why reverb sounds metallic
the metallic ring in a reverb tail is not a mystery: it is resonance from repeating delay patterns. where the ring comes from, why cheap reverbs have it, and how modern designs remove it.
the tail with a pitch
you know the sound. you put a reverb on a snare, and instead of a wash of air you get a boing: a metallic, pipe-like ring with an identifiable pitch hanging off the back of every hit. it is the signature of a cheap reverb, the reason “digital” used to be an insult in this corner of audio, and the first thing your ear checks now whether you asked it to or not.
here is the thing: metallic is not a vague texture. it is a measurable, specific failure. a good reverb tail is colorless: energy spread evenly across frequency, decaying evenly, no frequency favored. a metallic tail has favorites. a handful of frequencies ring louder and longer than their neighbours, and your ear does what it always does with a stack of related ringing frequencies: it hears a pitch.
a tail with a pitch is no longer a space. it is an instrument playing along with your mix, in a key nobody chose.
key takeaway
metallic reverb is resonance. somewhere in the algorithm, the same reflection pattern is repeating, and repetition at a fixed interval concentrates energy at a fixed set of frequencies. every fix for metallic reverb is a way of breaking that repetition.
where the ring comes from
an algorithmic reverb is built from delay lines with feedback: sound goes in, comes back a little later and a little quieter, goes around again. that loop is the engine of the whole effect, and it is also the source of the problem, because a delay with feedback is a comb filter, and a comb filter is a tuned resonator.
feed a delay of 10 milliseconds back into itself and it reinforces 100 Hz and every multiple of it: 200, 300, 400, up the spectrum like a comb’s teeth. that harmonic series is a pitch. one delay line, audibly, is not a room. it is a tube.
real reverbs use several delay lines to break this up, and this is where designs separate. schroeder knew it in 1962: he chose his comb delays to avoid simple ratios, precisely so their resonances would interleave instead of stack.[^1] if two lines share a common factor, their repeats line up regularly, reinforcing the same frequencies. this is why serious designs use mutually prime delay lengths: the repeats coincide as rarely as the math allows.
but spacing the resonances is only half the job. the other half is having enough of them.
mode density: the real currency
a reverberator’s resonances (modes) are spread across the spectrum roughly one per 1/τ Hz per delay line, where τ is the line’s delay. the total delay in the network sets how many modes you get. when modes are dense enough, they overlap and your ear hears a smooth wash; when they are sparse, individual modes stick out and ring. this is exactly the room-modes story from the resonance path, transplanted: a small bathroom has sparse, audible modes and sounds boxy; a concert hall has so many overlapping modes that none is audible alone. a reverb with too little total delay is, acoustically, a very small weird room.
a second, sneakier source: resonant elements inside the feedback loop. designers put allpass diffusers inside the loop to multiply echo density, and an allpass with a high feedback coefficient is itself a high-Q resonator. on paper an allpass is colorless: its long-term frequency response is flat. but a tail is not long-term; it is transient, and during the decay a hot allpass rings at its own frequencies inside every pass around the loop. flat on the analyzer, metallic on the snare. the analyzer is averaging over time; your ear is not. (i have lost real days of my life to that specific lie.)
how modern reverbs kill the ring
four tools, usually in combination.
more lines, longer lines. raise the mode density until no single resonance is audible alone. brute force, but it is the foundation: nothing else works if the underlying mode count is starved.
mutually prime, well-spread delays. spread the modes as evenly as the line count allows. no common factors, no clustering.
dense diffusion, kept polite. allpass diffusers multiply echo density so the tail smooths out fast, with feedback coefficients kept moderate so the diffusers do not become resonators themselves. density is bought with structure (nesting, cascading), not with hotter feedback.
modulation. the heavyweight. slowly drift the delay lengths and no mode stays at a fixed frequency long enough to accumulate energy. the ring smears into the surrounding spectrum and the tail comes alive in the way people describe as “expensive.” the trade is pitch wobble: modulate too fast or too deep and sustained material starts to detune audibly, so good designs use slow rates, shallow depths, and different rates per line so the drifts never synchronize. research has even produced ways to vary the network itself smoothly over time without touching pitch at all.[^2]
none of these is exotic anymore. the difference between a reverb that boings and one that breathes is not a secret ingredient. it is whether the designer did all four of these jobs, and how carefully.
note
diagnostic, two minutes: send a single dry snare hit into the reverb, mute the dry path, and listen to the tail alone at a healthy volume. then hum. if you can hum along with the tail, it has a pitch, and you have found your metallic culprit. repeat at different decay settings: rings that get worse as decay gets longer are feedback-loop resonances by definition, because the loop is the only part that repeats.
killing this ring is most of what i have spent the last few months on. OPEN is the algorithmic reverb that came out of it, built for synths and drum machines rather than another plate emulation, and it opens for beta soon. if you want first access, tick “i want to beta test future plugins” on the signup at /#signup.
frequently asked questions
frequently asked questions
why does my reverb sound metallic?
a metallic tail means the reverb is resonating at a small set of related frequencies instead of decaying evenly across the spectrum. the usual cause is repetition: too few delay lines, delay lengths that share common factors, or no modulation, so the same reflection pattern recycles and stacks energy at the same frequencies. your ear reads that stack as a pitch, and a pitched tail sounds like a struck pipe.
how do i fix a metallic reverb in my mix?
if you cannot change the algorithm: add slow modulation or chorus on the reverb return if the plugin allows it, shorten the decay so the ring has less time to build, EQ out the strongest ringing frequency on the return, or layer a second, denser reverb to mask the repeats. if the plugin exposes a diffusion or modulation control, raise it. if none of that works, the honest fix is a better-designed reverb.
what is echo density in a reverb?
echo density is how many reflections arrive per second. in a real room it grows with the square of time: a few sparse early reflections, then thousands per second within a few hundred milliseconds. a reverb whose echo density grows too slowly sounds grainy and fluttery early in the tail, and its sparse repeats are exactly what stack into metallic resonances later.
why do reverbs use prime number delay lengths?
if two delay lines share a common factor, their repeats line up regularly and reinforce the same frequencies, which concentrates energy into an audible ring. delay lengths that are mutually prime line up as rarely as possible, spreading the repeats (and the resonances) as evenly as the line count allows. it is the same reason you avoid speaker placement at exact fractions of a room dimension.
does modulation remove metallic ringing?
mostly, yes, and it is the single most effective tool. slowly varying the delay lengths means no resonant frequency stays fixed long enough for energy to accumulate at it; the ring is smeared into the surrounding spectrum. the trade-off is pitch wobble: too much modulation and sustained notes through the reverb start to shimmer or detune, so well-designed reverbs modulate as slowly and as little as they can get away with.
references
a note from the developer
i spent most of this spring inside exactly this problem, building and rebuilding feedback delay networks and listening to the tails until i could name which delay line was ringing. the humiliating discovery was the allpass one: i had a diffuser running hot inside a feedback loop, the frequency-response plot was ruler flat, and the tail still rang like a radiator. every measurement said colorless. my ears said pipe. the math was averaging over infinite time and my ears were living in the first 500 milliseconds, and the ears were right.
that is the lesson i would pass on, more than any of the four fixes: when a reverb sounds metallic, trust the boing over the plot. the plot is answering a different question.
if you have a metallic-reverb horror story, or a tail you cannot diagnose, send it over. jonas@kernaudio.io. i read every email.