Read the DR value of a track. It is the measurement behind the loudness-war database and the TT DR meter: how far the peaks stand above the loudest passages. A master that has been limited flat scores low because its peaks have been cut down to its body; an unmastered mix scores high because its peaks still stand.
The reading
The large number is the DR as the meter prints it, rounded to an integer, with the exact value in decibels beside it. Below it, the left channel’s parts: the RMS of the loudest fifth of blocks, the second-highest sample peak the DR is measured against, the sample peak itself, the RMS of the whole file and the crest factor, which is the peak over that whole-file RMS. A stereo file shows both channels’ DR.
The algorithm
The file is cut into 3-second blocks. Each block’s RMS is taken, with 3 dB added so a full-scale sine reads 0 dB rather than -3. The blocks are sorted and the loudest 20 percent are averaged in power. That average is subtracted from the channel’s second-highest sample peak, and the result is the channel’s DR. A trailing part block counts when it is at least a second long.
The plot
One bar per block: the solid bar is the block’s level, the faint bar behind it the block’s peak, and the gap between them is the dynamic range of that block. The dashed line is the loudest fifth, the level the reading is measured from.
DR against LUFS
Integrated loudness, loudness range and true peak are measured in the same pass and shown for comparison. Two masters at the same -14 LUFS can sit at DR6 and DR12: the first was limited to get there, the second was turned down. Loudness says how loud; DR says what was done to the peaks on the way.
Limits
The measurement is defined on sample peaks, not true peaks, as the reference meter is. A file shorter than a second has no block to measure. Heavy clipping lowers the second peak too, so a square-waved master reads lower still, which is the honest answer.