0 LUFS (stereo).
Here’s what I’m wondering:
Would the same effect happen with software that uses the Fraunhofer MP3 encoder, like Adobe Audition?
If you have Audition and you feel inspired to test this, please let me know your results!
And BTW, this effect DOES NOT happen when encoding to M4A files 😉
Want to receive the Daily Goody in your email, daily or weekly? Subscribe free here.
And please keep in mind, the Daily Goody is only a tiny little tip, fact or lesson everyday. Please don’t expect any of these posts to be long, earth-shattering masterpieces that instantly answer every single question you can think of and completely transform you into a world class podcast engineer. “Little by little, a little becomes a lot.”
.
DID YOU KNOW……..We exist for the purpose of helping you, so please comment below with any questions or remarks. We appreciate you listening.
Browse our online courses on podcasting and audio production to become a qualified engineer.
By continuing to use the site, you agree to the use of cookies. More information
The cookie settings on this website are set to "allow cookies" to give you the best browsing experience possible. If you continue to use this website without changing your cookie settings or you click "Accept" below then you are consenting to this.
2 Responses
I took my last podcast (39 minutes) , exported as a WAV file (48k, 24 bit, stereo) and an MP3 – (48k, 24 bit, 128 kbps CBR) and (48k 24bit, 128 kbps VBR) to see the difference.
I loaded them back into audition and checked the loudness.
WAV file – -17 LUFS. Total RMS – =21.71dB. Peak -0.0.06dB, Loudness -15.77dB, Perceived -09.90dB.
MP3 (CBR) – -17 LUFS. Total RMS -21.76dB, Peak -00.15dB, Loudness -15.84dB, perceived -09.96dB
MP3 (VBR) -17 LUFS, Total RMS -21.77dB, Peak -00.12dB, Loudness -15.89dB, Perceived -10.92dB
Same test with Fission. Opening the Wav from above and exporting
MP3 (High quality 256 kbps, VBR, 44.1 kHz, 16 bits Stereo) – -17 LUFS, Total RMS -21.74dB, Peak -00.37dB, Loudness -15.86dB, Perceived -10.12dB
MP3 (Medium quality 192 kbps, VBR 44.1 kHz, 16 bits Stereo) – -17 LUFS, Total RMS -21.74db, Peak -00.45dB, Loudness -15.90 dB, Perceived -10.12dB
MP3 (Low quality 128 kbps, VBR, 44.1 kHz, 16 bits Stereo) – -17 LUFS, Peak -21.73dB, Peak -00.12 dB, Loudness -15.78 dB, Perceived -09.94dB
Does this help?
Chris,
It’s common to notice minor differences in measured Program (Integrated) Loudness when analyzing results of various offline measurement utilities.
First, when comparing Program Loudness measurements between various audio clips, programs, etc. – the differential is described in Loudness Units (“LU”), not LUFS. LU == dB (1 LU == 1 dB, etc.).
For example: if Program (A) measures -16.0 LUFS, and Program (B) measures -14.5 LUFS … Program (B) is +1.5 LU louder than Program (A).
Secondly, read into Tolerance Margins. They are specified in all broadcast compliance specifications. Some include variables based on the attributes of the media. Bottom line is there is allowance for deviation relative to the spec. Integrated Loudness target. In fact Apple specifies a +1/-1 LU Tolerance Margin in their Podcast submission/distribution best practices documentation.
With all that said, the example that you provided is pretty much (with respect) irrelevant *from a perceptual perspective*. Typical listeners would never notice any difference. The deviation is well within any Tolerance Margin. I wouldn’t worry about it. However, there’s nothing wrong with digging deeper into these measurement anomalies.
Here’s some data:
Audition:
Test-Mix (WAV) checks in at -16.00 LUFS
Test-Mix (MP3) checks in at -16.02 LUFS
R128x (Open Source Measurement Utility)
Test-Mix (WAV) checks in at -16.00 LUFS
Test-Mix (MP3) checks in at -16.00 LUFS
Fraunhofer Pro Codec (by Sonnox)
Test-Mix (WAV) checks in at -16.00 LUFS
Test-Mix (MP3) checks in at -16.02 LUFS
Now I measure the source file and the MP3 in iZotope RX7 Advanced:
Test-Mix (WAV) checks in at -15.90 LUFS
Test-Mix (MP3) checks in at -15.90 LUFS
(MP3 attributes: 44.1kHz, 192 kbps, Stereo, CBR)
Remember this – Loudness Normalization is basically measurement, gain, and limiting (if necessary). If processed (Loudness Normalized) audio exhibits a slight (I) Loudness deviation, and ample headroom exists, a basic gain offset is certainly a viable option.
* * *
Lastly, keep this in mind: When converting “lossless” (WAV, AIFF) to “lossy” (MP3), you may notice headroom depreciation. In fact this is the reason why all specs. require Intersample Peak compliance – not Sample Peak compliance. In some cases the variation or depreciation as such may be fairly substantial. This is due to the attributes of lossy codec conversion, especially at lower data rates.
-paul.
@produceNewMedia