Castcore
Home Features Pricing Technical guidelines The Castcore index Migrate from RSS Help Start your show

Technical

True Peak limiting

What is the meaning of a True Peak limiter?

What is digital audio

Digital audio consists of a vast amount of little “images” of the audio, called samples. On a song stored on a CD, there are 44,100 of these samples in one second. In this format, each sample can have 65,536 different values. Digital audio relies on samples to represent fluctuation in continuous phenomena, such as sound traveling through the air, or analog audio being transferred through electrical components.

We don’t see it all

The issue, however, is that we’re blind to what happens between these samples. One sample may, on the digital full scale, have a value of -0 dB. We might think that alright, this is within limits, we let it pass. The next sample also reads the same -0 dB. Now we have two values that by themselves show no peaking at all. However, between these points, the audio wave has had a chance to exceed the threshold of -0 dB since we only see what happens every 44100th of a second and we’re blind to anything that happens in between.

The solution is to use magic magnifying glass

Or not. The actual solution is a process called oversampling using a special finite impulse response filter, which means we take more samples from the audio and use advanced mathematics to reconstruct the wave.

If we previously had 44,100 samples per second, we might now sample it 192,000 per second. Some tools sample even higher, such as the TNT Plus at a whopping rate of ~0.8 MHz. Oversampling the audio gives us better visibility into what happens between the samples, and lets us reconstruct the wave from its digital representation.

Why we use -5 dBTP specifically

We would love to have a full dynamic range for your content. Some distributors, however, don’t follow any industry standards as to what comes to loudness. Our limits of -18 LUFS and -5 dBTP are designed specifically to process your audio once, and deliver it to all distributors in such shape it allows maximum normalization on the distributor’s end.

Spotify loves to make things loud. Their loudness target is -14 LUFS. If we delivered audio following the broadcast standard of -23 LUFS, with peaks at -1 dBTP, Spotify would want to boost the audio file by 9 decibels, but they can’t because the peaks would clip (since the peaks would now sit at +8 dB).

Our normalization pipeline guarantees your audio comes out at -18 LUFS, -5 dBTP and <7 LRA. This is fine as-is for most platforms, and it allows Spotify to crank the levels up by 4 dB to reach their own targets, and the true peaks still sit nicely at -1 dBTP.