How to put reverb on bass without mud

The bass that lives outside the room

The drums are in a room, the keys have a plate, the vocal has a hall, and the bass is a dry DI that seems to sit in front of the speakers. In a dense track that can go unnoticed. In a sparse ballad or a slow jazz tune, the bass sounds like it was recorded somewhere else. Then you remember the rule, no reverb on bass, and leave it.

The rule protects you from real problems. Put a full-range hall on the bass and the low end turns to porridge. But the problems sit in specific places, and once you know where, you can give the bass a room and keep the low end clean.

Why it matters

Three ways a bass reverb goes wrong

The first is overlap. A long tail is still ringing when the next note starts, and two low notes at once do not separate the way two high notes do. The second is colouration. The early reflections are short delayed copies of the bass, and a delayed copy added to the dry signal cancels some frequencies and boosts others. The third is width. A stereo reverb spreads its output across both channels, and low end in the sides is the part that suffers when the mix is folded to mono, as the lesson on stereo low end explains.

Each of these has its own fix, and none of them requires the bass to stay dry.

The science

Critical bands, combs and harmonics

The overlap problem comes from how the ear splits sound into bands. Below about 500 Hz, a critical band is roughly 100 Hz wide (Fastl and Zwicker, 2007). Two tones inside one band are hard to hear as two clear pitches. A few hertz apart they beat at the difference between their frequencies, and further apart the beating speeds up into roughness. A1 is 55 Hz and B1 is 61.7 Hz, so a tail of A1 under a new B1 beats about 7 times a second.

How loud the old tail still is depends on the decay time $T$ and the note length $d$. A tail falls 60 dB in $T$ seconds, so by the next note it is down:

For eighth notes at 120 BPM, $d$ is 0.25 s. A 2.5 s hall is only 6 dB down when the next note arrives. A 0.5 s room is 30 dB down.

The colouration problem is a comb filter. Add a copy delayed by $\tau$ at relative level $a$, and the response is $|1 + a\,e^{-j 2\pi f \tau}|$ (Zölzer, 2011). Notches fall where the copy arrives half a cycle late, at $f = 1/(2\tau)$ and its odd multiples. For a 5 ms reflection that is 100 Hz, 300 Hz and 500 Hz. At half level the notches are $20 \log_{10}(1 - 0.5) \approx -6$ dB deep and the peaks $20 \log_{10}(1.5) \approx +3.5$ dB high.

A real reverb has many reflections at many delays, so instead of one tidy comb you get a ragged response, and with a moving bass line each note lands somewhere different on it. That is part of why reverbed bass sounds uneven from note to note. The demo uses an equal-level copy, which makes the effect much stronger than one reflection would. Try a few milliseconds of delay and listen to the low end thin out.

The fix for both problems is the same: keep the fundamental out of the reverb. For a bass with a saw-like spectrum, where harmonic $n$ has amplitude $1/n$, the first two harmonics hold about 76% of the power. A high-pass on the return removes most of what overlaps and combs, and the reverb still hears the upper harmonics, the string noise and the attack. Those are what tell the ear the bass is in the same room as the drums.