Why do small amounts of peak headroom and full scale lead to distortion during playback? What are interleaved sample overs?
Interleaved sample overs are contiguous virtual overs resulting from adjacent samples and from the "soft" waveform occurring after D/A converions (while monitoring). Interleaved sample overs don't actually exist as overs as a digital value, since the digital limit is 0dB and no more. Interleaved sample overs occur during oversampling or D/A conversion. When you have two contiguous full scale samples surrounded by lower-level samples, then an over has occured. But if you oversample (i.e. increase the sample rate by a factor of 2 or more), you have five contiguous full scale samples with 2x oversampling. You now have an over according to the mastering rule "3 contiguous full scale samples is considered an over" because on the analog side, levels over the 0dB limit occur which cannot be properly processed and therefore lead to distortion. Brickwall limiters use this technique by multiplying the sample frequency before limiting and then eliminating any interleaved sample overs before they actually occur.
Interleaved sample overs are often referred to as "inter sample overloads." Both the Loudness War and ignorance on the part of certain mastering engineers have led to a situation where a shockingly high percentage of current releases contain a large number of contiguous full scale samples, creating distortion. It is not rare to have 50 full scale samples occurring in a row, with this happening thousands of times on a single recording. This kind of product – depending on the quality of the the converter of the player – leads to significant distortion and is, in our opinion, unprofessional and simply faulty.
Adhering to the Pleasurize Music Standards (DR14 and 0.3dB minimum headroom after brickwall limiting) completely eliminates this problem.
Here are only two contiguous full scale (0dB) samples. The "analog" line shows the area which exceeds 0dB.
By oversampling, there are now 5 full scale samples.
The brickwall limiter "sees" the problem and uses complex algorithms to recalculate the sample values so that distortion is avoided.
The result is that there is no more distortion because there are no longer any interleaved sample overs.
Here you can see the catastrophic result present in current releases:
Passage from "Bleed Like Me" (Garbage) with a series of 10 contiguous full scale samples. This guarantees distortion.