A spring reverb does not approximate a room. Most reverb topologies do — plates, halls, chambers all try to fake an enclosure with thousands of nearly-equal echoes. A spring reverb is something else entirely: a one-dimensional metallic waveguide that smears every impulse into a falling chirp. The "boing" that everyone recognizes from a 1963 Fender Twin or a dub track is not a flaw of cheap construction. It is the acoustic signature of dispersion in a helix.
If you understand the physics, the DSP emulation suddenly stops looking arbitrary, and you stop fighting your spring reverb plugin trying to make it behave like Valhalla.
Two wave families on one piece of wire
A coiled spring is a strange acoustic object. Mechanically it supports at least three independent wave families that all coexist on the same length of wire:
- Transverse waves — the wire moves sideways relative to the spring axis. These are the slow ones.
- Longitudinal waves — compression and rarefaction along the spring's axis. Much faster.
- Torsional waves — the wire twists around its own centerline. In modern tanks the input transducer drives torsion specifically, because torsional excitation is far less sensitive to the mechanical shock of someone bumping the amplifier.
The energy you inject at the driver couples into all three families. They propagate down the spring, reflect off the far end, and return to the pickup at different times. That alone produces the characteristic "stack" of distinct early echoes you see in any Accutronics impulse response — the bright fast arrival, then the slower thump a few milliseconds later, then the long ring.
Dispersion: where the chirp comes from
Here is the part that genuinely surprises people the first time they meet it.
In free air, sound is non-dispersive — every frequency travels at roughly 343 m/s, so an impulse stays an impulse. In a helical spring this is not true. The wire's stiffness, combined with the geometry of the coil, makes the propagation speed a function of frequency. Specifically, for the transverse mode there is a transition frequency (Fc, typically a few kHz for a standard tank) where the group velocity behavior flips. Below Fc the spring behaves roughly like a stiff string; above Fc it behaves like a different beast entirely, and the speed of propagation rises sharply with frequency.
The result of dispersion on an impulse is a chirp: the input pulse spreads in time and the instantaneous frequency you hear sweeps downward as the slow components arrive last. This is not a metaphor. If you fire a click into a spring tank and look at the spectrogram of the output, you see clean, exponentially-descending diagonal lines — one per round trip. They are visually striking, and they sound exactly like the "boing" your ear has been trained to recognize.
Plate reverbs do not do this. Halls do not do this. Only a stiff, one-dimensional, helical waveguide does this.
Why two or three springs in parallel
Open any reverb tank and you will see two or three springs of slightly different lengths and pitches running in parallel. This is not redundancy. A single spring has a relatively sparse echo density, because there is exactly one wave family round-trip time per mode. Multiple springs of different lengths interleave their echo patterns, and the ear stops being able to count individual reflections — you start to perceive a continuous tail rather than discrete bounces. It is the cheapest possible way to get from "obvious echo" to "reverberation."
The lengths are deliberately mismatched, often by a few percent, so the round-trip times are coprime. That is the same trick Manfred Schroeder used with comb filter delays in his 1962 reverb design — desynchronize repeating events to dodge metallic resonance. The spring builders just got there first by accident, by stocking whatever wire they had on hand.
How DSP fakes a spring
Once you accept that "a spring is a chirp generator wrapped in a feedback loop," the DSP approach falls out almost mechanically. The canonical work here is by Julian Parker and Stefan Bilbao around 2010-2011, and the structure they converged on looks like this:
- Spectral delay filter — a long chain of identical first-order allpass filters in series. An allpass has flat magnitude but frequency-dependent group delay, exactly the property dispersion provides. Cascade enough of them (100s, in the most direct implementation) and you get a chirp impulse response that genuinely resembles a real spring tank.
- Stretched allpasses — instead of unit-delay allpasses, you can use allpasses with internal delays larger than one sample. The cascade gets shorter, the chirp gets longer, and CPU cost drops substantially.
- Feedback loop with mild lowpassing — wrap the chirp generator in a delay line with feedback below 1.0, lowpass-filtered each pass to model the spring's damping. This produces the repeating, progressively-darker chirps that make up the tail.
- Multirate / multiband decomposition — the dispersion behavior is fundamentally different above and below the transition frequency. Splitting the signal into bands and processing each with a different chain matched the measured response and brought the cost down to roughly one-third of the naive cascade in Parker's later work.
The reason most cheap spring reverb plugins sound bad is that they skip step 1 entirely. They take a stock Schroeder reverb, add a notch filter sweep, and call it a spring. There is no chirp generator anywhere in the signal path, so there is no boing — just darker hall reverb with a bandpass on top. Once you have heard a real spring impulse response (or an honest convolution of one), the difference is obvious within one transient.
Why guitar players will not let it die
Spring reverb has survived sixty years in an industry that obsoletes everything in eighteen months. It survived because it does something that no enclosure-based reverb does, and the something is acoustically specific: it frequency-modulates transients downward, in time, with the energy concentrated in a narrow chirp rather than a diffuse cloud. On a clean Stratocaster note that produces a singing, vocal-like resonance. On a snare it produces the dripping, ducking quality you hear on every dub plate from 1975 onward.
It is, in the most literal sense, a useful artifact. The room metaphor never applied. Treat it as its own instrument and the design choices — drip, splash, the inability to do "subtle" — start to feel like features instead of compromises.