True Peak vs Sample Peak: What -1 dB Really Means for Podcasts

Every loudness guide says the same thing: keep your peaks below -1 dB. So you check the meter, see the loudest moment at -1.2 dB, and export. Then the MP3 version clips.

Nothing was wrong with the meter. It was measuring sample peak, and the -1 dB rule is about true peak. They usually sit close together, but not always, and the gap between them is exactly where a “safe” file goes over.

Samples and the Waveform Between Them

A digital audio file is a long list of samples, each one a measurement of the waveform at an instant: 44,100 or 48,000 of them a second. A sample peak meter finds the largest of those numbers. That’s the loudest point the file stores.

Your speakers and headphones don’t play a list of numbers, though. The digital-to-analogue converter rebuilds a smooth, continuous waveform that passes through every sample, and that smooth curve can rise higher between two samples than either sample is. These are called inter-sample peaks.

The textbook example is a sine wave at a quarter of the sample rate, sampled at the wrong moments. Every stored sample can sit at about 70% of the wave’s real height, so the meter reads 3 dB lower than the peak the converter will produce. Real speech is rarely that extreme, but overshoots of a fraction of a decibel are common, especially on loud, heavily limited material.

True peak is an estimate of that reconstructed waveform. A true peak meter oversamples the audio, usually four times, to work out what happens between the samples, then reports the highest point in dBTP (decibels true peak). True peak is always the same as sample peak or higher, never lower.

Why MP3 and AAC Push Peaks Higher

Lossy encoders don’t store your waveform. They store an approximation that sounds the same, discarding detail you’re unlikely to hear and filtering off the top of the spectrum. When the decoder rebuilds the audio, the waveform has a slightly different shape, and its peaks land in slightly different places, often a little higher.

So a WAV master that peaks at -0.3 dBTP can decode from MP3 at above 0 dB and clip. It’s a short, sharp distortion, often on a plosive or a laugh, and it happens on the listener’s device, after every meter in your editor has done its job. The lower the bit rate, the more the encoder changes, and the more headroom it needs.

Where -1 dBTP Comes From

Apple’s podcast audio requirements ask for true peaks no higher than -1 dB, alongside the -16 LUFS loudness target. The EBU R128 broadcast standard uses the same -1 dBTP ceiling. That 1 dB is headroom for the encoder: enough for MP3 and AAC at normal podcast bit rates to decode without going over.

Some engineers leave more, -1.5 or -2 dBTP, for low bit rate or heavily compressed episodes. Audiobooks for ACX use -3 dB. For a typical podcast at 128 kbps or above, -1 dBTP is enough.

Setting a Limiter Ceiling

The last processor on your master should be a limiter, set to a ceiling of -1 dB. What matters is whether it measures true peak.

Many limiters have a true peak or ISP (inter-sample peak) mode, sometimes labelled oversampling. Turn it on and set the ceiling to -1 dBTP. If your limiter only works on sample peaks, set it lower, to about -1.5 or -2 dB, to leave room for the overshoots it can’t see.

Then measure the export. A true peak meter on the final file is the check that counts, and most free loudness meters show it. From the command line, ffmpeg reports both loudness and true peak:

ffmpeg -i episode.mp3 -af ebur128=peak=true -f null -

A limiter can’t repair clipping that happened while you were recording. If the input was too hot and the waveform has flat tops, that’s baked in, so keep an eye on your recording levels as well.

Reading Peak Results in Audio Audit

Our Peak Volume check reads sample peak and fails any episode above -1 dBFS. It’s Required in all three built-in standards. Because true peak can sit higher than sample peak, a pass means the file clears the -1 dB rule on the stored samples. A file that only just passes, at -1.1 dBFS say, may still have true peaks above -1 dBTP. If you see a result that close to the line, lower your limiter ceiling a little and export again.

If you’d rather leave it to us, the volume normalisation in our enhancements measures true peak. It brings the episode to -16 LUFS with a single gain change, then a true-peak limiter at -1 dBTP catches anything above the ceiling. It only acts when a peak goes over, so it doesn’t change how the episode sounds otherwise. Peaks are judged together with loudness, which our LUFS article covers.

To see where your own peaks sit, upload an episode to Audio Audit. Every account gets free credits each month.

Photos courtesy of Jumping Jax