Reply to thread

Message: [QUOTE="Queen, post: 90773, member: 27"] A loudness meter reading is built from 400-millisecond blocks that overlap by 75 percent, and it discards every block sitting below minus 70 LKFS. Then it does something less obvious. It averages what survives, drops that result by 10 LU, and throws away everything under the second line as well. The second line is where people get caught. It is not a fixed number. It is calculated from whatever audio you fed the meter, so the gate moves depending on what you played into it. Most advice about comparing your mix with a commercial record stops at three words. Match the levels. [B]Matching levels before you compare anything[/B] is correct as far as it goes, and it is where the useful part starts rather than finishes. [HEADING=2]The gate moves when the material moves[/HEADING] Loop fifteen seconds of a chorus and read integrated loudness on it. The meter builds its relative gate from those fifteen seconds alone, so the threshold ends up sitting 10 LU below a chorus average. Now read the whole track. The gate is built from verses, intros, and a quiet outro too, so it lands lower, and blocks that failed the first time around now pass. Same song. Different number. On dense, heavily limited material, the gap is tiny. On anything with real dynamic range, it is not, because the amount of audio the relative gate excludes depends entirely on how much quiet material you included in the measurement. The fix is dull but reliable. Measure the same span of both files, or measure both in full and work from the offset, and never read integrated loudness on a fifteen-second loop of your mix, then set it against a published album figure. [HEADING=2]Not every channel counts, and one does not count at all[/HEADING] Before any of that arithmetic runs, the signal passes through a two-stage K-weighting filter, a shelving stage followed by a high-pass, which pulls down what the low end contributes before a single block gets counted. The channel weighting is deliberately uneven too. Left, right, and center are each multiplied by 1.0. The surround channels are multiplied by 1.41, roughly plus 1.5 dB, on the basis that sound arriving from behind you contributes more than its level suggests. The LFE channel is not measured at all. The recommendation says so directly, stating that the low-frequency effects channel is not included in the measurement. For stereo work, none of that changes your reading. It changes what the reading means the moment you move between formats, because a stereo master and a surround master of the same program are not being judged by the same rules. There is also a limitation the document is refreshingly honest about. It describes the measured value as an estimate of subjective loudness, with uncertainty that depends on the listener, the material, and the listening conditions, and notes the algorithm is not generally suitable for estimating the loudness of pure tones. Useful for aligning two pieces of music. Not useful for settling an argument about a sine sweep. [HEADING=2]The equal-loudness curves stopped moving years ago[/HEADING] Level-matching matters in the first place because quiet and loud playback do not carry the same tonal balance, and the reference for that is a set of contours plenty of engineers still call Fletcher-Munson. The current standard is ISO 226, revised in 2023. Its values differ from the 2003 edition by at most 0.6 dB across the entire data set, and the largest change was the hearing threshold at 20 Hz, which dropped by 0.4 dB to align with the 2019 revision of ISO 389-7. Nothing is shifting under your feet. What constrains the curves is range. The 2023 edition covers preferred third-octave frequencies from 20 Hz up to 12,500 Hz, at loudness levels from 20 phon upward, and its ceiling drops to 80 phon above 5 kHz. So the air band you keep nudging sits outside the standard altogether. Anything you have read about how 16 kHz behaves at low listening levels is extrapolation rather than measurement, because the contours have never been published that high. Worth holding onto when [B][URL='https://goldmidi.com/community/threads/m-media-audio-dropped-reference-player-third-rail-tube-eq.76097/']a player with BS.1770 metering wired into its transport[/URL][/B] puts a number and a spectrum in front of you at the same time. The number is tightly specified, and the top of that spectrum is not, so align on the number, and treat the top octave as something you settle with your own monitors at a fixed listening level. [/QUOTE]

Name