🎓 Journalism Training
Courses › Sound Theory

Lesson 3 of 5 · about 16 minutes

Decibels, Loudness and Delivery Targets

You'll be able to

  • Tell dB SPL, dBFS, LUFS and dBTP apart and say what each measures
  • Explain why peak level says nothing about how loud a file sounds
  • Name the loudness targets for Spotify, Apple Podcasts, EBU R128 and ATSC A/85 and where they come from
  • Measure integrated loudness and true peak in SoundStudio and normalize to a target

Read first: SoundStudio theory section 3, Decibels, section 5, Loudness and section 6, Clipping. The tutorial gives you the rules of thumb (+6 dB doubles the wave height, about +10 dB sounds twice as loud) and a list of targets. This lesson sorts out which decibel is which, explains where each target comes from, and gives you the sources so you can check them when they change. They do change.

Four decibels that are not the same thing

A decibel is a ratio between two levels. "Minus 14" means nothing until you know what it is minus 14 from. Here are the ones you will meet:

UnitMeasuresZero meansTypical number
dB SPLAir pressure in a real roomThe quietest sound a young ear can hearConversation about 60, a loud bar about 100
dBFSSample level inside a fileThe ceiling: the biggest number the file can storeVoice peaks around -6 to -3
LUFS (or LKFS)Perceived loudness, averagedA full-scale reference signalPodcast about -16, streaming music about -14
dBTPTrue peak, between samples tooFull scaleCeiling of -1 or -2 for delivery

dB SPL lives in the air. Nothing in your editor measures it, because the file has no idea how loud the listener's speakers are turned up. Everything inside SoundStudio is relative to full scale: dBFS, LUFS and dBTP.

Why peaks lie about loudness

A peak meter shows the single highest instant. That tells you whether a file will clip. It says almost nothing about how loud the file sounds. A snare hit and a sustained organ chord can peak at the same level while the organ sounds twice as loud, because loudness depends on how much energy keeps coming, and at which frequencies.

That is why broadcasters, streaming services and podcast apps all moved to LUFS. The measurement comes from ITU-R BS.1770, now in its fifth revision (BS.1770-5, approved November 2023). In plain terms, it works like this:

  1. It filters the audio with a curve called K-weighting, which turns down the deep bass and gives a small boost to the upper mids, roughly the way the ear does (lesson 1).
  2. It measures the energy in overlapping 400 millisecond blocks.
  3. It throws away silent blocks (below -70 LUFS) and blocks more than 10 LU quieter than the rest, so pauses and silence do not drag the average down. Then it averages what is left.

The result for a whole file is the integrated loudness. LKFS and LUFS are the same scale; American broadcast documents tend to say LKFS, everyone else says LUFS. A difference between two loudness readings is in LU, and 1 LU is the same as 1 dB.

True peak

Samples are dots. The real wave that a speaker plays is the smooth curve through those dots, and that curve can swing higher between two dots than either dot. When a file is converted to analog, or squeezed into MP3 or AAC, those hidden peaks can clip even though no sample touches 0 dBFS. BS.1770 defines a true peak measurement that oversamples the audio (four times at 48 kHz) to find them. The result is in dBTP. A delivery spec that says "-1 dBTP" means the true peak, not the sample peak.

Delivery targets as of October 2026

Every platform below publishes its number somewhere, except one. Check the source before a paid delivery, because these get revised.

Where it is goingIntegrated loudnessPeak ceilingSource
Spotify (music)Normalizes to -14 LUFS (Premium listeners can pick Loud -11 or Quiet -19)-1 dBTP; -2 dBTP if your master is louder than -14 LUFSSpotify for Artists help
YouTubeAbout -14 LUFS, measured by third parties. YouTube does not publish a figure.-1 dBTP is a safe habitThird-party measurement only
Apple PodcastsAbout -16 LKFS, plus or minus 1 dB-1 dBFS true peakApple Podcasts audio requirements
European TV and radio (EBU R128)-23.0 LUFS (plus or minus 1 LU where hitting it exactly is not practical, such as live)-1 dBTPEBU R 128, version 5, 2023
U.S. TV (ATSC A/85)-24 LKFS, plus or minus 2-2 dBTPATSC A/85, via NAB and FCC

Some notes on that table, because the details matter.

Spotify turns loud tracks down to its reference. It is upfront that a master louder than -14 LUFS gains nothing in normal playback. It just gets turned down, and if it was crushed with a limiter to get loud, it now sounds both quiet and squashed.

YouTube is the odd one out. Engineers who measure YouTube's "Stats for nerds" readout have reported for years that it turns loud uploads down toward about -14 LUFS and does not turn quiet ones up. YouTube's own help pages do not state a number. So treat -14 as a measured habit, not a spec.

For U.S. television, the law is the CALM Act. The FCC rules that took effect December 13, 2012 require TV stations, cable and satellite providers to apply ATSC's A/85 practice to commercials, so ads match the average loudness of the programs around them. The FCC is clear that this sets no absolute cap, and it covers television commercials only, not radio or internet ads. ATSC published a revised A/85 on July 15, 2026, its first full update since 2013. The core numbers stayed put (-24 LKFS with about 2 dB of measurement tolerance, true peak below -2 dBTP), and it added an annex on loudness for streaming services. The FCC rule still points to the 2013 edition. California's SB 576, effective July 1, 2026, extends the same idea to ads on streaming services for California viewers.

Radio stations set their own specs for spots. The tutorial's radio workflow (section 16) uses -24 LUFS and -2 dBTP as an example. Ask the station for its sheet. Most stations have one, and the traffic director will be glad you asked.

Hit the target with gain and gentle compression, not with a limiter crushing everything. A spot at -24 LUFS that breathes beats one smashed to -10 and then turned down by the station.

Try it: measure, then normalize

Open SoundStudio with two files: a commercially released song and a voice recording you made yourself.

  1. Load the song. Open the Enhance tab and run Analyze only. It changes nothing and reports integrated loudness (LUFS), true peak and loudness range. Write all three down.
  2. Do the same for your voice recording. Most raw home recordings land somewhere around -25 to -30 LUFS. Most modern pop masters land far louder, often -10 or above.
  3. Open the Meters tab and play each one. Watch the peak bar in dBFS. Notice that both may peak near the top while sounding very different in loudness.
  4. On the song, use Tools, Apply Gain at -6 dB, as the tutorial suggests. Run Analyze only again. The LUFS reading should drop by 6 LU and the true peak by 6 dB. Ctrl+Z to undo.
  5. On the voice, use the Enhance tab's loudness normalize with a target of -16 LUFS (the podcast number). The help explains that without Use limiter it stops short if peaks would pass -1 dB, and with Use limiter it holds peaks at -1 dB while reaching the target. Run Analyze only and confirm you landed within 1 LU. The Delivery Check in the same tab lists named targets (podcast, music streaming, U.S. and European TV, social video) plus a Custom option for a station's sheet.
  6. Normalize a copy of the same voice to -23 LUFS. Play the -16 and -23 versions back to back. Seven LU is a big gap. Now you know what a European broadcaster expects compared with a podcast app.

Record your before and after numbers in a short table. That table is exactly what a station or editor will ask you for.

Checklist

  • I know which decibel I am reading: SPL in the room, dBFS, LUFS or dBTP in the file.
  • I judge loudness with integrated LUFS, not peaks.
  • I check true peak against the spec, not sample peak.
  • I look up the current target at the source before a delivery.
  • I write "as of" next to any platform number I put in a spec sheet.

Check yourself

1. What does 0 dBFS mean?

2. Two clips both peak at -1 dBFS. One measures -20 LUFS and the other -12 LUFS. Which sounds louder?

3. What integrated loudness does EBU R128 set for European broadcast programmes?

4. Apple's podcast audio guidance asks for roughly what loudness?

5. Why measure true peak instead of only sample peak?

6. Spotify normalizes music to about -14 LUFS. What happens to a master you made at -8 LUFS?

Terms in this lesson

dB SPL
Sound pressure level in the air, relative to the threshold of hearing. What a sound level meter in the room reads.
dBFS
Decibels relative to full scale in a digital file. 0 dBFS is the maximum.
LUFS
Loudness Units relative to Full Scale. A K-weighted, gated average of loudness defined by ITU-R BS.1770. LKFS is the same measurement under another name.
LU
Loudness Unit. A difference in loudness; 1 LU equals 1 dB of level change.
True peak (dBTP)
The highest level of the reconstructed wave, including peaks that fall between samples, found by oversampling.
Loudness normalization
A platform turning each file up or down so its integrated loudness matches a reference level.

Full glossary

Sources