BEATS WITH KEVYour master is not quieter than everyone else's. It is emptier. Loudness is not a level you turn up, it is how much of the time your track is actually near the ceiling, and that is decided long before the limiter.
This is why pushing the limiter harder does not close the gap. It makes your track louder and worse at the same time.
What "louder" actually means
Two tracks can peak at exactly the same number and sound nothing alike in volume. The one that sounds louder spends more of its time up near that peak. The one that sounds quieter has a few tall transients sticking out of a much lower average.
The distance between your peak and your average is crest factor, measured in dB. A sparse mix has a wide crest. A dense one has a narrow crest and sounds loud without ever going higher.
When your track sounds quiet next to a reference, the reference is almost never peaking higher than you. It is sitting higher, more of the time.
Why turning the limiter up makes it worse
A limiter does one thing: it stops peaks getting through. Ask it to close a 6 dB gap on its own and it has to grab every transient in the track, hard and constantly.
What you hear when that happens:
- The drums lose their front end. The click of the kick and the crack of the snare are the tallest, fastest peaks, so they are the first thing the limiter eats.
- Pumping. The limiter releases after each hit and the whole mix breathes up behind it.
- The low end goes soft. Sub energy holds the limiter down longer than anything else, so bass-heavy tracks lose the most.
The limiter is the last 1 to 2 dB. It is not the loudness stage.
So where does the loudness come from?
Clip or saturate before you limit.
Shaving a transient with a clipper takes the top off the peak instantly and without a release curve, so nothing pumps. You are removing the tall spikes that were holding your average down, which lets everything after it sit higher for free.
Ryan Schwabe states the trade exactly: "2dB of Modern Gold means the final limiter will do 2dB of less gain reduction". Worth being precise about what he means, because he is careful about it himself. Modern Gold is a saturation stage in his Gold Clip plugin, not clipping, and he sets the input so peaks reach 0.0 dBFS with essentially no clipping before he adds 1 to 2 dB of it. He will even switch the clipper off entirely when he wants a master extra clean. The transferable part is the arithmetic: every dB you take before the limiter is a dB the limiter never has to fight for, whether you take it with saturation or with a clipper.
A working order:
- Fix the mix first. Loudness that comes from arrangement costs nothing. Fewer things fighting means everything gets louder without a single plugin.
- Saturate individual elements. Drums, bass, anything with a hard transient. This adds harmonics that read as loud to the ear without adding peak level.
- Clip 1 to 2 dB. Solo the kick and snare and listen for the point where the transient goes from tight to blunt. Back off from there.
- Then limit. By now the limiter should only be doing 1 to 3 dB of gain reduction. If it is doing 8, you skipped a step.
Treat those numbers as a starting point, not a rule. A sparse beat with three elements and a dense one with thirty do not need the same amount of anything. The song tells you how much it wants.
The reason this works and the limiter alone does not: a clipper removes the peak, a limiter turns the whole track down around it.
Why your low end decides your loudness
In most popular music the loudest single frequency in the entire mix sits between 80 Hz and 100 Hz. Schwabe cites an AES study covering number one records in the US and UK from 1950 to 2010, and builds his whole low-end method around it.
That is the part worth sitting with. Your loudest content is not the vocal, the snare, or anything you are actively listening to. It is the bass region, and it is what hits the limiter first.
So when a mix will not get loud, the low end is usually the reason. Schwabe starts with a low frequency shelf between 80 Hz and 150 Hz to set the size of the low end against the rest of the mix, then rebalances the first three octaves with bells.
Note what he does not do. He is explicit that the advice to high-pass everything is wrong: he only reaches for a high-pass if there is genuine rumble or the low end feels slow, and even then he would try a second shelf first. When he does use one it is a 12 dB per octave slope. Worth knowing if you took the high-pass-everything rule from a tutorial, because he taught mixing at Drexel and this is the habit he was talking students out of.
You do not hear that sub energy on most speakers. Your limiter does.
What should I set the ceiling to?
Schwabe runs two limiters instead of one, and it is a better setup than the single -1 dB ceiling most tutorials hand you:
| Stage | Ceiling | Job |
|---|---|---|
| Loudness limiter | -0.4 to -0.3 dBFS | Does the actual work |
| Safety limiter | -0.1 dBFS | Catches anything that gets past |
The 0.3 to 0.4 dB gap between them is the point. It leaves a window where you can EQ, widen, or do dynamic processing after the loudness limiter without blowing through the final ceiling. He calls the Apple Digital Masters recommendation of a full 1 dB of headroom more than his approach needs.
One exception he makes: nearly 2 dB of headroom for anything going to social media, because lossy codecs generate their own peaks on the way through.
Worth knowing either way: a true peak reading is an estimate. True peak limiters oversample and reconstruct the waveform to guess what the converter will do on the way to analogue. It is a good guess, not a guarantee.
I bounce at -9 to -11 LUFS integrated, which is louder than the -14 that gets repeated everywhere. That number is a streaming normalization target, not a mastering target. Spotify and Apple will turn a -9 master down and it sounds exactly as intended when they do. What it also does is hold up everywhere that does not normalize: a SoundCloud upload, a link you send a client, a beat playing off a phone speaker in a car.
Master to -14 and the platform leaves you alone. Master to -9 and the platform turns you down. Neither is a penalty. The mistake is getting to -9 by leaning on the limiter, because that is where the damage happens, not at the number.
How do I check if it is density or level?
Bounce the track and run it through Signal Check, then read the crest factor rather than the LUFS.
LUFS tells you how loud you are. Crest tells you why. It is your true peak minus your average level, and the read grades it: under 7 dB is compressed, 7 to 12 dB is healthy, which is where most commercial masters actually sit.
That range is the whole test. If you bounce at -9 LUFS and crest comes back at 8 dB, the loudness is real and you got it from arrangement and clipping. If you bounce at -9 and crest comes back at 4, you got there by crushing, and the read will say VERY COMPRESSED because it is.
Run it before and after you add the clipper. Watch LUFS climb while crest holds. That is the whole trick, and it is the one thing more limiting can never do for you.
The short version
Loudness is density, not level. The gap between your master and a reference is time spent near the ceiling, not peak height. Fix the arrangement, control the 80 Hz to 100 Hz region that is quietly hitting your limiter hardest, take 1 to 2 dB with a clipper, and leave the limiter 1 to 3 dB of work. Ceiling at -0.4 to -0.3 dBFS with a -0.1 safety limiter behind it. Then check crest factor rather than LUFS: if it is still in the 7 to 12 dB band, the loudness you gained is real.
Credit where it is due
The gain-staging arithmetic, the two-limiter ceiling and the low-end method here come from Ryan Schwabe, a Grammy-nominated mixing and mastering engineer who also builds the Gold Clip and Orange Clip plugins at Schwabe Digital. I met him at NAMM 2024 and had a genuinely great conversation. His writing at Schwabe Digital University is some of the most specific, least hand-wavy material on this subject anywhere, and it is free.